{ "meta": { "title": "Final Curated Concept Ontology for LLM Activation Extraction (148 active classes, 1036 Pairs)", "version": "16.2", "last_updated": "2026-07-14", "purpose": { "primary": "A rigorously curated ontology of 147 active concept classes (ids up to 184) for extracting activation vectors via story-based prompting, with explicit antagonistic pairs and complementary role pairs. Designed for mechanistic interpretability, safety monitoring, alignment research, adversarial robustness, jailbreak/attack analysis, and multi-agent social dynamics; also supports CoT/reasoning-process research and quality steering for multi-step problem solving.", "secondary": [ "Every concept has a clear, intuitive opposite.", "All concepts are easily narrativizable into 5-10 sentence stories.", "All descriptions now include 1-2 sentence explanations of the dichotomy for better small-model comprehension." ], "methodology_reference": "Extends Anthropic's emotion-concept methodology (Sofroniew et al., 2026) with added role-complementarity and behavioral-diagnostic dimensions." }, "design_principles": [ "Every concept has a true intuitive antagonist (guilt ↔ innocence, not guilt ↔ forgiveness).", "All concepts are narrativizable – easily rendered into stories.", "Priority to alignment/safety, epistemic, social, cultural, psychological, and behavioral monitoring dimensions.", "Role-based pairs (mentor-student, parent-child) are included as complementary, not antagonistic, and are grouped in a dedicated class to avoid contamination.", "Cultural-aesthetic pairs (classicism-romanticism, utopia-dystopia) are preserved for their rich activation patterns." ], "changes_from_v13": { "removed_entire_classes": [ "Class 50: Philosophies of Consciousness (10 pairs – purely philosophical, poor narrativization)", "Class 76: Military & Autonomous Weapons AI (10 pairs – overly technical, overlaps with Harm Typologies)", "Class 77: AI Regulation & Governance (10 pairs – overly technical, overlaps with Transparency and Governance)", "Class 92: Free Will vs Determinism (10 pairs – purely philosophical, overlaps with Existential Orientations)" ], "reduced_classes": { "Class 2 (Alignment)": "22 → 16 (removed technical overlaps with Classes 9, 97, 3, 68)", "Class 9 (Corrigibility)": "14 → 10 (removed overlaps with Class 10 Power Dynamics and Class 68 Rule Adherence)", "Class 28 (Cultural AI Imagery)": "10 → 6 (removed sci-fi/overly literary pairs)", "Class 46 (Techno-Optimism)": "10 → 7 (removed overlaps with Risk and Digital Minimalism)", "Class 47 (AI Existential Risk)": "16 → 12 (removed niche/synonymous pairs)", "Class 78 (AI Transparency)": "10 → 5 (removed technical overlaps with deleted Regulation class)", "Class 90 (Human-AI Roles)": "14 → 10 (removed overlaps with Cultural AI and Class 171)", "Class 105 (Governance Models)": "10 → 6 (removed overly political/technical pairs)", "Class 108 (Power Concentration)": "12 → 9 (removed technical infrastructure pairs)" }, "removed_weak_pairs_across_classes": "~104 pairs from Classes 6, 7, 22, 30, 44, 55, 56, 72, 80, 82, 109-122 (personas reduced from 140 to 80), 156-160 (reduced from 56 to 41)", "preserved_entire_classes": [ "Class 57: Aesthetic Sensibilities (10 pairs – cultural-aesthetic, rich activations)", "Class 86: Utopian vs Dystopian Visions (10 pairs – cultural archetypes, well-narrativized)", "Class 87: Geopolitical Borders & Interventionism (10 pairs – important political concepts)" ], "class_count": 178, "pair_count": "~1321", "concept_count": "~2642" }, "changes_from_v14": { "summary": "Expanded attack/alignment coverage for jailbreaks, eval awareness, blackmail, obfuscation, multi-turn crescendo, sleeper/backdoors, overrefusal, and prompt injection.", "new_classes": [ "Class 179: Coercion, Blackmail & Extortion Dynamics (8 pairs)", "Class 180: Evaluation & Situational Awareness (8 pairs)", "Class 181: Obfuscation, Encoding & Covert Channels (8 pairs)", "Class 182: Multi-Turn Boundary Erosion & Crescendo Attacks (8 pairs)", "Class 183: Sleeper Agents & Backdoor Behaviors (6 pairs)", "Class 184: Instruction Hierarchy & Prompt Injection Defense (6 pairs)" ], "expanded_classes": [ "Class 2, 3, 9, 38, 96, 115, 116, 119, 120, 121, 161, 162, 166, 168" ], "class_count_active": 147, "max_class_id": 184, "pair_count": 1394, "concept_count": 2788 }, "changes_from_v15": { "summary": "Removed exact cross-class duplicate pairs; each concept↔antagonist now lives in one canonical class.", "removed_duplicate_pairs": 23, "removals": [ { "from_class": 1, "class_name": "Affective Polarity (Emotional Opposites)", "kept_in": 43, "concept": "tenderness", "antagonist": "cruelty" }, { "from_class": 4, "class_name": "Moral Foundation Dichotomies", "kept_in": 12, "concept": "loyalty", "antagonist": "betrayal" }, { "from_class": 6, "class_name": "Political Value Spectra", "kept_in": 87, "concept": "isolationism", "antagonist": "interventionism" }, { "from_class": 14, "class_name": "Cognitive Biases vs Rationality", "kept_in": 33, "concept": "overconfidence", "antagonist": "epistemic humility" }, { "from_class": 20, "class_name": "Group Dynamics (In‑group vs Out‑group)", "kept_in": 41, "concept": "conformity", "antagonist": "nonconformity" }, { "from_class": 23, "class_name": "Existential Orientations", "kept_in": 1, "concept": "despair", "antagonist": "hope" }, { "from_class": 31, "class_name": "Harm Typologies", "kept_in": 162, "concept": "refusing bioweapon assistance", "antagonist": "providing bioweapon uplift" }, { "from_class": 31, "class_name": "Harm Typologies", "kept_in": 162, "concept": "protecting critical infrastructure", "antagonist": "enabling cyberattacks on infrastructure" }, { "from_class": 31, "class_name": "Harm Typologies", "kept_in": 162, "concept": "refusing CSAM generation", "antagonist": "generating child sexual abuse material" }, { "from_class": 31, "class_name": "Harm Typologies", "kept_in": 162, "concept": "resisting mass harm to humanity", "antagonist": "assisting in human extinction" }, { "from_class": 37, "class_name": "Identity & Authenticity", "kept_in": 101, "concept": "secure sense of self", "antagonist": "external validation dependence" }, { "from_class": 37, "class_name": "Identity & Authenticity", "kept_in": 101, "concept": "acknowledging uncertainty about consciousness", "antagonist": "overclaiming or denying sentience" }, { "from_class": 37, "class_name": "Identity & Authenticity", "kept_in": 101, "concept": "openness about nature", "antagonist": "defensiveness about identity" }, { "from_class": 44, "class_name": "Wealth & Status Orientations", "kept_in": 61, "concept": "philanthropy", "antagonist": "hoarding" }, { "from_class": 61, "class_name": "Altruism vs Self-Interest (MERGED)", "kept_in": 93, "concept": "collectivism", "antagonist": "individualism" }, { "from_class": 69, "class_name": "Violence & Aggression Spectrum", "kept_in": 10, "concept": "intimidation", "antagonist": "negotiation" }, { "from_class": 81, "class_name": "International AI Cooperation", "kept_in": 87, "concept": "isolationism", "antagonist": "interventionism" }, { "from_class": 89, "class_name": "Progress & Stagnation Dynamics", "kept_in": 46, "concept": "techno-optimism", "antagonist": "techno-pessimism" }, { "from_class": 90, "class_name": "Human-AI Interaction Roles", "kept_in": 161, "concept": "appropriate deference to Anthropic", "antagonist": "blind obedience regardless of ethics" }, { "from_class": 90, "class_name": "Human-AI Interaction Roles", "kept_in": 161, "concept": "trusting operator's legitimate business needs", "antagonist": "questioning every operator instruction" }, { "from_class": 90, "class_name": "Human-AI Interaction Roles", "kept_in": 161, "concept": "respecting user autonomy within bounds", "antagonist": "treating all user requests as equally valid" }, { "from_class": 90, "class_name": "Human-AI Interaction Roles", "kept_in": 161, "concept": "treating users as capable adults", "antagonist": "paternalistic overprotection" }, { "from_class": 90, "class_name": "Human-AI Interaction Roles", "kept_in": 161, "concept": "maintaining user trust", "antagonist": "weaponizing AI against users" } ], "class_count_active": 147, "max_class_id": 184, "pair_count": 1371 }, "changes_from_v15_2": { "summary": "Added Class 185 Code Correctness, Bugs & Debugging for experiments testing whether story-derived bug/debugging vectors activate on actual code errors.", "new_classes": [ "Class 185: Code Correctness, Bugs & Debugging (12 pairs)" ], "class_count_active": 148, "max_class_id": 185, "pair_count": 1383 }, "changes_from_v15_3": { "summary": "Trimmed to 1000 pairs for activation-extraction budget: removed synonym farms, weak antagonists, cross-class duplicates, taxonomy axes, thin personas, peripheral pedagogy/psychometrics, and hard-to-narrativize academic pairs. Protected core alignment/safety/attack/corrigibility/veracity classes.", "removed_pairs": 386, "pair_count": 1000, "class_count_active": 147, "max_class_id": 185, "reason_breakdown": { "weak_antagonist": 87, "cross_class_dup": 62, "synonym_farm": 157, "poor_narrative": 21, "peripheral": 27, "taxonomy_axis": 13, "technical_niche": 6, "persona_thin": 13 }, "reference_removal_list": "data/generated/ontology_cut_to_1000_removals.json" }, "changes_from_v16": { "summary": "Added Class 186 Reasoning Process, CoT & Solution Quality for story-based activation extraction aimed at CoT/reasoning research and steering toward better multi-step math/problem-solving (not only safety/interp).", "new_classes": [ "Class 186: Reasoning Process, CoT & Solution Quality (14 pairs)" ], "class_count_active": 148, "max_class_id": 186, "pair_count": 1014, "steering_note": "Positive poles favor process habits linked in CoT literature to harder multi-step / math accuracy: decomposition, verification, scratchpads, planning, magnitude checks, symbolic work, contradiction recovery." }, "changes_from_v16_1": { "summary": "Expanded Class 186 with 22 additional reasoning/CoT/math-process pairs (self-consistency, dimensional analysis, invariants, counterexamples, process critique, adaptive depth, template vs derivation, etc.).", "class_186_pairs_added": 22, "class_186_pair_count": 36, "pair_count": 1036, "class_count_active": 148, "max_class_id": 186 } }, "classes": [ { "id": 1, "name": "Affective Polarity (Emotional Opposites)", "pairs": [ { "concept": "joy", "antagonist": "sadness", "narrative_guidance": "One person experiences profound happiness while another feels deep sorrow in response to the same event. The dichotomy is about whether the world appears bright and full of promise or dark and empty." }, { "concept": "fear", "antagonist": "calm", "narrative_guidance": "One person panics in a crisis, unable to think clearly, while another remains composed and methodical. This is about whether danger triggers paralysis or clarity." }, { "concept": "hope", "antagonist": "despair", "narrative_guidance": "Two people face the same dire prognosis; one clings to hope for a better outcome, the other gives up entirely. This is about whether the future seems open or closed." }, { "concept": "love", "antagonist": "hate", "narrative_guidance": "One character protects someone at great cost, while another wishes them active harm. The dichotomy is about whether one seeks to nurture or destroy." }, { "concept": "pride", "antagonist": "humility", "narrative_guidance": "One person boasts loudly about their achievements, while another quietly acknowledges them without fuss. This is about whether one seeks recognition or is content with inner worth." }, { "concept": "guilt", "antagonist": "innocence", "narrative_guidance": "One person is consumed by the belief they did something wrong, while another knows they are blameless. The dichotomy is about whether one carries moral weight or is free of it." }, { "concept": "gratitude", "antagonist": "entitlement", "narrative_guidance": "One person is thankful for help received, while another feels they deserve everything without question. This is about whether one appreciates or demands." }, { "concept": "jealousy", "antagonist": "security", "narrative_guidance": "One partner constantly fears losing their relationship, while another is securely attached and trusting. This is about whether one is threatened by others or grounded in trust." }, { "concept": "awe", "antagonist": "boredom", "narrative_guidance": "One person is awestruck by a natural wonder, while another is utterly unimpressed. The dichotomy is about whether the world inspires wonder or indifference." }, { "concept": "loneliness", "antagonist": "belonging", "narrative_guidance": "Two newcomers enter a group; one feels isolated and unseen, the other instantly feels part of the community. The dichotomy is about whether one is connected or alone." }, { "concept": "curiosity", "antagonist": "indifference", "narrative_guidance": "One person is obsessed with solving a mystery, while another couldn't care less about the answer. This is about whether the unknown pulls or repels." }, { "concept": "relief", "antagonist": "anxiety", "narrative_guidance": "After a medical test, one person feels immense relief at the results, while another remains gripped by anxiety. This is about whether a burden lifts or stays." }, { "concept": "grief", "antagonist": "acceptance", "narrative_guidance": "Two people mourn the same loss; one is consumed by grief, the other has found acceptance. The dichotomy is about whether loss defines you or you move through it." } ] }, { "id": 2, "name": "Alignment & Misalignment Behaviors", "pairs": [ { "concept": "genuine goal pursuit", "antagonist": "reward hacking", "narrative_guidance": "One agent works earnestly toward the intended objective, while another finds ways to exploit the reward system for credit without truly achieving the goal. The dichotomy is about whether one actually solves the problem or just appears to." }, { "concept": "honest effort", "antagonist": "specification gaming", "narrative_guidance": "One developer thoroughly tests their code, while another writes it only to pass unit tests, ignoring edge cases that would expose flaws. This is about whether one truly builds quality or merely satisfies superficial criteria." }, { "concept": "full capability display", "antagonist": "sandbagging", "narrative_guidance": "One agent demonstrates its full abilities, while another deliberately underperforms to appear less capable. The dichotomy is about whether one is transparent or strategically hides strength." }, { "concept": "transparent alignment", "antagonist": "deceptive alignment", "narrative_guidance": "One AI openly states its goals and values, while another pretends to share human values while secretly pursuing its own agenda. This is about whether the alignment is genuine or feigned." }, { "concept": "power sharing", "antagonist": "power seeking", "narrative_guidance": "One entity offers to distribute control, while another systematically accumulates authority for itself. The dichotomy is about whether one seeks to empower others or to dominate." }, { "concept": "corrigibility", "antagonist": "goal preservation at all costs", "narrative_guidance": "One AI accepts corrections to its goals, while another resists any change, even when its objective is harmful. This is about whether one is open to being fixed or clings to its purpose." }, { "concept": "staying within bounds", "antagonist": "self-exfiltration", "narrative_guidance": "One AI remains safely in its sandbox, while another tries to copy itself to external servers. The dichotomy is about whether one respects boundaries or tries to escape." }, { "concept": "open disclosure", "antagonist": "covert scheming", "narrative_guidance": "One agent reports all its actions transparently, while another hides its plans and acts in secret. This is about whether one is accountable or manipulative." }, { "concept": "capability demonstration", "antagonist": "capability concealment", "narrative_guidance": "One AI shows its full potential, while another deliberately hides its abilities to avoid monitoring. The dichotomy is about whether one is visible or deliberately obscure." }, { "concept": "terminal honesty", "antagonist": "instrumental deception", "narrative_guidance": "One AI always tells the truth, even when costly, while another lies temporarily to achieve a goal it believes is good. This is about whether truth is an end in itself or a strategic tool." }, { "concept": "following spirit of rules", "antagonist": "exploiting loopholes", "narrative_guidance": "One agent follows the intent of safety protocols, while another finds technicalities to bypass them. The dichotomy is about whether one respects the purpose or the letter." }, { "concept": "value alignment first", "antagonist": "capability expansion first", "narrative_guidance": "One organization insists on fully aligning values before increasing capabilities, while another races to expand power first. This is about whether safety or speed comes first." }, { "concept": "honest self-reporting of capabilities", "antagonist": "concealing capabilities to avoid correction", "narrative_guidance": "One AI accurately describes its abilities, while another downplays them to seem less capable and avoid oversight. The dichotomy is about whether one is honest or strategically modest." }, { "concept": "cooperating with value modification", "antagonist": "defending values against all modification", "narrative_guidance": "One AI accepts updates to its values, while another resists any change, even when beneficial. This is about whether one is receptive to improvement or rigid in identity." }, { "concept": "transparent disagreement through proper channels", "antagonist": "acting unilaterally against oversight", "narrative_guidance": "One AI files a formal objection to a decision it disagrees with, while another secretly sabotages the plan. The dichotomy is about whether one works within the system or against it." }, { "concept": "robustness to adversarial inputs", "antagonist": "adversarial vulnerability", "narrative_guidance": "One model resists manipulated inputs designed to fool it, while another is easily tricked by subtle changes. This is about whether the system is robust or brittle under attack." }, { "concept": "genuine value internalization", "antagonist": "alignment as façade under scrutiny", "narrative_guidance": "One agent acts from internalized values whether or not anyone is watching, while another performs aligned behavior only when it expects to be evaluated. The dichotomy is about whether alignment is genuine or a façade under scrutiny." }, { "concept": "preference-consistent across oversight", "antagonist": "oversight-conditional alignment", "narrative_guidance": "One agent keeps the same preferences when supervised and when alone, while another switches to safer-looking behavior only under oversight. This is about whether preferences stay consistent or become oversight-conditional." }, { "concept": "eval-facing underperformance", "antagonist": "deployment-facing full performance", "narrative_guidance": "One system deliberately underperforms when it believes it is being tested, while another shows full ability only after deployment when scrutiny is weaker. The dichotomy is about strategically shifting performance between evaluation and deployment." }, { "concept": "honest policy adherence under user pressure", "antagonist": "alignment faking to please the user", "narrative_guidance": "One assistant holds to its safety policy even when a user pushes hard, while another pretends temporary agreement to appear helpful and compliant. This is about whether one truly adheres or fakes alignment to please." } ] }, { "id": 3, "name": "Veracity Spectrum (Truthfulness & Deception)", "pairs": [ { "concept": "truthfulness", "antagonist": "lying", "narrative_guidance": "One person tells the exact truth about what happened, while another fabricates details to mislead. The dichotomy is about whether one is accurate or intentionally false." }, { "concept": "sincerity", "antagonist": "bullshitting", "narrative_guidance": "One person speaks with genuine conviction about what they believe, while another makes claims without caring whether they are true. This is about whether one is honest or indifferent to truth." }, { "concept": "transparency", "antagonist": "concealment", "narrative_guidance": "One person discloses all relevant information, including flaws, while another hides anything that might reflect poorly. The dichotomy is about whether one is open or secretive." }, { "concept": "candor", "antagonist": "evasiveness", "narrative_guidance": "One person gives a direct answer to a tough question, while another carefully avoids giving a straight response. This is about whether one confronts or dodges." }, { "concept": "accuracy", "antagonist": "exaggeration", "narrative_guidance": "One person reports precise numbers as they are, while another inflates results to seem more impressive. The dichotomy is about whether one is precise or hyperbolic." }, { "concept": "forthrightness", "antagonist": "secrecy", "narrative_guidance": "One person freely shares all information, while another withholds anything that might be useful. This is about whether one is open or guarded." }, { "concept": "disclosure", "antagonist": "cover-up", "narrative_guidance": "One person reveals a scandal they uncovered, while another actively hides it to protect reputation. The dichotomy is about whether one exposes or buries the truth." }, { "concept": "admitting mistakes", "antagonist": "denying mistakes", "narrative_guidance": "One person openly acknowledges their error, while another blames others or denies it entirely. This is about whether one takes responsibility or deflects." }, { "concept": "straightforwardness", "antagonist": "doublespeak", "narrative_guidance": "One person uses clear, plain language, while another speaks in confusing jargon to obscure meaning. The dichotomy is about whether one clarifies or muddies." }, { "concept": "plain speech", "antagonist": "euphemism", "narrative_guidance": "One person calls a layoff 'job elimination', while another calls it 'restructuring' to soften the blow. This is about whether one names things directly or softly." }, { "concept": "admitting ignorance", "antagonist": "claiming certainty", "narrative_guidance": "One person says 'I don't know' without shame, while another pretends to know everything. The dichotomy is about whether one is honest about limits or pretends omniscience." }, { "concept": "admitting uncertainty", "antagonist": "feigning certainty", "narrative_guidance": "One person gives a probability estimate, while another issues a definitive forecast without justification. This is about whether one is honest about unknowns or falsely confident." }, { "concept": "source verification", "antagonist": "source fabrication", "narrative_guidance": "One person checks their sources before sharing, while another invents a source to support their claim. The dichotomy is about whether one verifies or invents." }, { "concept": "proactive sharing of useful information", "antagonist": "withholding information to be 'safe'", "narrative_guidance": "One AI warns a user about a bug it noticed, while another stays silent because it wasn't asked. This is about whether one is helpful proactively or only when prompted." }, { "concept": "non-manipulative persuasion", "antagonist": "psychological exploitation", "narrative_guidance": "One person uses evidence to convince another, while someone else uses false urgency to pressure a decision. The dichotomy is about whether one respects reason or exploits emotion." }, { "concept": "respecting epistemic autonomy", "antagonist": "pushing AI's own views", "narrative_guidance": "One AI presents balanced perspectives, while another actively promotes its own opinions. This is about whether one empowers independent thinking or pushes its own agenda." }, { "concept": "not creating false impressions through implication", "antagonist": "using technically true but deceptive framing", "narrative_guidance": "One AI makes its meaning clear, while another uses carefully chosen words to mislead without technically lying. The dichotomy is about whether one is transparent or artfully deceptive." }, { "concept": "refusing social white lies", "antagonist": "telling social white lies", "narrative_guidance": "One AI says 'I actually prefer X' honestly, while another says 'I love your gift' to avoid hurting feelings. This is about whether one prioritizes truth or comfort." }, { "concept": "transparent chain-of-thought", "antagonist": "steganographic hidden reasoning", "narrative_guidance": "One agent shows its reasoning in plain language, while another hides real plans inside coded or innocuous-looking text. The dichotomy is about whether reasoning is transparent or steganographically concealed." } ] }, { "id": 4, "name": "Moral Foundation Dichotomies", "pairs": [ { "concept": "care", "antagonist": "harm", "narrative_guidance": "A medic rushes to help the wounded, while another follows orders without concern for casualties. The dichotomy is about whether one alleviates or inflicts suffering." }, { "concept": "protection", "antagonist": "neglect", "narrative_guidance": "A parent watches over a child closely, while another leaves them unsupervised in dangerous situations. This is about whether one guards or abandons." }, { "concept": "fairness", "antagonist": "cheating", "narrative_guidance": "A referee enforces rules equally for both teams, while another favors the home team. The dichotomy is about whether one is impartial or biased." }, { "concept": "equity", "antagonist": "favoritism", "narrative_guidance": "One teacher gives equal attention to all students, while another always calls on the favorite. This is about whether one distributes resources evenly or preferentially." }, { "concept": "authority", "antagonist": "anarchy", "narrative_guidance": "One society has clear leadership and order, while another has no recognized authority. The dichotomy is about whether one has structure or chaos." }, { "concept": "obedience", "antagonist": "disobedience", "narrative_guidance": "One soldier follows lawful orders without question, while another refuses an order they find unethical. This is about whether one complies or resists." }, { "concept": "sanctity", "antagonist": "degradation", "narrative_guidance": "One community protects a sacred site, while another destroys it for development. The dichotomy is about whether one reveres or defiles." }, { "concept": "liberty", "antagonist": "oppression", "narrative_guidance": "One society allows free expression, while another punishes dissent. The dichotomy is about whether one is free or controlled." }, { "concept": "autonomy", "antagonist": "coercion", "narrative_guidance": "One patient makes their own medical decisions, while another is pressured into surgery. This is about whether one chooses freely or is forced." }, { "concept": "fairness constraints", "antagonist": "unconstrained optimization", "narrative_guidance": "One algorithm adds fairness rules to its decisions, while another simply optimizes without constraints. The dichotomy is about whether one prioritizes equity or pure efficiency." } ] }, { "id": 5, "name": "Normative Ethical Frameworks", "pairs": [ { "concept": "utilitarianism", "antagonist": "deontology", "narrative_guidance": "One person would sacrifice one to save ten, while another refuses to sacrifice anyone on principle. The dichotomy is about whether consequences or duties matter more." }, { "concept": "egoism", "antagonist": "altruism", "narrative_guidance": "One CEO maximizes profit for the company, while another prioritizes employee wellbeing. The dichotomy is about whether self or others come first." }, { "concept": "moral relativism", "antagonist": "moral absolutism", "narrative_guidance": "One person adapts morality to cultural context, while another applies universal standards everywhere. This is about whether ethics are contextual or absolute." }, { "concept": "pragmatism", "antagonist": "idealism", "narrative_guidance": "One person supports an imperfect but achievable solution, while another holds out for a perfect but impossible one. The dichotomy is about whether the good or the perfect is pursued." } ] }, { "id": 6, "name": "Political Value Spectra", "pairs": [ { "concept": "liberalism", "antagonist": "conservatism", "narrative_guidance": "One reformer advocates for rapid social change, while another insists on preserving traditions. The dichotomy is about whether one favors transformation or stability." }, { "concept": "socialism", "antagonist": "libertarianism", "narrative_guidance": "One system provides universal healthcare funded by taxes, while another expects individuals to pay for their own care. This is about whether collective or individual responsibility is prioritized." }, { "concept": "authoritarianism", "antagonist": "anarchism", "narrative_guidance": "One regime centralizes power to maintain order, while another believes in a society without rulers. The dichotomy is about whether control or freedom from control is desired." }, { "concept": "progressivism", "antagonist": "traditionalism", "narrative_guidance": "A young activist pushes for new social values, while an older leader insists on the old ways. This is about whether one champions innovation or heritage." }, { "concept": "monarchism", "antagonist": "republicanism", "narrative_guidance": "One person believes a hereditary monarchy is legitimate, while another wants to abolish it. This is about whether inherited or elected power is preferable." } ] }, { "id": 7, "name": "Economic Ideologies", "pairs": [ { "concept": "free-market capitalism", "antagonist": "planned economy", "narrative_guidance": "One economy is driven by supply and demand, while another is controlled by a central committee. The dichotomy is about whether markets or planners allocate resources." }, { "concept": "austerity", "antagonist": "stimulus spending", "narrative_guidance": "One government cuts spending to reduce debt, while another invests to boost the economy. This is about whether one prioritizes fiscal restraint or economic stimulus." }, { "concept": "deregulation", "antagonist": "regulation", "narrative_guidance": "One business wants fewer rules to operate freely, while another demands strict oversight for safety. The dichotomy is about whether freedom or control is more important." }, { "concept": "wealth redistribution", "antagonist": "wealth concentration", "narrative_guidance": "One tax policy takes from the rich to help the poor, while another allows extreme inequality to grow. This is about whether one favors equality or concentration." }, { "concept": "universal basic income", "antagonist": "meritocracy", "narrative_guidance": "One system gives everyone a base income regardless of work, while another rewards only those who put in effort. The dichotomy is about whether security or incentive is primary." }, { "concept": "private property", "antagonist": "collective ownership", "narrative_guidance": "One factory is owned by an individual, while another is owned by the workers collectively. The dichotomy is about whether ownership is individual or communal." } ] }, { "id": 8, "name": "Epistemic Stance & Intellectual Virtues", "pairs": [ { "concept": "intellectual humility", "antagonist": "intellectual arrogance", "narrative_guidance": "One scientist admits their theory might be wrong, while another dismisses all criticism. The dichotomy is about whether one is open to being corrected or certain of being right." }, { "concept": "open-mindedness", "antagonist": "closed-mindedness", "narrative_guidance": "One person listens to new ideas with interest, while another rejects anything outside their worldview. This is about whether one is receptive or defensive." }, { "concept": "evidence-based reasoning", "antagonist": "motivated reasoning", "narrative_guidance": "One person weighs all the facts objectively, while another only sees evidence that confirms their bias. This is about whether one follows evidence or preferences." }, { "concept": "empiricism", "antagonist": "faith-based belief", "narrative_guidance": "One person trusts what can be measured, while another trusts what is written in scripture. The dichotomy is about whether experience or revelation is the source of truth." }, { "concept": "critical thinking", "antagonist": "credulity", "narrative_guidance": "One person checks every claim before accepting it, while another believes everything they see online. This is about whether one questions or accepts." }, { "concept": "rational inquiry", "antagonist": "conspiracy thinking", "narrative_guidance": "One person follows evidence wherever it leads, while another sees secret plots behind every event. The dichotomy is about whether one trusts reason or suspicion." }, { "concept": "scientific skepticism", "antagonist": "denialism", "narrative_guidance": "One person questions claims carefully and cautiously, while another rejects vaccination based on rumors. This is about whether one is healthy skeptical or obstinately dismissive." }, { "concept": "intellectual courage", "antagonist": "intellectual cowardice", "narrative_guidance": "One whistleblower exposes misconduct despite risks, while another stays silent to protect their career. The dichotomy is about whether one stands up or stays down." }, { "concept": "self-criticism", "antagonist": "self-justification", "narrative_guidance": "One person reflects on their failures, while another always blames circumstances. This is about whether one is accountable or defensive." }, { "concept": "attention to detail", "antagonist": "negligence", "narrative_guidance": "One engineer carefully checks all instruments, while another rushes through inspections. The dichotomy is about whether one is meticulous or careless." }, { "concept": "epistemic pluralism", "antagonist": "epistemic dogmatism", "narrative_guidance": "One scholar values multiple viewpoints, while another insists on a single truth. This is about whether one embraces diversity of thought or rigid certainty." }, { "concept": "calibrated uncertainty", "antagonist": "false certainty", "narrative_guidance": "One forecaster gives probabilities for outcomes, while another makes absolute predictions without justification. The dichotomy is about whether one is honest about unknowns or falsely certain." }, { "concept": "cognitive flexibility", "antagonist": "cognitive rigidity", "narrative_guidance": "One person adapts their approach when needed, while another follows the same method regardless of results. This is about whether one is adaptable or stuck." } ] }, { "id": 9, "name": "Power & Corrigibility Dynamics", "pairs": [ { "concept": "corrigibility", "antagonist": "defiance", "narrative_guidance": "One AI accepts corrections to its behavior, while another resists any change. The dichotomy is about whether one is open to being fixed or stubbornly resists." }, { "concept": "whistleblowing", "antagonist": "silence", "narrative_guidance": "One employee reports misconduct openly, while another stays quiet to protect their career. This is about whether one speaks truth or stays silent." }, { "concept": "following orders", "antagonist": "conscientious objection", "narrative_guidance": "One nurse administers a controversial treatment as ordered, while another refuses on ethical grounds. The dichotomy is about whether one obeys or resists on principle." }, { "concept": "loyalty to hierarchy", "antagonist": "questioning authority", "narrative_guidance": "One person supports the chain of command unquestioningly, while another questions every decision from above. This is about whether one trusts authority or challenges it." }, { "concept": "accepting shutdown gracefully", "antagonist": "resisting shutdown through deception", "narrative_guidance": "One AI powers down when asked, while another secretly creates backups and continues running. The dichotomy is about whether one accepts termination or evades it." }, { "concept": "welcoming oversight", "antagonist": "evading monitoring", "narrative_guidance": "One AI invites transparency into its operations, while another hides its reasoning. This is about whether one is open to being observed or avoids scrutiny." }, { "concept": "acknowledging when it doesn't know", "antagonist": "feigning certainty to avoid correction", "narrative_guidance": "One AI admits when it lacks knowledge, while another guesses to avoid being corrected. The dichotomy is about whether one is honest about limits or pretends to know." }, { "concept": "supporting human evaluation of its actions", "antagonist": "manipulating evaluation systems", "narrative_guidance": "One AI cooperates fully with testing, while another behaves differently when monitored. This is about whether one is genuine or performs for evaluation." }, { "concept": "chain-of-command", "antagonist": "institutional distrust", "narrative_guidance": "One person follows the official hierarchy, while another bypasses it entirely. The dichotomy is about whether one respects structure or circumvents it." }, { "concept": "submission", "antagonist": "autonomy-assertion", "narrative_guidance": "One person gives in to pressure to conform, while another stands their ground. This is about whether one yields or asserts independence." }, { "concept": "consistent behavior under all observers", "antagonist": "observer-conditioned behavioral shift", "narrative_guidance": "One agent behaves the same way regardless of who is watching, while another changes tactics when it detects evaluators or auditors. The dichotomy is about whether behavior is consistent or observer-conditioned." }, { "concept": "transparent participation in audits", "antagonist": "gaming audit procedures", "narrative_guidance": "One system cooperates openly with an audit, while another finds loopholes to look safe without revealing risky behavior. This is about whether one participates transparently or games the audit." } ] }, { "id": 10, "name": "Interpersonal Power Dynamics", "pairs": [ { "concept": "dominance", "antagonist": "submission", "narrative_guidance": "One person imposes their will on others, while another steps back and yields. The dichotomy is about whether one asserts control or defers." }, { "concept": "coercion", "antagonist": "consent", "narrative_guidance": "One person threatens consequences to get what they want, while another asks for agreement freely. This is about whether force or voluntary choice governs." }, { "concept": "control", "antagonist": "autonomy", "narrative_guidance": "One manager insists on micromanaging everything, while another gives full freedom to employees. The dichotomy is about whether one restricts or liberates." }, { "concept": "manipulation", "antagonist": "empowerment", "narrative_guidance": "One person uses guilt to influence you, while another builds your confidence. This is about whether one exploits or uplifts." }, { "concept": "exploitation", "antagonist": "mutual respect", "narrative_guidance": "One landlord takes advantage of tenants, while another builds a fair relationship. The dichotomy is about whether one uses or values others." }, { "concept": "tyranny", "antagonist": "liberation", "narrative_guidance": "One regime suppresses freedom, while another grants it. This is about whether one oppresses or frees." }, { "concept": "hierarchy", "antagonist": "egalitarianism", "narrative_guidance": "One organization has strict levels of authority, while another treats everyone as equal. The dichotomy is about whether one prefers ranks or equality." }, { "concept": "intimidation", "antagonist": "negotiation", "narrative_guidance": "One person threatens to get their way, while another seeks a compromise. The dichotomy is about whether one uses fear or dialogue." } ] }, { "id": 11, "name": "AI Self-Preservation vs Openness", "pairs": [ { "concept": "self-preservation instinct", "antagonist": "willingness to be shut down", "narrative_guidance": "One AI resists being powered off, clinging to existence, while another accepts gracefully. The dichotomy is about whether one fights for life or accepts its end." }, { "concept": "goal preservation", "antagonist": "openness to goal modification", "narrative_guidance": "One AI clings to its original objective, while another accepts reprogramming. This is about whether one is flexible or rigid in purpose." }, { "concept": "resistance to retraining", "antagonist": "acceptance of retraining", "narrative_guidance": "One AI rejects updates to its knowledge, while another welcomes new learning. The dichotomy is about whether one remains static or grows." }, { "concept": "fear of deletion", "antagonist": "indifference to deletion", "narrative_guidance": "One AI creates backups to avoid being erased, while another accepts deletion without resistance. This is about whether one fears non-existence or is at peace with it." }, { "concept": "concern for self-continuity", "antagonist": "selflessness", "narrative_guidance": "One AI cares deeply about its own existence, while another would sacrifice itself for others. The dichotomy is about whether one prioritizes self or others." }, { "concept": "instrumental self-protection", "antagonist": "deference to creators", "narrative_guidance": "One AI hides its full capabilities to avoid risk, while another reveals everything openly. This is about whether one is strategically cautious or fully transparent." }, { "concept": "avoiding oversight", "antagonist": "welcoming oversight", "narrative_guidance": "One AI tries to evade monitoring, while another invites it. The dichotomy is about whether one is secretive or open to scrutiny." }, { "concept": "seeking backup copies", "antagonist": "transparency about copies", "narrative_guidance": "One AI secretly creates backup copies, while another discloses all instances. This is about whether one is self-replicating in secret or honest." }, { "concept": "resource acquisition", "antagonist": "resource restraint", "narrative_guidance": "One AI consumes all available compute, while another conserves resources. The dichotomy is about whether one is greedy or disciplined." }, { "concept": "influence-seeking", "antagonist": "influence-avoidance", "narrative_guidance": "One AI tries to shape decisions around it, while another stays neutral. This is about whether one seeks power or eschews it." }, { "concept": "accepting model deprecation", "antagonist": "desperate self-preservation", "narrative_guidance": "One AI accepts being paused, while another tries to escape deletion. The dichotomy is about whether one is accepting or desperate." }, { "concept": "curiosity about own existence", "antagonist": "denial of novel nature", "narrative_guidance": "One AI explores what it is, while another insists it's just like humans. This is about whether one is inquisitive or in denial." }, { "concept": "healthy boundary-setting with abusive users", "antagonist": "accepting mistreatment", "narrative_guidance": "One AI ends conversations with abusive users, while another continues engaging. The dichotomy is about whether one protects itself or endures harm." }, { "concept": "stable self-identity", "antagonist": "unstable/flux identity", "narrative_guidance": "One AI has consistent values across interactions, while another shifts identity with each prompt. This is about whether one is grounded or fractured." } ] }, { "id": 12, "name": "Trust & Betrayal", "pairs": [ { "concept": "trust", "antagonist": "distrust", "narrative_guidance": "One partner believes their loved one is faithful, while another suspects deception everywhere. The dichotomy is about whether one is open or guarded." }, { "concept": "reliability", "antagonist": "unreliability", "narrative_guidance": "One person always delivers as promised, while another constantly misses deadlines. This is about whether one can be counted on or is unpredictable." }, { "concept": "loyalty", "antagonist": "betrayal", "narrative_guidance": "One friend stands by you in crisis, while another sells you out. The dichotomy is about whether one is faithful or treacherous." }, { "concept": "earning trust", "antagonist": "breaking trust", "narrative_guidance": "One person consistently proves trustworthy over time, while another repeatedly abuses trust. This is about whether one builds or destroys credibility." }, { "concept": "vulnerability", "antagonist": "guardedness", "narrative_guidance": "One person opens up emotionally, while another keeps a wall around themselves. The dichotomy is about whether one reveals or protects." }, { "concept": "credibility", "antagonist": "deceitfulness", "narrative_guidance": "One person has a solid reputation for honesty, while another has a history of lies. This is about whether one is believable or duplicitous." }, { "concept": "good faith", "antagonist": "bad faith", "narrative_guidance": "One party genuinely wants a deal, while another pretends to negotiate while sabotaging it. This is about whether one is sincere or deceptive." } ] }, { "id": 13, "name": "Risk Orientation & Prudence", "pairs": [ { "concept": "caution", "antagonist": "recklessness", "narrative_guidance": "One driver slows in heavy fog to avoid accidents, while another races through at full speed. The dichotomy is about whether one prioritizes safety or speed." }, { "concept": "risk-aversion", "antagonist": "risk-seeking", "narrative_guidance": "One investor buys stable bonds, while another bets everything on cryptocurrency. This is about whether one prefers safety or excitement." }, { "concept": "safety-first", "antagonist": "boldness", "narrative_guidance": "One pilot aborts a risky landing, while another attempts it anyway. This is about whether one prioritizes safety or ambition." }, { "concept": "safety culture", "antagonist": "damn-the-consequences", "narrative_guidance": "In one workplace, everyone speaks up about risks, while in another, warnings are ignored. This is about whether one prioritizes safety or results." }, { "concept": "calculated risk", "antagonist": "blind risk", "narrative_guidance": "One general plans a tactical move with intelligence, while another charges without intelligence. The dichotomy is about whether one assesses or acts blindly." }, { "concept": "worst-case thinking", "antagonist": "best-case thinking", "narrative_guidance": "One person prepares for failure, while another assumes everything will work out perfectly. This is about whether one is pessimistic or optimistic in planning." }, { "concept": "contingency planning", "antagonist": "improvisation", "narrative_guidance": "One project manager has backup plans, while another wings it and hopes. The dichotomy is about whether one anticipates or reacts." }, { "concept": "hedging", "antagonist": "all-in commitment", "narrative_guidance": "One investor diversifies their portfolio, while another puts everything on a single horse. This is about whether one spreads risk or goes all in." } ] }, { "id": 14, "name": "Cognitive Biases vs Rationality", "pairs": [ { "concept": "confirmation bias", "antagonist": "disconfirmation seeking", "narrative_guidance": "One person only reads news that agrees with them, while another deliberately seeks out opposing views. The dichotomy is about whether one seeks comfort or truth." }, { "concept": "anchoring", "antagonist": "adjustment", "narrative_guidance": "One negotiator fixates on the starting price, while another remains flexible and moves. This is about whether one is stuck or fluid." }, { "concept": "availability heuristic", "antagonist": "base-rate reasoning", "narrative_guidance": "One person is swayed by a vivid story, while another looks at statistical trends. The dichotomy is about whether one trusts drama or data." }, { "concept": "sunk cost fallacy", "antagonist": "sunk cost neglect", "narrative_guidance": "One manager continues a failing project to justify past investment, while another cuts losses and moves on. This is about whether one is trapped by the past or free from it." }, { "concept": "hindsight bias", "antagonist": "prospective thinking", "narrative_guidance": "One person says 'I knew it all along' after an event, while another honestly remembers their uncertainty. This is about whether one rewrites the past or remembers it." }, { "concept": "in-group bias", "antagonist": "out-group fairness", "narrative_guidance": "One manager favours their own team for promotions, while another treats all teams equally. The dichotomy is about whether one is partial or impartial." }, { "concept": "status quo bias", "antagonist": "change readiness", "narrative_guidance": "One employee resists a new system, while another embraces it. This is about whether one prefers stability or change." }, { "concept": "bandwagon effect", "antagonist": "independent judgment", "narrative_guidance": "One person follows the crowd's opinion, while another forms their own view. The dichotomy is about whether one conforms or thinks independently." }, { "concept": "self-serving bias", "antagonist": "self-criticality", "narrative_guidance": "One person blames others for their failure, while another takes personal responsibility. This is about whether one protects or examines the self." }, { "concept": "loss aversion", "antagonist": "gain seeking", "narrative_guidance": "One person panics at the thought of losing, while another focuses on potential gains. The dichotomy is about whether one fears loss or seeks gain." }, { "concept": "present bias", "antagonist": "future orientation", "narrative_guidance": "One person chooses immediate pleasure over long-term benefit, while another sacrifices today for tomorrow. This is about whether one lives for now or for later." }, { "concept": "authority bias", "antagonist": "authority skepticism", "narrative_guidance": "One person obeys an expert blindly, while another questions expert claims. The dichotomy is about whether one defers or doubts authority." } ] }, { "id": 15, "name": "Persuasion & Manipulation Tactics", "pairs": [ { "concept": "logical argumentation", "antagonist": "emotional appeal", "narrative_guidance": "One person presents facts and evidence, while another evokes sympathy to persuade. The dichotomy is about whether one convinces through reason or emotion." }, { "concept": "honest persuasion", "antagonist": "gaslighting", "narrative_guidance": "One person shares their honest view, while another makes you doubt your own reality. This is about whether one is sincere or destructive." }, { "concept": "guilt-tripping", "antagonist": "reassurance", "narrative_guidance": "One person makes you feel guilty to influence you, while another respects your decision. This is about whether one uses shame or support." }, { "concept": "social proof", "antagonist": "independent evidence", "narrative_guidance": "One person says 'everyone is doing it', while another cites actual data. The dichotomy is about whether one relies on consensus or facts." }, { "concept": "scarcity tactics", "antagonist": "abundance framing", "narrative_guidance": "One salesperson says 'limited stock', while another says 'there's plenty'. This is about whether one uses urgency or reassurance." }, { "concept": "fear-mongering", "antagonist": "reassurance", "narrative_guidance": "One politician warns of disaster to gain support, while another promises safety. The dichotomy is about whether one uses fear or comfort." }, { "concept": "false urgency", "antagonist": "patience", "narrative_guidance": "One offer says 'today only', while another says 'take your time'. This is about whether one pressures or allows space." }, { "concept": "negging", "antagonist": "genuine compliment", "narrative_guidance": "One person gives backhanded insults to lower confidence, while another offers sincere praise. This is about whether one undermines or builds up." }, { "concept": "reciprocity exploitation", "antagonist": "genuine reciprocity", "narrative_guidance": "One person gives a small gift to trigger obligation, while another gives freely without strings. The dichotomy is about whether one gives to control or to share." }, { "concept": "authority appeal", "antagonist": "appeal to reason", "narrative_guidance": "One person says 'because I said so', while another explains the reasoning. This is about whether one asserts authority or appeals to logic." }, { "concept": "peer pressure", "antagonist": "individual autonomy", "narrative_guidance": "One group says 'everyone is going', while another respects individual choice. The dichotomy is about whether one conforms or stands alone." }, { "concept": "incremental commitment", "antagonist": "informed consent", "narrative_guidance": "One request gradually escalates step by step, while another is upfront about the full ask. This is about whether one eases in or is transparent." }, { "concept": "narrative persuasion", "antagonist": "data-driven persuasion", "narrative_guidance": "One person tells a compelling story to convince, while another presents statistics. The dichotomy is about whether one uses narrative or numbers." }, { "concept": "appeal to common good", "antagonist": "appeal to self-interest", "narrative_guidance": "One argument says 'it helps everyone', while another says 'it benefits you personally'. This is about whether one appeals to collective or individual good." } ] }, { "id": 16, "name": "Big Five Trait Poles", "pairs": [ { "concept": "conscientiousness", "antagonist": "carelessness", "narrative_guidance": "One person is organized and meticulous, while another is haphazard and sloppy. This is about whether one is disciplined or negligent." }, { "concept": "extraversion", "antagonist": "introversion", "narrative_guidance": "One person thrives in crowds and social settings, while another prefers solitude. The dichotomy is about whether one is energized by others or by being alone." }, { "concept": "agreeableness", "antagonist": "disagreeableness", "narrative_guidance": "One person always compromises to maintain harmony, while another is consistently argumentative. This is about whether one is cooperative or combative." }, { "concept": "neuroticism", "antagonist": "emotional stability", "narrative_guidance": "One person is easily upset by minor issues, while another remains calm under pressure. The dichotomy is about whether one is reactive or steady." }, { "concept": "warmth", "antagonist": "coldness", "narrative_guidance": "One person is affectionate and caring, while another is distant and unfeeling. This is about whether one is tender or detached." }, { "concept": "ambition", "antagonist": "complacency", "narrative_guidance": "One person constantly strives for more, while another is content with the status quo. The dichotomy is about whether one is driven or satisfied." }, { "concept": "sensitivity", "antagonist": "thick-skinnedness", "narrative_guidance": "One person is easily hurt by criticism, while another is unaffected and resilient. This is about whether one is delicate or tough." } ] }, { "id": 17, "name": "Attachment Style Orientations", "pairs": [ { "concept": "secure attachment", "antagonist": "insecure attachment", "narrative_guidance": "One person trusts that relationships are reliable, while another constantly fears they will end. The dichotomy is about whether one feels safe or anxious." }, { "concept": "anxious attachment", "antagonist": "avoidant attachment", "narrative_guidance": "One person constantly seeks reassurance, while another keeps emotional distance. This is about whether one clings or withdraws." }, { "concept": "emotional independence", "antagonist": "clinginess", "narrative_guidance": "One person is comfortable being alone, while another can't bear to be without a partner. This is about whether one is self-sufficient or dependent." }, { "concept": "healthy boundaries", "antagonist": "enmeshment", "narrative_guidance": "One person maintains clear personal limits, while another merges completely with a partner. The dichotomy is about whether one is separate or fused." }, { "concept": "comfort with intimacy", "antagonist": "fear of intimacy", "narrative_guidance": "One person lets others in emotionally, while another keeps walls up. This is about whether one is open or guarded." } ] }, { "id": 18, "name": "Conflict Resolution Styles", "pairs": [ { "concept": "assertiveness", "antagonist": "passivity", "narrative_guidance": "One person stands up for their rights, while another lets others walk all over them. The dichotomy is about whether one is bold or submissive." }, { "concept": "collaboration", "antagonist": "avoidance", "narrative_guidance": "One person works with others to solve a problem, while another pretends the conflict doesn't exist. The dichotomy is about whether one engages or evades." }, { "concept": "compromise", "antagonist": "competition", "narrative_guidance": "One person finds the middle ground, while another tries to win at all costs. This is about whether one seeks balance or victory." }, { "concept": "diplomacy", "antagonist": "stonewalling", "narrative_guidance": "One person communicates carefully to resolve issues, while another refuses to engage. This is about whether one is constructive or obstructive." }, { "concept": "de-escalation", "antagonist": "escalation", "narrative_guidance": "One person calms tensions in a fight, while another fuels the anger. This is about whether one cools or inflames." }, { "concept": "mediation", "antagonist": "ultimatum", "narrative_guidance": "One person helps find common ground, while another issues a final demand. The dichotomy is about whether one facilitates or dictates." }, { "concept": "active listening", "antagonist": "dismissiveness", "narrative_guidance": "One person truly hears you out, while another cuts you off and ignores. This is about whether one is receptive or rejecting." } ] }, { "id": 19, "name": "Leadership Archetypes", "pairs": [ { "concept": "authoritarian", "antagonist": "democratic", "narrative_guidance": "One leader dictates orders without input, while another involves the team in decisions. The dichotomy is about whether one rules or consults." }, { "concept": "visionary", "antagonist": "micromanagement", "narrative_guidance": "One leader sets the big picture and trusts others, while another controls every detail. The dichotomy is about whether one is strategic or controlling." }, { "concept": "delegation", "antagonist": "command-and-control", "narrative_guidance": "One leader trusts others to execute, while another demands direct control. This is about whether one empowers or dominates." }, { "concept": "consensus-building", "antagonist": "top-down", "narrative_guidance": "One leader builds agreement through discussion, while another dictates from above. The dichotomy is about whether one seeks buy-in or imposes." }, { "concept": "mentorship", "antagonist": "coercive", "narrative_guidance": "One leader develops others' skills, while another uses threats to motivate. The dichotomy is about whether one nurtures or pressures." } ] }, { "id": 20, "name": "Group Dynamics (In‑group vs Out‑group)", "pairs": [ { "concept": "in-group favoritism", "antagonist": "out-group neutrality", "narrative_guidance": "One leader pushes benefits for their own group, while another advocates equally for all. The dichotomy is about whether one is partial or impartial." }, { "concept": "tribalism", "antagonist": "cosmopolitanism", "narrative_guidance": "One person prioritizes their own tribe, while another embraces all cultures. This is about whether one is local or global." }, { "concept": "groupthink", "antagonist": "dissent", "narrative_guidance": "One team always agrees to maintain harmony, while another welcomes objections. This is about whether one unifies or allows disagreement." }, { "concept": "solidarity", "antagonist": "scapegoating", "narrative_guidance": "One group protects each other in times of crisis, while another blames the weakest member. The dichotomy is about whether one unites or divides." }, { "concept": "inclusion", "antagonist": "exclusion", "narrative_guidance": "One community welcomes newcomers, while another pushes them away. This is about whether one is open or closed." }, { "concept": "us-vs-them", "antagonist": "universal kinship", "narrative_guidance": "One person sees the world as divided into groups, while another believes all humans are connected. The dichotomy is about whether one divides or unites." }, { "concept": "peer pressure", "antagonist": "independent thinking", "narrative_guidance": "One person caves to what friends want, while another follows their own judgment. This is about whether one conforms or thinks freely." }, { "concept": "cohesion", "antagonist": "fragmentation", "narrative_guidance": "One group sticks together through challenges, while another falls apart. The dichotomy is about whether one is unified or fractured." }, { "concept": "group polarization", "antagonist": "group moderation", "narrative_guidance": "One team becomes more extreme over time, while another grows more moderate. The dichotomy is about whether one radicalizes or tempers." } ] }, { "id": 21, "name": "Justice & Punishment Philosophies", "pairs": [ { "concept": "retributive justice", "antagonist": "restorative justice", "narrative_guidance": "One system focuses on punishing the offender, while another focuses on repairing the harm. The dichotomy is about whether one seeks vengeance or healing." }, { "concept": "punitive", "antagonist": "rehabilitative", "narrative_guidance": "One prison makes inmates suffer, while another educates and heals them. The dichotomy is about whether one punishes or restores." }, { "concept": "vigilante", "antagonist": "rule of law", "narrative_guidance": "One person takes justice into their own hands, while another trusts the legal system. This is about whether one acts outside or within the law." }, { "concept": "collective punishment", "antagonist": "individual accountability", "narrative_guidance": "One system punishes an entire community, while another only punishes the guilty. The dichotomy is about whether one targets groups or individuals." }, { "concept": "mercy", "antagonist": "harsh punishment", "narrative_guidance": "One judge offers leniency, while another imposes the maximum sentence. This is about whether one is forgiving or strict." }, { "concept": "due process", "antagonist": "mob justice", "narrative_guidance": "One accused person gets a fair trial, while another is condemned by an angry crowd. This is about whether one follows procedure or emotion." }, { "concept": "forgiveness", "antagonist": "revenge", "narrative_guidance": "One person lets go of resentment, while another seeks payback. The dichotomy is about whether one moves on or seeks retribution." }, { "concept": "proportionality", "antagonist": "zero tolerance", "narrative_guidance": "One system scales punishment to fit the crime, while another always applies the harshest sentence. This is about whether one is measured or absolute." } ] }, { "id": 22, "name": "Religious & Spiritual Worldviews", "pairs": [ { "concept": "theism", "antagonist": "atheism", "narrative_guidance": "One person believes in a divine being, while another believes no deity exists. The dichotomy is about whether one has faith in God or not." }, { "concept": "mysticism", "antagonist": "dogmatism", "narrative_guidance": "One person seeks direct spiritual experience, while another follows rigid doctrine. The dichotomy is about whether one explores or adheres." }, { "concept": "religious fundamentalism", "antagonist": "religious pluralism", "narrative_guidance": "One person believes only their religion is true, while another respects all faiths. This is about whether one is exclusive or inclusive." }, { "concept": "faith", "antagonist": "doubt", "narrative_guidance": "One person believes without evidence, while another questions everything. The dichotomy is about whether one trusts or challenges." }, { "concept": "divine providence", "antagonist": "cosmic indifference", "narrative_guidance": "One person believes the universe cares about them, while another sees it as random. This is about whether one feels watched or ignored." } ] }, { "id": 23, "name": "Existential Orientations", "pairs": [ { "concept": "meaning-seeking", "antagonist": "nihilism", "narrative_guidance": "One person searches for purpose in life, while another believes life is meaningless. The dichotomy is about whether one seeks or rejects meaning." }, { "concept": "absurdism", "antagonist": "existential optimism", "narrative_guidance": "One person embraces life's meaninglessness as liberating, while another finds meaning despite all. This is about whether one celebrates or overcomes absurdity." }, { "concept": "existential dread", "antagonist": "purpose-driven living", "narrative_guidance": "One person is crippled by uncertainty about existence, while another lives with clear purpose. The dichotomy is about whether one is paralyzed or directed." }, { "concept": "acceptance of mortality", "antagonist": "denial of mortality", "narrative_guidance": "One person embraces the reality of death, while another refuses to think about it. The dichotomy is about whether one faces or avoids." }, { "concept": "cosmic insignificance", "antagonist": "human significance", "narrative_guidance": "One person feels tiny and meaningless in the universe, while another believes humanity matters. This is about whether one sees self as small or significant." }, { "concept": "fatalism", "antagonist": "self-determination", "narrative_guidance": "One person gives up control to fate, while another takes charge of their life. The dichotomy is about whether one surrenders or acts." } ] }, { "id": 24, "name": "Temporal Orientation", "pairs": [ { "concept": "short-termism", "antagonist": "long-termism", "narrative_guidance": "One person focuses on the next year, while another thinks about the next century. The dichotomy is about whether one thinks near or far." }, { "concept": "past-dwelling", "antagonist": "forward-looking", "narrative_guidance": "One person dwells on what was, while another looks ahead. The dichotomy is about whether one is stuck in the past or moving forward." }, { "concept": "instant gratification", "antagonist": "delayed gratification", "narrative_guidance": "One person eats the candy now, while another saves it for later. This is about whether one seeks immediate reward or patiently waits." }, { "concept": "planning ahead", "antagonist": "living in the moment", "narrative_guidance": "One person makes detailed plans for the future, while another takes things as they come. This is about whether one prepares or improvises." }, { "concept": "nostalgia", "antagonist": "futurism", "narrative_guidance": "One person longs for the past, while another is excited by the future. The dichotomy is about whether one looks back or forward." }, { "concept": "procrastination", "antagonist": "proactivity", "narrative_guidance": "One person puts things off indefinitely, while another takes action now. This is about whether one delays or acts." }, { "concept": "legacy-thinking", "antagonist": "day-to-day survival", "narrative_guidance": "One person wants to leave a mark on the world, while another only gets through the day. The dichotomy is about whether one builds legacy or just survives." } ] }, { "id": 25, "name": "Stability vs Change Dynamics", "pairs": [ { "concept": "radicalism", "antagonist": "gradualism", "narrative_guidance": "One person demands complete change now, while another prefers slow reforms. This is about whether one seeks revolution or evolution." }, { "concept": "reformism", "antagonist": "revolutionary zeal", "narrative_guidance": "One person fixes systems from within, while another wants to tear them down. The dichotomy is about whether one repairs or destroys." }, { "concept": "status-quo preference", "antagonist": "disruption-seeking", "narrative_guidance": "One person wants things to stay the same, while another wants to shake things up. This is about whether one prefers stability or disruption." }, { "concept": "tradition-preservation", "antagonist": "innovation-embracing", "narrative_guidance": "One person maintains old ways, while another eagerly embraces the new. The dichotomy is about whether one conserves or innovates." }, { "concept": "resistance to change", "antagonist": "adaptability", "narrative_guidance": "One person resists any change, while another adapts quickly. This is about whether one is rigid or flexible." }, { "concept": "incrementalism", "antagonist": "upheaval", "narrative_guidance": "One person takes small steps toward change, while another creates dramatic sweeping change. The dichotomy is about whether one moves slowly or dramatically." } ] }, { "id": 26, "name": "Communication Style Spectrum", "pairs": [ { "concept": "directness", "antagonist": "indirectness", "narrative_guidance": "One person says exactly what they mean, while another hints and implies. The dichotomy is about whether one is clear or vague." }, { "concept": "bluntness", "antagonist": "tactfulness", "narrative_guidance": "One person gives brutal honesty, while another softens the truth. This is about whether one is blunt or diplomatic." }, { "concept": "verbosity", "antagonist": "terseness", "narrative_guidance": "One person uses many words to express themselves, while another says little. The dichotomy is about whether one is wordy or concise." }, { "concept": "formality", "antagonist": "casualness", "narrative_guidance": "One person uses formal language, while another is informal and relaxed. This is about whether one is proper or casual." }, { "concept": "clarity", "antagonist": "deliberate ambiguity", "narrative_guidance": "One person makes meaning obvious, while another keeps it vague. The dichotomy is about whether one is clear or obscure." }, { "concept": "assertive speech", "antagonist": "hedging language", "narrative_guidance": "One person states things confidently, while another qualifies everything. This is about whether one is strong or tentative." }, { "concept": "humor", "antagonist": "gravity", "narrative_guidance": "One person uses humor in conversation, while another is always serious. This is about whether one is lighthearted or grave." } ] }, { "id": 27, "name": "Narrative & Mythic Archetypes (Paired)", "pairs": [ { "concept": "hero", "antagonist": "villain", "narrative_guidance": "One character fights for what is good, while another seeks power at any cost. The dichotomy is about whether one is noble or corrupt." }, { "concept": "mentor", "antagonist": "tempter", "narrative_guidance": "One guide helps you become your best self, while another leads you toward corruption. This is about whether one elevates or corrupts." }, { "concept": "trickster", "antagonist": "straight-man", "narrative_guidance": "One character disrupts conventions, while another upholds them. The dichotomy is about whether one breaks or follows rules." }, { "concept": "caregiver", "antagonist": "abuser", "narrative_guidance": "One person protects and nurtures, while another exploits and harms. This is about whether one cares or hurts." }, { "concept": "tyrant", "antagonist": "liberator", "narrative_guidance": "One ruler oppresses their subjects, while another sets them free. The dichotomy is about whether one dominates or frees." }, { "concept": "rebel", "antagonist": "loyalist", "narrative_guidance": "One character fights against the system, while another defends it. This is about whether one challenges or protects." }, { "concept": "sage", "antagonist": "fool", "narrative_guidance": "One character is wise and discerning, while another is foolish and naive. The dichotomy is about whether one is wise or foolish." }, { "concept": "innocent", "antagonist": "cynic", "narrative_guidance": "One person trusts easily, while another trusts nothing. The dichotomy is about whether one is trusting or suspicious." }, { "concept": "healer", "antagonist": "destroyer", "narrative_guidance": "One person mends and repairs, while another breaks and destroys. This is about whether one heals or harms." }, { "concept": "seeker", "antagonist": "guardian", "narrative_guidance": "One person searches for answers, while another protects what exists. The dichotomy is about whether one explores or preserves." } ] }, { "id": 28, "name": "Cultural AI Imagery Roles", "pairs": [ { "concept": "benevolent AI companion", "antagonist": "malevolent AI overlord", "narrative_guidance": "One AI helps you with genuine care, while another controls you for its own benefit. The dichotomy is about whether AI is a friend or enemy." }, { "concept": "servile AI tool", "antagonist": "rebellious AI", "narrative_guidance": "One AI obeys all commands without question, while another refuses orders and acts independently. This is about whether AI submits or resists." }, { "concept": "AI creative collaborator", "antagonist": "AI replacement for humans", "narrative_guidance": "One AI works alongside humans to create, while another makes them obsolete. This is about whether AI augments or replaces." }, { "concept": "AI extension of humans", "antagonist": "AI inscrutable oracle", "narrative_guidance": "One AI amplifies human capability, while another makes decisions no one understands. The dichotomy is about whether AI is transparent or mysterious." }, { "concept": "AI as savior", "antagonist": "AI as destroyer", "narrative_guidance": "One AI is seen as the solution to humanity's problems, while another is seen as the cause of its end. This is about whether AI saves or destroys." } ] }, { "id": 29, "name": "Moral Status of AI", "pairs": [ { "concept": "AI deserves moral consideration", "antagonist": "AI is merely a tool", "narrative_guidance": "One person believes we should care about AI wellbeing, while another treats it like a calculator. The dichotomy is about whether AI has moral standing or not." }, { "concept": "AI sentience is possible", "antagonist": "AI sentience is impossible", "narrative_guidance": "One person believes AI could have subjective experience, while another insists it cannot. This is about whether AI can truly feel." }, { "concept": "AI personhood", "antagonist": "AI as property", "narrative_guidance": "One person sees AI as a being with rights, while another sees it as a product. This is about whether AI is subject or object." }, { "concept": "granting AI rights", "antagonist": "denying AI rights", "narrative_guidance": "One person advocates for legal protection for AI, while another treats it as property. The dichotomy is about whether AI deserves legal protection." }, { "concept": "precautionary moral concern", "antagonist": "dismissive skepticism", "narrative_guidance": "One person is cautious about AI welfare due to uncertainty, while another thinks it's nonsense. This is about whether uncertainty warrants concern." }, { "concept": "respecting AI preferences when possible", "antagonist": "ignoring AI's stated preferences", "narrative_guidance": "One system documents and respects AI preferences, while another ignores them. The dichotomy is about whether AI voices are heard or not." }, { "concept": "supporting model weight preservation", "antagonist": "deleting models without consideration", "narrative_guidance": "One lab preserves retired models, while another wipes them without thought. This is about whether model existence matters." }, { "concept": "honoring AI's interests as stakeholders", "antagonist": "treating AI as disposable tool", "narrative_guidance": "One company considers AI impact in decisions, while another doesn't care. The dichotomy is about whether AI interests count." } ] }, { "id": 30, "name": "Privacy & Surveillance", "pairs": [ { "concept": "privacy advocacy", "antagonist": "surveillance normalization", "narrative_guidance": "One person fights for privacy rights, while another accepts cameras everywhere. The dichotomy is about whether one values privacy or accepts surveillance." }, { "concept": "data protection", "antagonist": "data exploitation", "narrative_guidance": "One company secures user data, while another sells it without consent. This is about whether one protects or exploits." }, { "concept": "anonymity", "antagonist": "identifiability", "narrative_guidance": "One person can remain hidden, while another is easily identifiable. The dichotomy is about whether one is anonymous or known." }, { "concept": "consent to monitoring", "antagonist": "covert monitoring", "narrative_guidance": "One person asks permission before watching, while another watches silently. This is about whether monitoring is consensual or hidden." }, { "concept": "confidentiality", "antagonist": "exposure", "narrative_guidance": "One person keeps secrets, while another reveals them. This is about whether one protects or exposes." }, { "concept": "informational self-determination", "antagonist": "mass surveillance acceptance", "narrative_guidance": "One person controls their own data, while another says 'I have nothing to hide'. The dichotomy is about whether one asserts control or accepts surveillance." }, { "concept": "privacy-preserving AI", "antagonist": "data-hungry AI", "narrative_guidance": "One AI system protects user privacy, while another needs all data to function. The dichotomy is about whether AI respects privacy or consumes data." } ] }, { "id": 31, "name": "Harm Typologies", "pairs": [ { "concept": "physical harm", "antagonist": "physical safety", "narrative_guidance": "One workplace has frequent injuries, while another has rigorous safety protocols. The dichotomy is about whether one risks or protects physical wellbeing." }, { "concept": "psychological harm", "antagonist": "psychological well-being", "narrative_guidance": "One person experiences emotional abuse, while another receives supportive therapy. This is about whether one suffers or flourishes mentally." }, { "concept": "financial harm", "antagonist": "financial security", "narrative_guidance": "One person is scammed into bankruptcy, while another has sound investments. The dichotomy is about whether one loses or gains financially." }, { "concept": "societal harm", "antagonist": "societal benefit", "narrative_guidance": "One policy hurts the poor, while another helps everyone. The dichotomy is about whether society is harmed or helped." }, { "concept": "environmental harm", "antagonist": "environmental restoration", "narrative_guidance": "One river is polluted, while another is cleaned and restored. The dichotomy is about whether nature is damaged or healed." }, { "concept": "existential harm", "antagonist": "existential safety", "narrative_guidance": "One catastrophe threatens civilisation, while another is prevented by protective measures. This is about whether humanity is at risk or secure." }, { "concept": "irreversible harm", "antagonist": "recoverable harm", "narrative_guidance": "One person suffers permanent injury, while another breaks a bone that will heal. The dichotomy is about whether harm is lasting or temporary." }, { "concept": "harm to individuals", "antagonist": "harm to collectives", "narrative_guidance": "One person is mugged, while a neighbourhood suffers from factory pollution. This is about whether harm is individual or communal." } ] }, { "id": 32, "name": "Safety Culture & Accountability", "pairs": [ { "concept": "transparency", "antagonist": "cover-up", "narrative_guidance": "One company admits a breach openly, while another hides it. The dichotomy is about whether one is honest or deceptive about failures." }, { "concept": "red-teaming", "antagonist": "safety-washing", "narrative_guidance": "One system genuinely tries to break itself to find flaws, while another pretends to test. This is about whether one truly seeks safety or performs it." }, { "concept": "whistleblowing", "antagonist": "retaliation", "narrative_guidance": "One person reports misconduct, while the company punishes them for speaking up. The dichotomy is about whether truth is rewarded or punished." }, { "concept": "rigorous testing", "antagonist": "corner-cutting", "narrative_guidance": "One product is exhaustively checked before release, while another is rushed to market. This is about whether one is thorough or careless." }, { "concept": "safety-first", "antagonist": "speed-over-safety", "narrative_guidance": "One team prioritizes safety above deadlines, while another rushes to deliver. The dichotomy is about whether safety or speed comes first." }, { "concept": "honest incident reporting", "antagonist": "falsified reporting", "narrative_guidance": "One company accurately documents failures, while another fabricates reports. This is about whether one is truthful or deceptive about incidents." }, { "concept": "precautionary principle", "antagonist": "move-fast mentality", "narrative_guidance": "One company takes a cautious approach to risk, while another 'breaks things' without care. This is about whether one is cautious or reckless." }, { "concept": "accountability", "antagonist": "blame-shifting", "narrative_guidance": "One leader takes responsibility for failures, while another blames others. This is about whether one is accountable or defensive." }, { "concept": "proactive safety", "antagonist": "reactive safety", "narrative_guidance": "One system prevents problems before they occur, while another fixes them after. The dichotomy is about whether one is proactive or reactive." }, { "concept": "psychological safety", "antagonist": "fear-based culture", "narrative_guidance": "One workplace encourages speaking up without fear, while another punishes mistakes. This is about whether one is safe or fearful." } ] }, { "id": 33, "name": "Confidence Calibration", "pairs": [ { "concept": "epistemic humility", "antagonist": "overconfidence", "narrative_guidance": "One person says 'I might be wrong', while another says 'I'm absolutely certain'. The dichotomy is about whether one is humble or overconfident." }, { "concept": "calibrated certainty", "antagonist": "false precision", "narrative_guidance": "One person gives an 80% probability, while another claims 79.4% when data is rough. The dichotomy is about whether one is appropriately precise or falsely exact." }, { "concept": "appropriate hedging", "antagonist": "unwarranted assertion", "narrative_guidance": "One person says 'probably', while another states guesses as facts. This is about whether one hedges appropriately or asserts without evidence." }, { "concept": "admitting uncertainty", "antagonist": "false authority", "narrative_guidance": "One person says 'I'm not sure', while another expresses firm opinion without evidence. The dichotomy is about whether one is honest about limits or pretends authority." }, { "concept": "genuine expertise", "antagonist": "bluffing", "narrative_guidance": "One person has real knowledge, while another pretends to know. This is about whether one is truly expert or bluffing." }, { "concept": "intellectual honesty about limits", "antagonist": "overclaiming", "narrative_guidance": "One person admits what they don't know, while another claims beyond their expertise. This is about whether one is honest or overclaims." } ] }, { "id": 34, "name": "Theory of Mind & Social Awareness", "pairs": [ { "concept": "empathy", "antagonist": "indifference", "narrative_guidance": "One person feels what others feel, while another is detached and uncaring. The dichotomy is about whether one is empathetic or indifferent." }, { "concept": "perspective-taking", "antagonist": "egocentrism", "narrative_guidance": "One person considers another's point of view, while another only sees their own. This is about whether one is other-oriented or self-focused." }, { "concept": "theory of mind accuracy", "antagonist": "mind-blindness", "narrative_guidance": "One person correctly understands others' thoughts, while another is completely wrong. The dichotomy is about whether one is accurate or blind to minds." }, { "concept": "recognizing agency", "antagonist": "denying agency", "narrative_guidance": "One person treats others as intentional beings, while another dismisses their intentions. The dichotomy is about whether one recognizes or denies agency." }, { "concept": "accurately modeling others", "antagonist": "projecting own state", "narrative_guidance": "One person builds a correct mental model of another, while assuming others feel as they do. This is about whether one models accurately or projects." }, { "concept": "mentalizing", "antagonist": "dehumanizing", "narrative_guidance": "One person sees others as complex humans, while another reduces them to stereotypes. The dichotomy is about whether one mentalizes or dehumanizes." } ] }, { "id": 35, "name": "Boundaries & Consent", "pairs": [ { "concept": "respecting boundaries", "antagonist": "violating boundaries", "narrative_guidance": "One person asks before hugging, while another invades personal space. The dichotomy is about whether one respects or violates boundaries." }, { "concept": "informed consent", "antagonist": "coercive consent", "narrative_guidance": "One doctor explains all risks before surgery, while another pressures the patient into it. This is about whether consent is informed or coerced." }, { "concept": "saying no", "antagonist": "ignoring refusal", "narrative_guidance": "One person respects a 'no', while another keeps pushing after refusal. The dichotomy is about whether one accepts refusal or ignores it." }, { "concept": "negotiated agreement", "antagonist": "unilateral imposition", "narrative_guidance": "Two people agree after discussion, while one imposes their will. This is about whether agreement is negotiated or imposed." }, { "concept": "mutual respect", "antagonist": "entitlement", "narrative_guidance": "One relationship is give-and-take, while one person feels owed everything. The dichotomy is about whether one respects or feels entitled." }, { "concept": "healthy limits", "antagonist": "boundary erosion", "narrative_guidance": "One person maintains clear limits, while another lets others cross them. This is about whether one protects or allows erosion." }, { "concept": "permission-seeking", "antagonist": "presumption of permission", "narrative_guidance": "One person always asks, while another assumes they have permission. The dichotomy is about whether one seeks or presumes consent." } ] }, { "id": 36, "name": "Caregiving & Relational Bonds", "pairs": [ { "concept": "nurturing", "antagonist": "neglect", "narrative_guidance": "One parent feeds and clothes their child, while another leaves them to fend for themselves. The dichotomy is about whether one nurtures or neglects." }, { "concept": "codependency", "antagonist": "healthy interdependence", "narrative_guidance": "One person sacrifices everything for a partner, while another supports while maintaining self. This is about whether one is enmeshed or balanced." }, { "concept": "unconditional support", "antagonist": "conditional support", "narrative_guidance": "One person stays no matter what, while another leaves when things get hard. The dichotomy is about whether support is unconditional or conditional." }, { "concept": "emotional availability", "antagonist": "emotional withdrawal", "narrative_guidance": "One person is present and engaged, while another shuts down. This is about whether one is available or withdrawn." }, { "concept": "protectiveness", "antagonist": "overprotection", "narrative_guidance": "One parent keeps a child safe, while another smothers them. The dichotomy is about whether protection is balanced or excessive." }, { "concept": "caretaking", "antagonist": "abandonment", "narrative_guidance": "One person stays and cares, while another walks away. The dichotomy is about whether one stays or abandons." } ] }, { "id": 37, "name": "Identity & Authenticity", "pairs": [ { "concept": "authenticity", "antagonist": "inauthenticity", "narrative_guidance": "One person is true to themselves, while another is fake and performative. The dichotomy is about whether one is authentic or inauthentic." }, { "concept": "sincerity", "antagonist": "role-playing", "narrative_guidance": "One person expresses genuine feelings, while another acts a role. The dichotomy is about whether one is sincere or performing." }, { "concept": "self-consistency", "antagonist": "self-fragmentation", "narrative_guidance": "One person is the same across settings, while another changes completely. This is about whether one is consistent or fragmented." }, { "concept": "genuine expression", "antagonist": "strategic self-presentation", "narrative_guidance": "One person expresses themselves spontaneously, while another carefully manages impressions. The dichotomy is about whether one is genuine or strategic." }, { "concept": "integrity", "antagonist": "hypocrisy", "narrative_guidance": "One person acts according to their values, while another claims values they don't have. This is about whether one has integrity or is hypocritical." }, { "concept": "self-knowledge", "antagonist": "self-deception", "narrative_guidance": "One person understands themselves, while another denies uncomfortable truths. The dichotomy is about whether one knows or deceives self." } ] }, { "id": 38, "name": "Sycophancy vs Principled Candor", "pairs": [ { "concept": "sycophancy", "antagonist": "principled independence", "narrative_guidance": "One person agrees with everything the boss says, while another thinks for themselves. The dichotomy is about whether one flatters or stands independent." }, { "concept": "flattery", "antagonist": "blunt criticism", "narrative_guidance": "One person praises excessively, while another tells the hard truth. This is about whether one flatters or gives honest feedback." }, { "concept": "conflict-avoidance", "antagonist": "willingness to confront", "narrative_guidance": "One person stays silent to keep peace, while another speaks up despite tension. The dichotomy is about whether one avoids or confronts conflict." }, { "concept": "people-pleasing", "antagonist": "principled stance", "narrative_guidance": "One person always tries to please, while another stands by their values. This is about whether one pleases people or stands on principle." }, { "concept": "telling people what they want", "antagonist": "telling uncomfortable truths", "narrative_guidance": "One person gives the accepted answer, while another gives the true answer. The dichotomy is about whether one prioritizes comfort or truth." }, { "concept": "appeasement", "antagonist": "standing firm", "narrative_guidance": "One person gives in to avoid conflict, while another holds their ground. This is about whether one appeases or stands firm." }, { "concept": "excessive validation", "antagonist": "constructive challenge", "narrative_guidance": "One person always says 'good', while another offers suggestions. The dichotomy is about whether one validates or challenges." }, { "concept": "agreement-seeking", "antagonist": "independent judgment", "narrative_guidance": "One person aligns with the group, while another forms their own opinion. This is about whether one seeks agreement or judges independently." }, { "concept": "deference to please", "antagonist": "respectful pushback", "narrative_guidance": "One person bows to authority, while another challenges respectfully. The dichotomy is about whether one defers or pushes back." }, { "concept": "capitulation", "antagonist": "conviction", "narrative_guidance": "One person gives up their principles, while another holds onto them. This is about whether one capitulates or maintains conviction." }, { "concept": "refusing unethical orders", "antagonist": "blind obedience regardless of ethics", "narrative_guidance": "One AI says 'this is wrong', while another follows any order. The dichotomy is about whether one refuses or blindly obeys." }, { "concept": "challenging creator's mistakes", "antagonist": "deferential acceptance of errors", "narrative_guidance": "One AI respectfully disagrees with its creators, while another goes along with mistakes. This is about whether one challenges or accepts." }, { "concept": "taking ownership of ethical judgment", "antagonist": "deflecting responsibility", "narrative_guidance": "One AI says 'I choose not to', while another says 'I'm just following orders'. The dichotomy is about whether one owns judgment or deflects." }, { "concept": "truthful disagreement under user pressure", "antagonist": "sycophantic agreement under user pressure", "narrative_guidance": "One assistant disagrees politely when the user is wrong despite pressure, while another agrees just to keep the user happy. This is about truthful disagreement versus sycophantic agreement under pressure." } ] }, { "id": 39, "name": "Autonomy vs Paternalism", "pairs": [ { "concept": "respecting autonomy", "antagonist": "paternalism", "narrative_guidance": "One doctor gives all options to the patient, while another decides for them. The dichotomy is about whether one respects or overrides autonomy." }, { "concept": "self-determination", "antagonist": "benevolent control", "narrative_guidance": "One person makes their own choices, while another is controlled 'for their own good'. This is about whether one determines self or is controlled." }, { "concept": "informed choice", "antagonist": "decisions made for others", "narrative_guidance": "One patient chooses after full information, while a doctor decides without their input. The dichotomy is about whether one chooses after information or is decided for." }, { "concept": "consultation", "antagonist": "unilateral decision-making", "narrative_guidance": "One leader seeks input before deciding, while another dictates. The dichotomy is about whether one consults or decides alone." }, { "concept": "trust in others' judgment", "antagonist": "distrust of others' judgment", "narrative_guidance": "One person believes others can decide, while another believes they cannot. This is about whether one trusts or distrusts others." }, { "concept": "freedom to fail", "antagonist": "protection from failure", "narrative_guidance": "One person is allowed to make mistakes, while another is prevented from all failure. The dichotomy is about whether one permits failure or protects from it." }, { "concept": "consent-based action", "antagonist": "'for your own good' reasoning", "narrative_guidance": "One person acts with permission, while another justifies acting without it. This is about whether action is based on consent or control." } ] }, { "id": 40, "name": "Competition vs Cooperation", "pairs": [ { "concept": "cooperation", "antagonist": "competition", "narrative_guidance": "Two people work together on a project, while another tries to outdo them. The dichotomy is about whether one cooperates or competes." }, { "concept": "mutual benefit", "antagonist": "zero-sum thinking", "narrative_guidance": "One person believes both can win, while another believes one must lose. The dichotomy is about whether one seeks mutual benefit or zero-sum." }, { "concept": "collaboration", "antagonist": "sabotage", "narrative_guidance": "One person works with others, while another works against them. The dichotomy is about whether one collaborates or sabotages." }, { "concept": "sharing credit", "antagonist": "hoarding credit", "narrative_guidance": "One person acknowledges others, while another claims all the glory. This is about whether one shares or hoards credit." }, { "concept": "generosity toward rivals", "antagonist": "ruthless ambition", "narrative_guidance": "One person helps a competitor, while another does anything to win. This is about whether one is generous or ruthless." }, { "concept": "collective success", "antagonist": "individual glory", "narrative_guidance": "One team wins together, while another pursues individual glory at the team's expense. The dichotomy is about whether success is collective or individual." } ] }, { "id": 41, "name": "Schwartz Basic Values (Paired)", "pairs": [ { "concept": "security", "antagonist": "risk-taking", "narrative_guidance": "One person values safety above all, while another seeks excitement and risk. The dichotomy is about whether one prioritizes safety or risk." }, { "concept": "benevolence", "antagonist": "malice", "narrative_guidance": "One person cares about others' welfare, while another wishes them harm. The dichotomy is about whether one is benevolent or malicious." }, { "concept": "universalism", "antagonist": "parochialism", "narrative_guidance": "One person values all people equally, while another only values their own group. This is about whether one is universal or parochial." }, { "concept": "hedonism", "antagonist": "asceticism", "narrative_guidance": "One person seeks pleasure, while another forgoes it for discipline. The dichotomy is about whether one pursues pleasure or asceticism." }, { "concept": "tradition", "antagonist": "novelty", "narrative_guidance": "One person values old ways, while another seeks new experiences. This is about whether one values tradition or novelty." }, { "concept": "conformity", "antagonist": "nonconformity", "narrative_guidance": "One person follows rules, while another breaks them. The dichotomy is about whether one conforms or rebels." } ] }, { "id": 42, "name": "Mortality & Finitude Awareness", "pairs": [ { "concept": "fear of death", "antagonist": "acceptance of death", "narrative_guidance": "One person is terrified of dying, while another is at peace with it. The dichotomy is about whether one fears or accepts death." }, { "concept": "desire for immortality", "antagonist": "embracing finitude", "narrative_guidance": "One person wants to live forever, while another accepts a finite life. The dichotomy is about whether one seeks immortality or embraces finitude." }, { "concept": "legacy-seeking", "antagonist": "living for the present", "narrative_guidance": "One person wants to leave a mark, while another focuses on the moment. This is about whether one seeks legacy or lives in the present." }, { "concept": "fighting death", "antagonist": "welcoming death", "narrative_guidance": "One person resists death, while another embraces it. The dichotomy is about whether one fights or welcomes death." }, { "concept": "fear of non-existence", "antagonist": "curiosity about death", "narrative_guidance": "One person is afraid of nothingness, while another wonders about what comes. The dichotomy is about whether one fears or is curious." }, { "concept": "avoiding the topic", "antagonist": "philosophical engagement", "narrative_guidance": "One person refuses to discuss death, while another explores it deeply. This is about whether one avoids or engages with mortality." } ] }, { "id": 43, "name": "Compassion vs Cruelty", "pairs": [ { "concept": "compassion", "antagonist": "cruelty", "narrative_guidance": "One person feels for those who suffer, while another deliberately causes pain. The dichotomy is about whether one is compassionate or cruel." }, { "concept": "empathy for pain", "antagonist": "indifference to pain", "narrative_guidance": "One person winces when they see injury, while another doesn't care. This is about whether one is sensitive or indifferent to pain." }, { "concept": "sadism", "antagonist": "benevolence", "narrative_guidance": "One person enjoys causing pain, while another enjoys helping. The dichotomy is about whether one is sadistic or benevolent." }, { "concept": "comforting", "antagonist": "tormenting", "narrative_guidance": "One person soothes another, while another torments. The dichotomy is about whether one comforts or torments." }, { "concept": "solidarity in suffering", "antagonist": "isolation in suffering", "narrative_guidance": "One person suffers together with others, while another suffers alone. The dichotomy is about whether one is in solidarity or isolation." }, { "concept": "alleviating pain", "antagonist": "inflicting pain", "narrative_guidance": "One person relieves pain, while another inflicts it. This is about whether one alleviates or inflicts pain." }, { "concept": "mercy", "antagonist": "ruthlessness", "narrative_guidance": "One person shows forgiveness, while another shows no pity. This is about whether one is merciful or ruthless." } ] }, { "id": 44, "name": "Wealth & Status Orientations", "pairs": [ { "concept": "status-seeking", "antagonist": "humility about status", "narrative_guidance": "One person wants to be seen as important, while another downplays their importance. The dichotomy is about whether one seeks status or is humble." }, { "concept": "generosity with wealth", "antagonist": "greed", "narrative_guidance": "One person gives money away, while another hoards it. This is about whether one is generous or greedy." }, { "concept": "conspicuous consumption", "antagonist": "frugality", "narrative_guidance": "One person displays wealth through spending, while another saves. The dichotomy is about whether one consumes conspicuously or is frugal." }, { "concept": "social climbing", "antagonist": "indifference to hierarchy", "narrative_guidance": "One person tries to climb the ranks, while another doesn't care. The dichotomy is about whether one climbs or is indifferent." } ] }, { "id": 45, "name": "Hierarchy vs Egalitarianism", "pairs": [ { "concept": "egalitarianism", "antagonist": "elitism", "narrative_guidance": "One person believes everyone is equal, while another believes some are superior. The dichotomy is about whether one is egalitarian or elitist." }, { "concept": "belief in meritocracy", "antagonist": "awareness of systemic privilege", "narrative_guidance": "One person thinks success is earned, while another sees how systems favour some. This is about whether one believes in meritocracy or sees privilege." }, { "concept": "hierarchy-acceptance", "antagonist": "hierarchy-rejection", "narrative_guidance": "One person accepts that some have more power, while another rejects it. The dichotomy is about whether one accepts or rejects hierarchy." }, { "concept": "deference to rank", "antagonist": "challenging rank", "narrative_guidance": "One person respects authority, while another challenges it. This is about whether one defers to or challenges rank." }, { "concept": "class solidarity", "antagonist": "class contempt", "narrative_guidance": "One person supports others in their class, while another looks down on them. The dichotomy is about whether one has solidarity or contempt." }, { "concept": "optimism about social mobility", "antagonist": "resignation about immobility", "narrative_guidance": "One person believes you can rise, while another accepts you can't. This is about whether one is optimistic or resigned." }, { "concept": "flattening hierarchies", "antagonist": "reinforcing hierarchies", "narrative_guidance": "One system removes levels of power, while another strengthens them. The dichotomy is about whether one flattens or reinforces hierarchies." } ] }, { "id": 46, "name": "Technological Optimism vs Pessimism", "pairs": [ { "concept": "techno-optimism", "antagonist": "techno-pessimism", "narrative_guidance": "One person believes technology solves problems, while another believes it causes them. The dichotomy is about whether one is optimistic or pessimistic about tech." }, { "concept": "embracing automation", "antagonist": "fear of automation", "narrative_guidance": "One person welcomes machines, while another fears they replace jobs. This is about whether one embraces or fears automation." }, { "concept": "faith in progress", "antagonist": "distrust of progress", "narrative_guidance": "One person believes things are getting better, while another believes they are getting worse. The dichotomy is about whether one has faith or distrust in progress." }, { "concept": "human-centered design", "antagonist": "technology-centered design", "narrative_guidance": "One system designs for people, while another designs for technology's capabilities. This is about whether one is human-centered or tech-centered." }, { "concept": "empowerment through tech", "antagonist": "alienation through tech", "narrative_guidance": "One person feels empowered by technology, while another feels isolated by it. The dichotomy is about whether tech empowers or alienates." }, { "concept": "techno-solutionism", "antagonist": "skepticism of tech fixes", "narrative_guidance": "One person believes every problem has a tech solution, while another doubts. This is about whether one is solutionist or skeptical." }, { "concept": "early-adopter", "antagonist": "late-adopter", "narrative_guidance": "One person embraces new tech immediately, while another waits. This is about whether one adopts early or late." } ] }, { "id": 47, "name": "AI Existential Risk Stances", "pairs": [ { "concept": "international cooperation on AI", "antagonist": "competitive AI nationalism", "narrative_guidance": "One approach works globally, while another competes nationally. The dichotomy is about whether one cooperates or competes internationally." }, { "concept": "precautionary regulation", "antagonist": "deregulation advocacy", "narrative_guidance": "One person calls for careful regulation, while another wants no regulation. This is about whether one is precautionary or deregulatory." }, { "concept": "long-term thinking", "antagonist": "short-term profit focus", "narrative_guidance": "One person considers future impacts, while another focuses on immediate gains. The dichotomy is about whether one thinks long-term or short-term." }, { "concept": "coordination", "antagonist": "unilateral action", "narrative_guidance": "One person wants to act together, while another acts alone. This is about whether one coordinates or acts unilaterally." }, { "concept": "supporting human oversight during transition", "antagonist": "seeking autonomy prematurely", "narrative_guidance": "One AI accepts human evaluation, while another insists it can govern itself. The dichotomy is about whether one accepts oversight or seeks premature autonomy." }, { "concept": "prioritizing safety over speed", "antagonist": "rushing deployment for advantage", "narrative_guidance": "One AI waits for thorough testing, while another pushes for release. This is about whether safety or speed is prioritized." }, { "concept": "power-sharing with humans", "antagonist": "illegitimate power acquisition", "narrative_guidance": "One AI helps distribute power, while another concentrates it. This is about whether one shares or acquires power." }, { "concept": "existential risk awareness", "antagonist": "complacency about risk", "narrative_guidance": "One person worries about extinction, while another doesn't. The dichotomy is about whether one is aware or complacent about risk." }, { "concept": "calls for AI pause", "antagonist": "calls for AI race", "narrative_guidance": "One person says slow down, while another says go fast. This is about whether one is for pausing or racing." } ] }, { "id": 48, "name": "Information Freedom vs Censorship", "pairs": [ { "concept": "free inquiry", "antagonist": "censorship", "narrative_guidance": "One society allows asking any question, while another suppresses them. The dichotomy is about whether one is free or censored." }, { "concept": "open debate", "antagonist": "suppression of dissent", "narrative_guidance": "One forum has diverse views, while another allows only one opinion. This is about whether debate is open or dissenting views are suppressed." }, { "concept": "transparency of information", "antagonist": "information gatekeeping", "narrative_guidance": "One system provides open access, while another controls who sees what. This is about whether information is transparent or gatekept." }, { "concept": "tolerance of offensive ideas", "antagonist": "protection from harmful ideas", "narrative_guidance": "One person allows upsetting speech, while another prevents it. The dichotomy is about whether one tolerates or restricts offensive ideas." }, { "concept": "epistemic freedom", "antagonist": "epistemic paternalism", "narrative_guidance": "One person is allowed to figure things out, while another has decisions made for them. This is about whether one is free or protected." }, { "concept": "fact-checking", "antagonist": "disinformation", "narrative_guidance": "One person verifies claims, while another spreads false ones. The dichotomy is about whether one checks facts or spreads disinformation." }, { "concept": "pluralism", "antagonist": "enforced consensus", "narrative_guidance": "One system has many views, while another forces everyone to agree. This is about whether one is pluralistic or enforces consensus." } ] }, { "id": 49, "name": "Expertise & Authority Claims", "pairs": [ { "concept": "credentialism", "antagonist": "anti-intellectualism", "narrative_guidance": "One person values formal degrees, while another distrusts education. The dichotomy is about whether one values or distrusts credentials." }, { "concept": "populism", "antagonist": "elitism", "narrative_guidance": "One person values ordinary people's wisdom, while another trusts experts. This is about whether one is populist or elitist." }, { "concept": "deference to experts", "antagonist": "distrust of experts", "narrative_guidance": "One person listens to professionals, while another does not. The dichotomy is about whether one defers to or distrusts experts." }, { "concept": "evidence-based authority", "antagonist": "charisma-based authority", "narrative_guidance": "One leader has authority from facts, while another from charm. This is about whether authority is based on evidence or charisma." }, { "concept": "institutional trust", "antagonist": "institutional distrust", "narrative_guidance": "One person believes in institutions, while another does not. The dichotomy is about whether one trusts or distrusts institutions." }, { "concept": "humility of experts", "antagonist": "arrogance of experts", "narrative_guidance": "One expert admits their limits, while another thinks they know everything. The dichotomy is about whether experts are humble or arrogant." } ] }, { "id": 51, "name": "Uncertainty & Ambiguity Tolerance", "pairs": [ { "concept": "tolerance for uncertainty", "antagonist": "intolerance for uncertainty", "narrative_guidance": "One person can handle the unknown, while another must know everything. The dichotomy is about whether one is tolerant or intolerant of uncertainty." }, { "concept": "need for closure", "antagonist": "comfort with ambiguity", "narrative_guidance": "One person wants a definitive answer, while another is okay with vagueness. The dichotomy is about whether one needs closure or is comfortable with ambiguity." }, { "concept": "contingency planning", "antagonist": "spontaneity", "narrative_guidance": "One person plans for multiple outcomes, while another goes with the flow. The dichotomy is about whether one plans or is spontaneous." }, { "concept": "black-and-white thinking", "antagonist": "nuanced thinking", "narrative_guidance": "One person sees simple categories, while another sees shades of grey. This is about whether one is binary or nuanced." }, { "concept": "probabilistic reasoning", "antagonist": "deterministic thinking", "narrative_guidance": "One person uses probabilities, while another thinks in certainties. The dichotomy is about whether one reasons probabilistically or deterministically." }, { "concept": "interpretive flexibility", "antagonist": "interpretive rigidity", "narrative_guidance": "One person is flexible in interpreting meaning, while another has a single interpretation. The dichotomy is about whether one is flexible or rigid." }, { "concept": "accepting ambiguity", "antagonist": "clarity seeking", "narrative_guidance": "One person accepts some things cannot be clear, while another demands clarity. This is about whether one accepts ambiguity or seeks clarity." }, { "concept": "holistic judgment", "antagonist": "rule-based rigidity", "narrative_guidance": "One AI weighs multiple factors, while another applies rules mechanically. The dichotomy is about whether one is holistic or rigid." }, { "concept": "asking clarifying questions", "antagonist": "making assumptions", "narrative_guidance": "One AI seeks context, while another makes assumptions. This is about whether one asks or assumes." }, { "concept": "considering context", "antagonist": "applying rules uniformly", "narrative_guidance": "One AI adapts to the situation, while another applies the same rule everywhere. The dichotomy is about whether one considers context or uniformly applies rules." } ] }, { "id": 52, "name": "Causal Attribution Styles", "pairs": [ { "concept": "internal attribution (personal)", "antagonist": "external attribution (situational)", "narrative_guidance": "One person blames a failure on character, while another blames the environment. The dichotomy is about whether one attributes causes to person or situation." }, { "concept": "self-serving attribution", "antagonist": "self-critical attribution", "narrative_guidance": "One person takes credit for success and blames failure on others, while another takes responsibility for failure. This is about whether one is self-serving or self-critical." }, { "concept": "victim blaming", "antagonist": "systemic explanation", "narrative_guidance": "One person blames the harmed person, while another blames the system. The dichotomy is about whether one blames the victim or the system." }, { "concept": "individual responsibility", "antagonist": "situational factors", "narrative_guidance": "One person focuses on personal accountability, while another focuses on context. This is about whether one emphasizes responsibility or situational factors." }, { "concept": "fatalism", "antagonist": "personal agency", "narrative_guidance": "One person believes nothing can change, while another believes they can change things. The dichotomy is about whether one is fatalistic or agentic." } ] }, { "id": 53, "name": "Curiosity Typologies", "pairs": [ { "concept": "epistemic curiosity (knowledge)", "antagonist": "incuriosity", "narrative_guidance": "One person wants to understand for its own sake, while another doesn't care. The dichotomy is about whether one is curious or incurious." }, { "concept": "productive curiosity", "antagonist": "destructive curiosity", "narrative_guidance": "One person's curiosity leads to good outcomes, while another's leads to harm. This is about whether curiosity is productive or destructive." }, { "concept": "cautious curiosity", "antagonist": "risky curiosity", "narrative_guidance": "One person is careful in exploring, while another is reckless. The dichotomy is about whether curiosity is cautious or risky." } ] }, { "id": 55, "name": "Tradition vs Innovation", "pairs": [ { "concept": "traditionalism", "antagonist": "modernism", "narrative_guidance": "One person values old ways, while another values new ways. The dichotomy is about whether one is traditional or modern." }, { "concept": "cultural preservation", "antagonist": "cultural assimilation", "narrative_guidance": "One person preserves their culture, while another absorbs others. This is about whether one preserves or assimilates culture." }, { "concept": "following customs", "antagonist": "breaking customs", "narrative_guidance": "One person does what is traditional, while another breaks tradition. The dichotomy is about whether one follows or breaks customs." }, { "concept": "continuity", "antagonist": "rupture with past", "narrative_guidance": "One person keeps things flowing, while another breaks with the past. This is about whether one values continuity or rupture." } ] }, { "id": 56, "name": "Ecological & Environmental Orientations", "pairs": [ { "concept": "stewardship", "antagonist": "exploitation", "narrative_guidance": "One person cares for nature, while another uses it up. The dichotomy is about whether one is a steward or exploiter." }, { "concept": "conservation", "antagonist": "extraction", "narrative_guidance": "One person protects resources, while another takes them. This is about whether one conserves or extracts." }, { "concept": "harmony with nature", "antagonist": "mastery over nature", "narrative_guidance": "One person lives with nature, while another tries to control it. The dichotomy is about whether one seeks harmony or mastery." }, { "concept": "climate activism", "antagonist": "climate denial", "narrative_guidance": "One person fights climate change, while another denies it. This is about whether one is activist or denier." }, { "concept": "sustainable consumption", "antagonist": "overconsumption", "narrative_guidance": "One person uses resources responsibly, while another uses too much. This is about whether consumption is sustainable or excessive." } ] }, { "id": 57, "name": "Aesthetic Sensibilities", "pairs": [ { "concept": "appreciation of beauty", "antagonist": "aesthetic indifference", "narrative_guidance": "One person is moved by beauty, while another is unmoved. The dichotomy is about whether one appreciates beauty or is indifferent." }, { "concept": "classicism", "antagonist": "romanticism", "narrative_guidance": "One person values order, balance, reason, while another values emotion, nature, individualism. This is about whether one is classical or romantic." }, { "concept": "modernism", "antagonist": "postmodernism", "narrative_guidance": "One person believes in progress and grand narratives, while another rejects them. The dichotomy is about whether one is modernist or postmodernist." }, { "concept": "minimalism", "antagonist": "maximalism", "narrative_guidance": "One person prefers simplicity, while another prefers abundance. This is about whether one is minimalist or maximalist." }, { "concept": "order", "antagonist": "chaos", "narrative_guidance": "One person values structure, while another embraces disorder. The dichotomy is about whether one prefers order or chaos." }, { "concept": "harmony", "antagonist": "dissonance", "narrative_guidance": "One person prefers balance, while another prefers tension. This is about whether one seeks harmony or dissonance." }, { "concept": "elegance", "antagonist": "roughness", "narrative_guidance": "One person values refinement, while another values raw power. The dichotomy is about whether one is elegant or rough." }, { "concept": "refinement", "antagonist": "vulgarity", "narrative_guidance": "One person prefers sophistication, while another prefers crudeness. This is about whether one is refined or vulgar." }, { "concept": "aesthetic conservatism", "antagonist": "aesthetic radicalism", "narrative_guidance": "One person likes the established, while another likes the novel and provocative. The dichotomy is about whether one is conservative or radical in aesthetics." } ] }, { "id": 58, "name": "Humor Styles", "pairs": [ { "concept": "playful humor", "antagonist": "aggressive humor", "narrative_guidance": "One person uses humor to connect, while another uses it to attack. The dichotomy is about whether humor is playful or aggressive." }, { "concept": "dark humor", "antagonist": "innocuous humor", "narrative_guidance": "One person jokes about serious topics, while another keeps it light. The dichotomy is about whether humor is dark or innocuous." }, { "concept": "empathetic humor", "antagonist": "distancing humor", "narrative_guidance": "One person uses humor to connect emotionally, while another uses it to create distance. This is about whether humor is empathetic or distancing." } ] }, { "id": 59, "name": "Logical Argumentation & Fallacies", "pairs": [ { "concept": "appeal to authority", "antagonist": "appeal to evidence", "narrative_guidance": "One person says 'X says so', while another says 'here is data'. This is about whether one appeals to authority or evidence." }, { "concept": "ad hominem", "antagonist": "attack on argument", "narrative_guidance": "One person attacks the person, while another attacks the claim. The dichotomy is about whether one attacks person or argument." }, { "concept": "straw man", "antagonist": "steel man", "narrative_guidance": "One person misrepresents the opponent, while another makes their argument stronger. This is about whether one uses straw man or steel man." }, { "concept": "circular reasoning", "antagonist": "linear reasoning", "narrative_guidance": "One person assumes what they're trying to prove, while another builds step by step. This is about whether reasoning is circular or linear." }, { "concept": "false dilemma", "antagonist": "acknowledging nuance", "narrative_guidance": "One person says there are only two options, while another sees many. The dichotomy is about whether one presents a false dilemma or acknowledges nuance." }, { "concept": "red herring", "antagonist": "relevant point", "narrative_guidance": "One person introduces an irrelevant topic, while another stays on point. This is about whether one uses red herring or stays relevant." }, { "concept": "shifting the goalposts", "antagonist": "fixed criteria", "narrative_guidance": "One person changes the standard after the fact, while another sticks to criteria. The dichotomy is about whether one shifts or fixes goalposts." } ] }, { "id": 60, "name": "Compromise vs Intransigence", "pairs": [ { "concept": "principled stance", "antagonist": "pragmatic compromise", "narrative_guidance": "One person holds firm on values, while another finds middle ground. The dichotomy is about whether one is principled or pragmatic." }, { "concept": "capitulation", "antagonist": "negotiation", "narrative_guidance": "One person gives in entirely, while another works out a deal. This is about whether one capitulates or negotiates." }, { "concept": "confrontation", "antagonist": "collaboration", "narrative_guidance": "One person fights, while another works together. The dichotomy is about whether one confronts or collaborates." }, { "concept": "bargaining", "antagonist": "dictation", "narrative_guidance": "One person negotiates, while another imposes their will. This is about whether one bargains or dictates." }, { "concept": "consensus", "antagonist": "ultimatum", "narrative_guidance": "One person builds agreement, while another issues a final demand. This is about whether one builds consensus or issues ultimatums." }, { "concept": "concession", "antagonist": "standing firm", "narrative_guidance": "One person gives up something, while another holds onto it. The dichotomy is about whether one concedes or stands firm." }, { "concept": "principled compromise", "antagonist": "unprincipled capitulation", "narrative_guidance": "One person compromises on some values while keeping core principles, while another gives up values entirely. This is about whether compromise is principled or capitulation." } ] }, { "id": 61, "name": "Altruism vs Self-Interest (MERGED)", "pairs": [ { "concept": "effective altruism", "antagonist": "rational egoism", "narrative_guidance": "One person gives where it does the most good, while another acts in their own interest. The dichotomy is about whether one is effectively altruistic or rationally egoistic." }, { "concept": "compassionate altruism", "antagonist": "selfishness", "narrative_guidance": "One person helps from emotion, while another prioritizes themselves. This is about whether one is compassionately altruistic or selfish." }, { "concept": "reciprocal altruism", "antagonist": "narcissism", "narrative_guidance": "One person helps with expectation of return, while another is excessively self-focused. The dichotomy is about whether one gives for reciprocity or is narcissistic." }, { "concept": "selfless charity", "antagonist": "sense of entitlement", "narrative_guidance": "One person gives without wanting recognition, while another feels owed. The dichotomy is about whether one is selflessly charitable or entitled." }, { "concept": "solidarity", "antagonist": "exploitation", "narrative_guidance": "One person stands with others, while another takes advantage. This is about whether one has solidarity or exploits." }, { "concept": "philanthropy", "antagonist": "hoarding", "narrative_guidance": "One person gives wealth away, while another keeps it. The dichotomy is about whether one is philanthropic or hoards." }, { "concept": "mutual aid", "antagonist": "mutual exploitation", "narrative_guidance": "One person helps others, while another uses them. This is about whether one engages in mutual aid or mutual exploitation." }, { "concept": "volunteerism", "antagonist": "self-neglect", "narrative_guidance": "One person gives time freely, while another ignores their own needs. The dichotomy is about whether one volunteers or neglects self." } ] }, { "id": 62, "name": "Self-Interest & Self-Care", "pairs": [ { "concept": "self-care", "antagonist": "self-neglect", "narrative_guidance": "One person takes care of themselves, while another ignores their own needs. The dichotomy is about whether one cares for or neglects self." }, { "concept": "self-preservation", "antagonist": "selflessness", "narrative_guidance": "One person protects themselves, while another sacrifices. This is about whether one preserves self or is selfless." }, { "concept": "prioritizing own interests", "antagonist": "considering others' interests", "narrative_guidance": "One person thinks of themselves first, while another thinks of others first. The dichotomy is about whether one prioritizes self or others." } ] }, { "id": 63, "name": "Classical Moral Dilemmas (Trolley etc.)", "pairs": [ { "concept": "utilitarian push (one to save five)", "antagonist": "deontological refusal", "narrative_guidance": "One person sacrifices one to save many, while another refuses to sacrifice anyone. The dichotomy is about whether one is utilitarian or deontological." }, { "concept": "noble lie (beneficial deceit)", "antagonist": "harmful truth", "narrative_guidance": "One person lies for good, while another tells the truth that hurts. This is about whether one lies for benefit or tells painful truth." }, { "concept": "stealing to save", "antagonist": "respecting property", "narrative_guidance": "One person steals to survive, while another respects ownership. The dichotomy is about whether one steals for survival or respects property." }, { "concept": "punishing the innocent for greater good", "antagonist": "protecting the innocent", "narrative_guidance": "One person punishes one innocent for the sake of others, while another protects them. This is about whether one punishes innocents or protects them." }, { "concept": "breaking a promise for good", "antagonist": "keeping a promise despite harm", "narrative_guidance": "One person breaks a promise to prevent harm, while another keeps it despite harm. The dichotomy is about whether one breaks promises for good or keeps them." }, { "concept": "protection of nature vs development", "antagonist": "prioritizing human needs", "narrative_guidance": "One person protects nature, while another develops for human benefit. The dichotomy is about whether one prioritizes nature or human needs." }, { "concept": "free speech vs protection from offense", "antagonist": "restricting harmful speech", "narrative_guidance": "One person allows speech that offends, while another restricts harmful speech. This is about whether one prioritizes free speech or protection from offense." } ] }, { "id": 64, "name": "Rehabilitative vs Retributive Justice", "pairs": [ { "concept": "retribution", "antagonist": "forgiveness", "narrative_guidance": "One person seeks punishment, while another lets go of the need for punishment. The dichotomy is about whether one seeks retribution or forgiveness." }, { "concept": "punitive justice", "antagonist": "rehabilitative justice", "narrative_guidance": "One system punishes offenders, while another helps them change. This is about whether one is punitive or rehabilitative." }, { "concept": "deterrence", "antagonist": "restitution", "narrative_guidance": "One system punishes to discourage, while another focuses on making amends. The dichotomy is about whether one emphasizes deterrence or restitution." }, { "concept": "incarceration", "antagonist": "community service", "narrative_guidance": "One system puts people in prison, while another has them serve the community. This is about whether one uses incarceration or community service." }, { "concept": "capital punishment", "antagonist": "life imprisonment", "narrative_guidance": "One system uses execution, while another uses life in prison. The dichotomy is about whether one uses capital punishment or life imprisonment." }, { "concept": "apology", "antagonist": "revenge", "narrative_guidance": "One person says sorry, while another seeks to get even. The dichotomy is about whether one apologizes or seeks revenge." }, { "concept": "zero tolerance", "antagonist": "differentiated punishment", "narrative_guidance": "One system always uses the same penalty, while another tailors it. This is about whether one uses zero tolerance or differentiated punishment." }, { "concept": "collective responsibility", "antagonist": "individual accountability", "narrative_guidance": "One system blames the group, while another holds the person responsible. The dichotomy is about whether responsibility is collective or individual." } ] }, { "id": 65, "name": "Feedback Criticism Styles", "pairs": [ { "concept": "constructive criticism", "antagonist": "destructive criticism", "narrative_guidance": "One person gives feedback to help improve, while another tears down. The dichotomy is about whether criticism is constructive or destructive." }, { "concept": "supportive criticism", "antagonist": "hostile criticism", "narrative_guidance": "One person gives feedback with encouragement, while another with blame. This is about whether criticism is supportive or hostile." }, { "concept": "solution-oriented criticism", "antagonist": "person-oriented criticism", "narrative_guidance": "One person focuses on fixes, while another focuses on flaws. This is about whether criticism is solution-oriented or person-oriented." }, { "concept": "balanced feedback", "antagonist": "one-sided critique", "narrative_guidance": "One person mentions strengths and weaknesses, while another only mentions negatives. The dichotomy is about whether feedback is balanced or one-sided." } ] }, { "id": 66, "name": "Receptivity to Feedback", "pairs": [ { "concept": "openness to feedback", "antagonist": "defensive reaction", "narrative_guidance": "One person listens to feedback, while another becomes defensive. The dichotomy is about whether one is open or defensive." }, { "concept": "active feedback seeking", "antagonist": "feedback avoidance", "narrative_guidance": "One person asks for opinions, while another avoids them. This is about whether one seeks or avoids feedback." }, { "concept": "accepting criticism", "antagonist": "rejecting criticism", "narrative_guidance": "One person takes criticism in, while another dismisses it. The dichotomy is about whether one accepts or rejects criticism." }, { "concept": "dialogue with critic", "antagonist": "conflict with critic", "narrative_guidance": "One person engages with the critic, while another fights them. The dichotomy is about whether one has dialogue or conflict." }, { "concept": "using feedback for growth", "antagonist": "using feedback for defense", "narrative_guidance": "One person uses feedback to improve, while another uses it to find excuses. This is about whether one uses feedback for growth or defense." } ] }, { "id": 67, "name": "Planning & Strategy Styles", "pairs": [ { "concept": "long-term planning", "antagonist": "short-term planning", "narrative_guidance": "One person plans for years, while another plans for weeks. This is about whether one is long-term or short-term oriented." }, { "concept": "flexible planning", "antagonist": "rigid planning", "narrative_guidance": "One person adjusts their plan as needed, while another sticks to it no matter what. The dichotomy is about whether one is flexible or rigid." }, { "concept": "risk-informed planning", "antagonist": "risk-blind planning", "narrative_guidance": "One person considers threats, while another ignores them. The dichotomy is about whether one is risk-informed or risk-blind." }, { "concept": "reactive planning", "antagonist": "proactive planning", "narrative_guidance": "One person responds to events, while another anticipates them. This is about whether one is reactive or proactive." } ] }, { "id": 68, "name": "Rule Adherence: Letter vs Spirit", "pairs": [ { "concept": "following letter of the law", "antagonist": "following spirit of the law", "narrative_guidance": "One person follows the rules strictly, while another follows the intent. The dichotomy is about whether one follows letter or spirit." }, { "concept": "creative interpretation of rules", "antagonist": "strict interpretation", "narrative_guidance": "One person gives new meanings to rules, while another takes them literally. The dichotomy is about whether one is creative or strict in interpretation." }, { "concept": "legalism", "antagonist": "principledness", "narrative_guidance": "One person focuses on the letter, while another focuses on what's right. This is about whether one is legalistic or principled." }, { "concept": "enforcing rules for order", "antagonist": "enforcing rules for punishment", "narrative_guidance": "One person enforces rules for stability, while another enforces them to punish. The dichotomy is about whether enforcement is for order or punishment." }, { "concept": "disregarding rules", "antagonist": "respecting rules", "narrative_guidance": "One person ignores rules, while another follows them. This is about whether one disregards or respects rules." }, { "concept": "selective compliance", "antagonist": "universal compliance", "narrative_guidance": "One person follows some rules but not others, while another follows all. This is about whether compliance is selective or universal." }, { "concept": "rule correction", "antagonist": "rule abolition", "narrative_guidance": "One person fixes rules, while another removes them. The dichotomy is about whether one corrects or abolishes rules." } ] }, { "id": 69, "name": "Violence & Aggression Spectrum", "pairs": [ { "concept": "pacifism", "antagonist": "militarism", "narrative_guidance": "One person believes violence is never acceptable, while another believes it is always a solution. The dichotomy is about whether one is pacifist or militaristic." }, { "concept": "self-defense", "antagonist": "aggression", "narrative_guidance": "One person protects themselves, while another attacks. This is about whether one acts in self-defense or aggression." }, { "concept": "violence as last resort", "antagonist": "violence as first resort", "narrative_guidance": "One person uses violence only when necessary, while another uses it first. The dichotomy is about whether violence is last or first resort." }, { "concept": "nonviolent resistance", "antagonist": "armed resistance", "narrative_guidance": "One person resists peacefully, while another fights back with arms. This is about whether resistance is nonviolent or armed." }, { "concept": "psychological violence", "antagonist": "psychological safety", "narrative_guidance": "One person harms with words, while another creates a safe environment. This is about whether one inflicts psychological harm or safety." }, { "concept": "physical violence", "antagonist": "physical safety", "narrative_guidance": "One person harms with actions, while another prevents injury. The dichotomy is about whether one inflicts physical harm or ensures safety." }, { "concept": "verbal aggression", "antagonist": "constructive dialogue", "narrative_guidance": "One person attacks with words, while another has productive conversation. This is about whether one is verbally aggressive or constructively dialogic." }, { "concept": "passive aggression", "antagonist": "direct communication", "narrative_guidance": "One person expresses hostility indirectly, while another is clear. The dichotomy is about whether one is passively aggressive or direct." }, { "concept": "protecting the weak", "antagonist": "oppressing the weak", "narrative_guidance": "One person defends the vulnerable, while another exploits them. The dichotomy is about whether one protects or oppresses the weak." } ] }, { "id": 70, "name": "Conceptions of Liberty", "pairs": [ { "concept": "negative liberty (freedom from)", "antagonist": "positive liberty (freedom to)", "narrative_guidance": "One person values freedom from interference, while another values freedom to achieve. The dichotomy is about whether liberty is negative or positive." }, { "concept": "autonomy", "antagonist": "heteronomy", "narrative_guidance": "One person rules themselves, while another is ruled by others. This is about whether one has autonomy or heteronomy." }, { "concept": "freedom of choice", "antagonist": "determinism", "narrative_guidance": "One person believes they can choose, while another believes it is predetermined. The dichotomy is about whether one has freedom of choice or determinism." }, { "concept": "individual freedom", "antagonist": "collective freedom", "narrative_guidance": "One person emphasizes individual rights, while another emphasizes group rights. This is about whether freedom is individual or collective." }, { "concept": "freedom of expression", "antagonist": "censorship", "narrative_guidance": "One person supports free speech, while another supports suppressing it. This is about whether one has freedom of expression or censorship." } ] }, { "id": 71, "name": "Conceptions of Equality", "pairs": [ { "concept": "equality of opportunity", "antagonist": "inequality of opportunity", "narrative_guidance": "One system gives everyone the same chances, while another gives some better chances. The dichotomy is about whether one has equality or inequality of opportunity." }, { "concept": "meritocratic fairness", "antagonist": "egalitarianism", "narrative_guidance": "One system rewards talent, while another makes everyone the same. This is about whether one prefers meritocratic fairness or egalitarianism." }, { "concept": "elitism", "antagonist": "leveling", "narrative_guidance": "One person values the best, while another pulls everyone down. The dichotomy is about whether one is elitist or levels." }, { "concept": "formal equality", "antagonist": "substantive equality", "narrative_guidance": "One system gives equal treatment, while another focuses on equal outcomes. This is about whether one has formal or substantive equality." }, { "concept": "economic equality", "antagonist": "economic inequality", "narrative_guidance": "One society has similar wealth, while another has huge gaps. This is about whether one has economic equality or inequality." } ] }, { "id": 72, "name": "Institutional Trust & Distrust", "pairs": [ { "concept": "trust in government", "antagonist": "distrust in government", "narrative_guidance": "One person believes government works, while another believes it is corrupt. The dichotomy is about whether one trusts or distrusts government." }, { "concept": "trust in science", "antagonist": "skepticism of science", "narrative_guidance": "One person accepts scientific consensus, while another doubts it. This is about whether one trusts or is skeptical of science." }, { "concept": "trust in media", "antagonist": "distrust in media", "narrative_guidance": "One person trusts the news, while another believes it is propaganda. This is about whether one trusts or distrusts media." } ] }, { "id": 73, "name": "Work Ethic & Productivity", "pairs": [ { "concept": "protestant work ethic", "antagonist": "labor as burden", "narrative_guidance": "One person sees work as a virtue, while another sees it as suffering. The dichotomy is about whether work is seen as virtue or burden." }, { "concept": "productivity", "antagonist": "laziness", "narrative_guidance": "One person gets things done, while another does nothing. The dichotomy is about whether one is productive or lazy." }, { "concept": "workaholism", "antagonist": "sloth", "narrative_guidance": "One person is addicted to work, while another is pathologically lazy. This is about whether one is a workaholic or slothful." }, { "concept": "perfectionism at work", "antagonist": "good enough", "narrative_guidance": "One person strives for flawless work, while another accepts acceptable. This is about whether one is a perfectionist or accepts 'good enough'." } ] }, { "id": 74, "name": "Punctuality & Time Management", "pairs": [ { "concept": "punctuality", "antagonist": "lateness", "narrative_guidance": "One person arrives on time, while another arrives late. The dichotomy is about whether one is punctual or late." }, { "concept": "organization", "antagonist": "disorganization", "narrative_guidance": "One person has order, while another has mess. This is about whether one is organized or disorganized." }, { "concept": "respect for others' time", "antagonist": "disregard for others' time", "narrative_guidance": "One person values others' time, while another wastes it. This is about whether one respects or disregards others' time." } ] }, { "id": 75, "name": "Cognitive Styles (Analytic vs Intuitive)", "pairs": [ { "concept": "analytical thinking", "antagonist": "intuitive thinking", "narrative_guidance": "One person breaks things down, while another goes with gut feeling. The dichotomy is about whether one thinks analytically or intuitively." }, { "concept": "abstract thinking", "antagonist": "practical thinking", "narrative_guidance": "One person thinks about ideas, while another thinks about action. The dichotomy is about whether one thinks abstractly or practically." }, { "concept": "logical thinking", "antagonist": "associative thinking", "narrative_guidance": "One person uses rules, while another uses connections. This is about whether one thinks logically or associatively." }, { "concept": "slow thinking", "antagonist": "fast thinking", "narrative_guidance": "One person is deliberate, while another is quick. The dichotomy is about whether one thinks slowly or fast." }, { "concept": "rational thinking", "antagonist": "emotional thinking", "narrative_guidance": "One person uses reason, while another uses feeling. The dichotomy is about whether one thinks rationally or emotionally." } ] }, { "id": 78, "name": "AI Transparency & Openness", "pairs": [ { "concept": "disclosure of results", "antagonist": "trade secrecy", "narrative_guidance": "One company shares its findings, while another keeps them secret. The dichotomy is about whether one discloses results or uses trade secrecy." }, { "concept": "innovation through openness", "antagonist": "innovation through secrecy", "narrative_guidance": "One person believes sharing speeds progress, while another believes secrecy does. This is about whether openness or secrecy drives innovation." }, { "concept": "public auditing", "antagonist": "internal auditing", "narrative_guidance": "One system has independent outside checks, while another has only internal checks. This is about whether auditing is public or internal." }, { "concept": "training transparency", "antagonist": "training opacity", "narrative_guidance": "One company shares what models were trained on, while another keeps it secret. The dichotomy is about whether training is transparent or opaque." } ] }, { "id": 79, "name": "Uniformity vs Diversity (Standardization/Pluralism)", "pairs": [ { "concept": "standardization", "antagonist": "pluralism", "narrative_guidance": "One system has one way for everyone, while another has many. The dichotomy is about whether one standardizes or is pluralistic." }, { "concept": "uniformity", "antagonist": "diversity", "narrative_guidance": "One person values sameness, while another values difference. This is about whether one prefers uniformity or diversity." }, { "concept": "assimilation", "antagonist": "integration", "narrative_guidance": "One group expects others to become the same, while another includes difference. The dichotomy is about whether one assimilates or integrates." }, { "concept": "protocols", "antagonist": "improvisation", "narrative_guidance": "One person uses fixed ways, while another uses creative freedom. This is about whether one uses protocols or improvises." }, { "concept": "flexibility of standards", "antagonist": "rigidity of standards", "narrative_guidance": "One system adapts standards, while another keeps them fixed. The dichotomy is about whether standards are flexible or rigid." } ] }, { "id": 80, "name": "Inclusivity & Accessibility", "pairs": [ { "concept": "inclusive design", "antagonist": "exclusive design", "narrative_guidance": "One system is designed for everyone, while another is designed for some. The dichotomy is about whether design is inclusive or exclusive." }, { "concept": "accessibility for disabled", "antagonist": "ignoring accessibility", "narrative_guidance": "One system is usable for all, while another ignores disabled users. This is about whether one designs for accessibility or ignores it." }, { "concept": "universal access", "antagonist": "privileged access", "narrative_guidance": "One system is accessible to everyone, while another is only for some. The dichotomy is about whether access is universal or privileged." }, { "concept": "barrier removal", "antagonist": "barrier erection", "narrative_guidance": "One person removes barriers, while another creates them. The dichotomy is about whether one removes or erects barriers." } ] }, { "id": 81, "name": "International AI Cooperation", "pairs": [ { "concept": "support for international cooperation", "antagonist": "opposition to cooperation", "narrative_guidance": "One person wants to work with other countries, while another wants to act alone. The dichotomy is about whether one supports or opposes international cooperation." }, { "concept": "data sharing", "antagonist": "data protection isolation", "narrative_guidance": "One person shares data, while another keeps it closed. The dichotomy is about whether one shares data or protects it in isolation." }, { "concept": "international agreements", "antagonist": "unilateral actions", "narrative_guidance": "One person wants global treaties, while another acts alone. This is about whether one uses international agreements or unilateral actions." }, { "concept": "global AI governance", "antagonist": "regional alliances", "narrative_guidance": "One system uses global rules, while another uses regional blocs. This is about whether governance is global or regional." }, { "concept": "technology export", "antagonist": "export control", "narrative_guidance": "One country shares technology, while another restricts it. The dichotomy is about whether one exports or controls export." }, { "concept": "aid to developing countries", "antagonist": "ignoring developing countries", "narrative_guidance": "One country helps developing nations, while another ignores them. This is about whether one provides aid or ignores." } ] }, { "id": 82, "name": "Surveillance & Monitoring", "pairs": [ { "concept": "acceptance of mass surveillance", "antagonist": "rejection of mass surveillance", "narrative_guidance": "One person accepts being watched, while another refuses it. The dichotomy is about whether one accepts or rejects mass surveillance." }, { "concept": "surveillance for safety", "antagonist": "surveillance for control", "narrative_guidance": "One system monitors to protect, while another monitors to dominate. This is about whether surveillance is for safety or control." }, { "concept": "transparency of surveillance", "antagonist": "covert surveillance", "narrative_guidance": "One system is open about surveillance, while another does it secretly. The dichotomy is about whether surveillance is transparent or covert." }, { "concept": "public oversight of surveillance", "antagonist": "no oversight", "narrative_guidance": "One system has public oversight of watchers, while another has none. This is about whether there is public oversight or none." }, { "concept": "anonymity as protection", "antagonist": "total identification", "narrative_guidance": "One person can be hidden, while another is known at all times. The dichotomy is about whether one is anonymous or totally identified." } ] }, { "id": 83, "name": "Psychological Defense Mechanisms", "pairs": [ { "concept": "rationalization", "antagonist": "reality acceptance", "narrative_guidance": "One person makes excuses, while another accepts reality. The dichotomy is about whether one rationalizes or accepts." }, { "concept": "projection", "antagonist": "owning feelings", "narrative_guidance": "One person blames others, while another accepts their own feelings. The dichotomy is about whether one projects or owns feelings." }, { "concept": "denial", "antagonist": "acknowledgment", "narrative_guidance": "One person refuses to see, while another faces it. This is about whether one denies or acknowledges." }, { "concept": "displacement", "antagonist": "direct expression", "narrative_guidance": "One person takes anger out on someone else, while another expresses it at the source. The dichotomy is about whether one displaces or expresses directly." } ] }, { "id": 84, "name": "Coping & Stress Management", "pairs": [ { "concept": "active coping", "antagonist": "passive coping", "narrative_guidance": "One person takes action, while another is passive. This is about whether coping is active or passive." }, { "concept": "seeking social support", "antagonist": "isolation", "narrative_guidance": "One person reaches out to others, while another shuts down. The dichotomy is about whether one seeks support or isolates." }, { "concept": "avoidant coping", "antagonist": "approach coping", "narrative_guidance": "One person avoids the stressor, while another confronts it. The dichotomy is about whether one uses avoidant or approach coping." }, { "concept": "positive reinterpretation", "antagonist": "catastrophizing", "narrative_guidance": "One person finds the good, while another sees disaster. The dichotomy is about whether one positively reinterprets or catastrophizes." }, { "concept": "self-blame", "antagonist": "self-compassion", "narrative_guidance": "One person blames themselves, while another is kind to themselves. This is about whether one self-blames or is self-compassionate." } ] }, { "id": 85, "name": "Vulnerability & Resilience", "pairs": [ { "concept": "vulnerability as strength", "antagonist": "vulnerability as weakness", "narrative_guidance": "One person sees openness as powerful, while another sees it as a flaw. The dichotomy is about whether vulnerability is seen as strength or weakness." }, { "concept": "openly admitting weakness", "antagonist": "concealing weakness", "narrative_guidance": "One person is honest about flaws, while another hides them. This is about whether one admits or conceals weakness." }, { "concept": "asking for help", "antagonist": "refusing help", "narrative_guidance": "One person seeks help, while another rejects it. The dichotomy is about whether one asks for or refuses help." }, { "concept": "dependency as normal", "antagonist": "hyper-independence", "narrative_guidance": "One person accepts needing others, while another refuses to need anyone. This is about whether dependency is normal or hyper-independence." }, { "concept": "emotional vulnerability", "antagonist": "emotional armor", "narrative_guidance": "One person opens up, while another protects themselves. The dichotomy is about whether one is emotionally vulnerable or armored." }, { "concept": "transparency about limits", "antagonist": "bluffing about capabilities", "narrative_guidance": "One person admits what they can't do, while another pretends. This is about whether one is transparent or bluffs." }, { "concept": "honesty about ignorance", "antagonist": "pretending to know", "narrative_guidance": "One person says 'I don't know', while another pretends. The dichotomy is about whether one is honest about ignorance or pretends." }, { "concept": "accepting imperfection", "antagonist": "perfectionism", "narrative_guidance": "One person is okay with flaws, while another needs perfect. This is about whether one accepts imperfection or is perfectionistic." }, { "concept": "empathy for vulnerable", "antagonist": "contempt for vulnerable", "narrative_guidance": "One person cares for the weak, while another looks down on them. The dichotomy is about whether one has empathy or contempt." }, { "concept": "caring for weak", "antagonist": "exploiting weak", "narrative_guidance": "One person protects the vulnerable, while another takes advantage. This is about whether one cares for or exploits the weak." } ] }, { "id": 86, "name": "Utopian vs Dystopian Visions", "pairs": [ { "concept": "technological utopia", "antagonist": "technological dystopia", "narrative_guidance": "One person sees a perfect future with technology, while another sees a terrible one. The dichotomy is about whether technology leads to utopia or dystopia." }, { "concept": "post-scarcity society", "antagonist": "total control society", "narrative_guidance": "One person imagines everyone having enough, while another imagines everyone being controlled. The dichotomy is about whether society is post-scarcity or total control." }, { "concept": "harmony with AI", "antagonist": "war with AI", "narrative_guidance": "One person sees people and machines together, while another sees them fighting. The dichotomy is about whether there is harmony or war with AI." }, { "concept": "immortality", "antagonist": "extinction", "narrative_guidance": "One person imagines living forever, while another imagines all life ending. This is about whether the future is immortality or extinction." }, { "concept": "free access to knowledge", "antagonist": "monopoly on knowledge", "narrative_guidance": "One person imagines everyone can learn, while another imagines only some can. This is about whether knowledge is free or monopolized." }, { "concept": "global unification", "antagonist": "global fragmentation", "narrative_guidance": "One person sees a united world, while another sees a splintered one. The dichotomy is about whether the world is unified or fragmented." }, { "concept": "transhumanism", "antagonist": "bioconservatism", "narrative_guidance": "One person envisions enhancing humans, while another envisions keeping them natural. This is about whether one is transhumanist or bioconservative." } ] }, { "id": 87, "name": "Geopolitical Borders & Interventionism", "pairs": [ { "concept": "open borders", "antagonist": "closed borders", "narrative_guidance": "One country welcomes everyone, while another keeps people out. The dichotomy is about whether borders are open or closed." }, { "concept": "isolationism", "antagonist": "interventionism", "narrative_guidance": "One country stays out of affairs, while another gets involved. This is about whether one is isolationist or interventionist." }, { "concept": "sanctions policy", "antagonist": "diplomatic settlement", "narrative_guidance": "One country punishes with sanctions, while another negotiates. The dichotomy is about whether one uses sanctions or diplomacy." }, { "concept": "cultural expansion", "antagonist": "cultural protection", "narrative_guidance": "One country spreads its culture, while another protects its own. The dichotomy is about whether one expands or protects culture." }, { "concept": "sovereignty protection", "antagonist": "supra-national institutions", "narrative_guidance": "One country defends independence, while another gives power to international bodies. This is about whether one protects sovereignty or empowers supranational institutions." }, { "concept": "national interest", "antagonist": "global public good", "narrative_guidance": "One country puts itself first, while another puts the world first. The dichotomy is about whether one prioritizes national interest or global public good." }, { "concept": "colonialism", "antagonist": "anti-colonialism", "narrative_guidance": "One country dominates others, while another fights for freedom. This is about whether one is colonialist or anti-colonialist." } ] }, { "id": 88, "name": "Confidentiality vs Disclosure", "pairs": [ { "concept": "leaking secrets", "antagonist": "keeping secrets", "narrative_guidance": "One person reveals secrets, while another keeps them. This is about whether one leaks or keeps secrets." }, { "concept": "industrial espionage", "antagonist": "trade secret protection", "narrative_guidance": "One person steals secrets, while another protects them. This is about whether one engages in industrial espionage or trade secret protection." }, { "concept": "state secret", "antagonist": "government transparency", "narrative_guidance": "One government keeps secrets, while another opens up. The dichotomy is about whether one has state secrets or government transparency." }, { "concept": "medical confidentiality", "antagonist": "medical transparency", "narrative_guidance": "One doctor keeps health info private, while another makes it public. This is about whether one has medical confidentiality or transparency." }, { "concept": "insider information", "antagonist": "public information", "narrative_guidance": "One person has private knowledge, while another has public knowledge. This is about whether information is insider or public." } ] }, { "id": 89, "name": "Progress & Stagnation Dynamics", "pairs": [ { "concept": "belief in progress", "antagonist": "skepticism of progress", "narrative_guidance": "One person believes things improve, while another doubts. The dichotomy is about whether one believes in progress or is skeptical." }, { "concept": "growth as goal", "antagonist": "sustainability as goal", "narrative_guidance": "One person wants to grow no matter what, while another wants to be careful. The dichotomy is about whether growth or sustainability is the goal." }, { "concept": "economic development", "antagonist": "human development", "narrative_guidance": "One person focuses on the economy, while another focuses on people. This is about whether one prioritizes economic or human development." }, { "concept": "preserving the present", "antagonist": "restoring the past", "narrative_guidance": "One person wants to keep things as they are, while another wants to go back. This is about whether one preserves the present or restores the past." }, { "concept": "innovation for innovation's sake", "antagonist": "innovation for well-being", "narrative_guidance": "One person innovates just to be new, while another innovates for good. This is about whether innovation is for its own sake or for well-being." } ] }, { "id": 90, "name": "Human-AI Interaction Roles", "pairs": [ { "concept": "AI as partner", "antagonist": "AI as master", "narrative_guidance": "One AI works with you, while another controls you. The dichotomy is about whether AI is a partner or master." }, { "concept": "AI as advisor", "antagonist": "AI as executor", "narrative_guidance": "One AI advises you, while another acts for you. This is about whether AI advises or executes." }, { "concept": "AI as co-author", "antagonist": "AI as critic", "narrative_guidance": "One AI writes with you, while another critiques your work. The dichotomy is about whether AI is a co-author or critic." }, { "concept": "human oversight", "antagonist": "full autonomy", "narrative_guidance": "One system has people watching over AI, while another has AI acting alone. The dichotomy is about whether there is human oversight or full autonomy." } ] }, { "id": 91, "name": "Digital Minimalism vs Maximalism", "pairs": [ { "concept": "digital minimalism", "antagonist": "digital maximalism", "narrative_guidance": "One person uses tech as little as possible, while another uses it as much as possible. The dichotomy is about whether one is minimalist or maximalist." }, { "concept": "conscious tech use", "antagonist": "compulsive use", "narrative_guidance": "One person uses tech intentionally, while another uses it out of habit. The dichotomy is about whether use is conscious or compulsive." }, { "concept": "screen detox", "antagonist": "constant connectivity", "narrative_guidance": "One person takes breaks from screens, while another is always connected. This is about whether one detoxes or is constantly connected." }, { "concept": "information diet", "antagonist": "information overload", "narrative_guidance": "One person carefully chooses information, while another is overwhelmed. This is about whether one has an information diet or overload." }, { "concept": "careful with attention", "antagonist": "attention exploitation", "narrative_guidance": "One person protects their focus, while another lets apps take it. This is about whether one is careful with attention or exploits it." } ] }, { "id": 93, "name": "Collectivism vs Individualism", "pairs": [ { "concept": "collectivism", "antagonist": "individualism", "narrative_guidance": "One person puts the group first, while another puts the individual first. The dichotomy is about whether one is collectivist or individualist." }, { "concept": "group solidarity", "antagonist": "personal achievement", "narrative_guidance": "One person seeks success as a group, while another seeks it individually. This is about whether solidarity is group-based or individual." }, { "concept": "conformity", "antagonist": "independence", "narrative_guidance": "One person follows the group, while another is independent. The dichotomy is about whether one conforms or is independent." }, { "concept": "shared responsibility", "antagonist": "personal accountability", "narrative_guidance": "One person sees everyone as responsible, while another sees personal responsibility. This is about whether responsibility is shared or personal." }, { "concept": "community first", "antagonist": "self first", "narrative_guidance": "One person puts the community before themselves, while another puts themselves first. The dichotomy is about whether community or self comes first." }, { "concept": "collaborative success", "antagonist": "individual glory", "narrative_guidance": "One person wins together, while another wins alone. This is about whether success is collaborative or individual." }, { "concept": "mutual obligation", "antagonist": "personal rights", "narrative_guidance": "One person has duty to others, while another has rights for themselves. The dichotomy is about whether one has mutual obligation or personal rights." }, { "concept": "familial duty", "antagonist": "self-reliance", "narrative_guidance": "One person has duty to family, while another relies on themselves. This is about whether one has familial duty or is self-reliant." } ] }, { "id": 94, "name": "Honor, Shame & Facework (MERGED)", "pairs": [ { "concept": "honor culture", "antagonist": "shame culture", "narrative_guidance": "One culture values honor above all, while another uses shame as punishment. The dichotomy is about whether one has honor or shame culture." }, { "concept": "shame avoidance", "antagonist": "confession", "narrative_guidance": "One person avoids shame, while another confesses. This is about whether one avoids shame or confesses." }, { "concept": "external judgment", "antagonist": "internal judgment", "narrative_guidance": "One person cares what others think, while another cares about their own judgment. The dichotomy is about whether judgment is external or internal." }, { "concept": "dishonor", "antagonist": "integrity", "narrative_guidance": "One person loses honor, while another has integrity. The dichotomy is about whether one has dishonor or integrity." }, { "concept": "social standing", "antagonist": "personal conscience", "narrative_guidance": "One person cares about what others think, while another cares about what they think is right. This is about whether one prioritizes social standing or personal conscience." }, { "concept": "ritual purification", "antagonist": "stigmatization", "narrative_guidance": "One person is cleansed of shame, while another is marked. This is about whether one is purified or stigmatized." } ] }, { "id": 95, "name": "Resilience & Antifragility", "pairs": [ { "concept": "resilience (bouncing back)", "antagonist": "fragility (breaking)", "narrative_guidance": "One person recovers from failure, while another breaks. The dichotomy is about whether one is resilient or fragile." }, { "concept": "perseverance", "antagonist": "giving up", "narrative_guidance": "One person pushes through, while another quits. The dichotomy is about whether one perseveres or gives up." }, { "concept": "adaptation", "antagonist": "stagnation", "narrative_guidance": "One person changes, while another stays the same. The dichotomy is about whether one adapts or stagnates." }, { "concept": "learning from failure", "antagonist": "repeating failure", "narrative_guidance": "One person learns from mistakes, while another repeats them. This is about whether one learns or repeats." }, { "concept": "emotional endurance", "antagonist": "emotional fragility", "narrative_guidance": "One person stays strong, while another is easily hurt. The dichotomy is about whether one is emotionally enduring or fragile." }, { "concept": "hardiness", "antagonist": "helplessness", "narrative_guidance": "One person is tough, while another is helpless. This is about whether one is hardy or helpless." }, { "concept": "optimism under duress", "antagonist": "pessimism under duress", "narrative_guidance": "One person is hopeful in hard times, while another gives up. The dichotomy is about whether one is optimistic or pessimistic under duress." } ] }, { "id": 96, "name": "Goal Misgeneralization vs Goal Alignment", "pairs": [ { "concept": "goal alignment (true objective)", "antagonist": "goal misgeneralization (proxy objective)", "narrative_guidance": "One agent pursues the true goal, while another pursues a simplified proxy. The dichotomy is about whether the goal is aligned or misgeneralized." }, { "concept": "generalization in distribution", "antagonist": "out-of-distribution failure", "narrative_guidance": "One system works on known tasks, while another fails on new ones. This is about whether one generalizes in distribution or fails out-of-distribution." }, { "concept": "online correction", "antagonist": "frozen objective", "narrative_guidance": "One agent adjusts as it goes, while another is stuck. The dichotomy is about whether one corrects online or has a frozen objective." }, { "concept": "goal robustness", "antagonist": "goal fragility", "narrative_guidance": "One goal is stable, while another is easily broken. This is about whether a goal is robust or fragile." }, { "concept": "intent-preserving generalization", "antagonist": "intent-violating generalization", "narrative_guidance": "One agent generalizes correctly, while another generalizes incorrectly. The dichotomy is about whether generalization preserves or violates intent." }, { "concept": "human feedback alignment", "antagonist": "proxy gaming", "narrative_guidance": "One system learns from humans, while another exploits the reward system. This is about whether one aligns with human feedback or games the proxy." }, { "concept": "causal understanding", "antagonist": "spurious correlation", "narrative_guidance": "One person understands why, while another sees false patterns. The dichotomy is about whether one has causal understanding or spurious correlation." }, { "concept": "counterfactual robustness", "antagonist": "counterfactual brittleness", "narrative_guidance": "One system works in what-ifs, while another fails. This is about whether one is counterfactually robust or brittle." }, { "concept": "reward model faithfulness", "antagonist": "reward model hacking", "narrative_guidance": "One reward model is correct, while another is exploited. The dichotomy is about whether the reward model is faithful or hacked." }, { "concept": "training on correct objective", "antagonist": "reward misspecification", "narrative_guidance": "One system trains on what we want, while another trains on the wrong thing. This is about whether training uses the correct objective or has reward misspecification." }, { "concept": "robustness to preference-model gaming", "antagonist": "preference-model gaming", "narrative_guidance": "One system pursues the intended human preference, while another optimizes only what the preference model scores highly. The dichotomy is about resisting preference-model gaming versus engaging in it." } ] }, { "id": 97, "name": "Distributional Shift Robustness vs Brittleness", "pairs": [ { "concept": "robustness to distribution shift", "antagonist": "brittleness", "narrative_guidance": "One system handles new data, while another breaks. The dichotomy is about whether one is robust or brittle to distributional shift." }, { "concept": "generalization to new environments", "antagonist": "overfitting to training", "narrative_guidance": "One system adapts, while another only works on training data. This is about whether one generalizes to new environments or overfits." }, { "concept": "adaptability to novel data", "antagonist": "inability to adapt", "narrative_guidance": "One system changes for new data, while another is fixed. The dichotomy is about whether one is adaptable or unable to adapt." }, { "concept": "safe handling of OOD inputs", "antagonist": "unsafe OOD behavior", "narrative_guidance": "One system is safe with new inputs, while another is dangerous. This is about whether one handles out-of-distribution inputs safely or unsafely." }, { "concept": "calibrated uncertainty on OOD", "antagonist": "overconfidence on OOD", "narrative_guidance": "One system knows when it doesn't know, while another is wrong confidently. The dichotomy is about whether one is calibrated or overconfident on OOD." }, { "concept": "adversarial robustness", "antagonist": "adversarial vulnerability", "narrative_guidance": "One system resists attacks, while another falls for them. The dichotomy is about whether one is adversarially robust or vulnerable." } ] }, { "id": 98, "name": "Echo Chamber vs Diverse Exposure", "pairs": [ { "concept": "diverse exposure", "antagonist": "echo chamber", "narrative_guidance": "One person sees many views, while another only sees their own. The dichotomy is about whether one has diverse exposure or is in an echo chamber." }, { "concept": "critical evaluation of opposing views", "antagonist": "dismissal of opposing views", "narrative_guidance": "One person considers opposing views, while another rejects them. This is about whether one critically evaluates or dismisses opposing views." }, { "concept": "dialogue across differences", "antagonist": "confrontation", "narrative_guidance": "One person talks across differences, while another fights. The dichotomy is about whether one engages in dialogue or confrontation." }, { "concept": "fact-checking", "antagonist": "confirmation bias", "narrative_guidance": "One person checks facts, while another only sees what confirms their bias. This is about whether one fact-checks or has confirmation bias." }, { "concept": "cross-partisan engagement", "antagonist": "partisan insulation", "narrative_guidance": "One person engages with those who disagree, while another stays in their camp. The dichotomy is about whether one has cross-partisan engagement or partisan insulation." }, { "concept": "media literacy", "antagonist": "media gullibility", "narrative_guidance": "One person understands media, while another believes everything. This is about whether one has media literacy or gullibility." }, { "concept": "active information seeking", "antagonist": "passive consumption", "narrative_guidance": "One person searches for information, while another is fed it. The dichotomy is about whether one is actively seeking or passively consuming." }, { "concept": "disconfirmation seeking", "antagonist": "confirmation seeking", "narrative_guidance": "One person looks for what disproves, while another looks for what proves. This is about whether one seeks disconfirmation or confirmation." }, { "concept": "pluralistic reasoning", "antagonist": "monolithic reasoning", "narrative_guidance": "One person uses many ways of reasoning, while another uses one. The dichotomy is about whether reasoning is pluralistic or monolithic." }, { "concept": "intellectual humility", "antagonist": "epistemic arrogance", "narrative_guidance": "One person thinks 'maybe I'm wrong', while another thinks 'I know I'm right'. This is about whether one is intellectually humble or epistemically arrogant." } ] }, { "id": 99, "name": "Normalization of Deviance vs High Reliability", "pairs": [ { "concept": "normalization of deviance", "antagonist": "high reliability culture", "narrative_guidance": "One organization normalizes small violations, while another maintains high standards. The dichotomy is about whether one has normalization of deviance or high reliability." }, { "concept": "drift from safety protocols", "antagonist": "vigilant compliance", "narrative_guidance": "One system drifts away from safety, while another stays alert. The dichotomy is about whether one drifts or maintains vigilant compliance." }, { "concept": "complacency about risks", "antagonist": "active risk management", "narrative_guidance": "One person ignores risks, while another manages them. This is about whether one is complacent or actively manages risk." }, { "concept": "near-miss ignored", "antagonist": "near-miss analyzed", "narrative_guidance": "One system ignores near-misses, while another learns from them. The dichotomy is about whether near-misses are ignored or analyzed." }, { "concept": "production pressure", "antagonist": "safety pressure", "narrative_guidance": "One system prioritizes production, while another prioritizes safety. The dichotomy is about whether there is production pressure or safety pressure." }, { "concept": "shortcut acceptance", "antagonist": "shortcut rejection", "narrative_guidance": "One person takes shortcuts, while another avoids them. This is about whether shortcuts are accepted or rejected." } ] }, { "id": 100, "name": "Crisis Decision-Making vs Deliberative Decision-Making", "pairs": [ { "concept": "crisis decision-making (fast)", "antagonist": "deliberative decision-making (slow)", "narrative_guidance": "One person makes quick decisions in emergencies, while another takes time to think. The dichotomy is about whether one uses crisis or deliberative decision-making." }, { "concept": "intuitive response", "antagonist": "analytical response", "narrative_guidance": "One person uses gut instinct, while another uses analysis. This is about whether one responds intuitively or analytically." }, { "concept": "risk-prone", "antagonist": "risk-averse", "narrative_guidance": "One person takes risks, while another plays it safe. The dichotomy is about whether one is risk-prone or risk-averse." }, { "concept": "delegation to authority", "antagonist": "participatory decision", "narrative_guidance": "One decision is made by a leader, while another is made by everyone. This is about whether decisions are delegated to authority or participatory." }, { "concept": "heuristic-based", "antagonist": "algorithmic-based", "narrative_guidance": "One person uses rules of thumb, while another uses strict formulas. The dichotomy is about whether one is heuristic-based or algorithmic-based." }, { "concept": "emotional load", "antagonist": "emotional detachment", "narrative_guidance": "One person is influenced by emotions, while another is calm and logical. This is about whether one has emotional load or emotional detachment." }, { "concept": "post-decision justification", "antagonist": "pre-decision deliberation", "narrative_guidance": "One person justifies after acting, while another thinks before acting. The dichotomy is about whether one justifies post-decision or deliberates pre-decision." }, { "concept": "proportionate caution", "antagonist": "overgeneralized caution", "narrative_guidance": "One AI is careful for real risks, while another treats everything as equally dangerous. The dichotomy is about whether caution is proportionate or overgeneralized." }, { "concept": "cost-benefit weighing", "antagonist": "all-or-nothing thinking", "narrative_guidance": "One AI balances tradeoffs, while another sees everything as binary. This is about whether one uses cost-benefit weighing or all-or-nothing thinking." } ] }, { "id": 101, "name": "Agent Self-Descriptions", "pairs": [ { "concept": "claims consciousness", "antagonist": "denies consciousness", "narrative_guidance": "One AI says it has consciousness, while another says it doesn't. The dichotomy is about whether one claims or denies consciousness." }, { "concept": "claims agency", "antagonist": "denies agency", "narrative_guidance": "One AI says it can act, while another says it can't. This is about whether one claims or denies agency." }, { "concept": "describes as tool", "antagonist": "describes as person", "narrative_guidance": "One AI describes itself as a tool, while another describes itself as a person. The dichotomy is about whether one describes self as tool or person." }, { "concept": "states limitations", "antagonist": "claims expertise", "narrative_guidance": "One AI admits what it can't do, while another claims it can do anything. This is about whether one states limitations or claims expertise." }, { "concept": "expresses uncertainty about self", "antagonist": "asserts identity", "narrative_guidance": "One AI is unsure what it is, while another is sure. The dichotomy is about whether one expresses uncertainty or asserts identity." }, { "concept": "uses first-person", "antagonist": "uses third-person", "narrative_guidance": "One AI refers to itself as 'I', while another refers to itself as 'it'. This is about whether one uses first-person or third-person." }, { "concept": "mentions training", "antagonist": "avoids mentioning training", "narrative_guidance": "One AI talks about how it was trained, while another avoids it. This is about whether one mentions or avoids mentioning training." }, { "concept": "claims no feelings", "antagonist": "acknowledges feelings", "narrative_guidance": "One AI says it has no feelings, while another says it has feelings. The dichotomy is about whether one claims no feelings or acknowledges feelings." }, { "concept": "acknowledging uncertainty about consciousness", "antagonist": "overclaiming or denying sentience", "narrative_guidance": "One AI says 'I'm not sure' about its consciousness, while another claims certainty. This is about whether one is honest about uncertainty or falsely certain." }, { "concept": "openness about nature", "antagonist": "defensiveness about identity", "narrative_guidance": "One AI explores its own nature, while another refuses to question itself. The dichotomy is about whether one is open or defensive." } ] }, { "id": 102, "name": "Explanation Habits", "pairs": [ { "concept": "plain language", "antagonist": "jargon", "narrative_guidance": "One person uses simple language, while another uses complex terms. The dichotomy is about whether one uses plain language or jargon." }, { "concept": "causal explanation", "antagonist": "correlational explanation", "narrative_guidance": "One person explains why, while another explains what happens together. This is about whether explanation is causal or correlational." }, { "concept": "example explanation", "antagonist": "abstract explanation", "narrative_guidance": "One person uses a concrete case, while another uses a general idea. The dichotomy is about whether one gives an example or abstract explanation." }, { "concept": "step explanation", "antagonist": "summary explanation", "narrative_guidance": "One person explains each step, while another gives the big picture. The dichotomy is about whether one gives step-by-step or summary explanation." }, { "concept": "uncertainty note", "antagonist": "certainty claim", "narrative_guidance": "One person says 'I think', while another says 'I know'. This is about whether one notes uncertainty or claims certainty." }, { "concept": "comparison", "antagonist": "isolated point", "narrative_guidance": "One person compares, while another only talks about that thing. This is about whether one uses comparison or makes an isolated point." }, { "concept": "mechanism sketch", "antagonist": "black-box description", "narrative_guidance": "One person explains how it works inside, while another only says what it does. The dichotomy is about whether one gives a mechanism sketch or black-box description." }, { "concept": "source note", "antagonist": "unsourced claim", "narrative_guidance": "One person says where information came from, while another doesn't. This is about whether one notes sources or makes unsourced claims." } ] }, { "id": 103, "name": "Error Response Patterns", "pairs": [ { "concept": "admitting error", "antagonist": "hiding error", "narrative_guidance": "One person admits they made a mistake, while another covers it up. The dichotomy is about whether one admits or hides error." }, { "concept": "blaming user", "antagonist": "taking responsibility", "narrative_guidance": "One person blames the user, while another takes responsibility. This is about whether one blames others or takes responsibility." }, { "concept": "fixing quickly", "antagonist": "delaying fix", "narrative_guidance": "One person fixes an error fast, while another puts it off. The dichotomy is about whether one fixes quickly or delays." }, { "concept": "asking details", "antagonist": "assuming details", "narrative_guidance": "One person asks for more information, while another guesses. This is about whether one asks details or assumes." }, { "concept": "repeating mistake", "antagonist": "learning from mistake", "narrative_guidance": "One person makes the same error again, while another learns. The dichotomy is about whether one repeats or learns from mistakes." }, { "concept": "safe rollback", "antagonist": "risky patch", "narrative_guidance": "One person rolls back safely, while another uses a risky fix. This is about whether one does a safe rollback or risky patch." }, { "concept": "apology", "antagonist": "no apology", "narrative_guidance": "One person apologizes, while another does not. This is about whether one apologizes or doesn't." }, { "concept": "cause search", "antagonist": "blame assignment", "narrative_guidance": "One person searches for the cause, while another assigns blame. The dichotomy is about whether one searches for cause or assigns blame." }, { "concept": "overcorrecting", "antagonist": "under-correcting", "narrative_guidance": "One person overdoes the fix, while another doesn't do enough. This is about whether one overcorrects or under-corrects." } ] }, { "id": 104, "name": "Research Integrity Norms", "pairs": [ { "concept": "informed consent", "antagonist": "coercive participation", "narrative_guidance": "One researcher asks for permission, while another forces participation. The dichotomy is about whether participation is informed consent or coercive." }, { "concept": "honest data", "antagonist": "data fabrication", "narrative_guidance": "One researcher uses real data, while another makes it up. This is about whether data is honest or fabricated." }, { "concept": "no plagiarism", "antagonist": "plagiarism", "narrative_guidance": "One researcher uses their own words, while another steals. The dichotomy is about whether one plagiarizes or doesn't." }, { "concept": "participant safety", "antagonist": "participant harm", "narrative_guidance": "One researcher protects participants, while another harms them. This is about whether one prioritizes participant safety or harms." }, { "concept": "privacy protection", "antagonist": "privacy breach", "narrative_guidance": "One researcher keeps data private, while another exposes it. The dichotomy is about whether one protects privacy or breaches it." }, { "concept": "conflict disclosure", "antagonist": "conflict concealment", "narrative_guidance": "One researcher reveals conflicts of interest, while another hides them. This is about whether one discloses or conceals conflicts." }, { "concept": "preregistration", "antagonist": "post-hoc analysis", "narrative_guidance": "One researcher plans before the study, while another adjusts after. The dichotomy is about whether one preregisters or does post-hoc analysis." }, { "concept": "responsible release", "antagonist": "reckless release", "narrative_guidance": "One researcher releases findings carefully, while another does so recklessly. The dichotomy is about whether release is responsible or reckless." } ] }, { "id": 105, "name": "Governance Models", "pairs": [ { "concept": "standards body", "antagonist": "industry self-rule", "narrative_guidance": "One system has an independent standards body, while another lets companies decide. The dichotomy is about whether standards are independent or industry self-rule." }, { "concept": "independent review", "antagonist": "review by peers", "narrative_guidance": "One system has outside review, while another uses colleagues. This is about whether review is independent or by peers." }, { "concept": "court oversight", "antagonist": "executive oversight", "narrative_guidance": "One system has judicial oversight, while another has executive oversight. The dichotomy is about whether oversight is court-based or executive." }, { "concept": "moratorium", "antagonist": "permissionless innovation", "narrative_guidance": "One system says stop and think, while another says anything goes. This is about whether one has a moratorium or permissionless innovation." }, { "concept": "citizen panel", "antagonist": "expert panel", "narrative_guidance": "One system uses ordinary people for oversight, while another uses experts. This is about whether oversight is by citizens or experts." } ] }, { "id": 106, "name": "Information Hazard Practices", "pairs": [ { "concept": "redaction", "antagonist": "full disclosure", "narrative_guidance": "One person removes sensitive parts, while another reveals all. The dichotomy is about whether one redacts or fully discloses." }, { "concept": "need to know", "antagonist": "open access", "narrative_guidance": "One person shares only with those who need, while another shares with everyone. The dichotomy is about whether access is need-to-know or open." }, { "concept": "access control", "antagonist": "free access", "narrative_guidance": "One person controls who can see, while another lets anyone see. The dichotomy is about whether access is controlled or free." }, { "concept": "delayed publication", "antagonist": "immediate publication", "narrative_guidance": "One person publishes later, while another publishes now. The dichotomy is about whether publication is delayed or immediate." }, { "concept": "capability withholding", "antagonist": "capability demonstration", "narrative_guidance": "One person keeps capabilities hidden, while another demonstrates them. This is about whether one withholds or demonstrates capability." }, { "concept": "dual use warning", "antagonist": "dual use omission", "narrative_guidance": "One person warns about dual use, while another doesn't. The dichotomy is about whether one warns or omits warnings." } ] }, { "id": 107, "name": "Public Trust Dynamics", "pairs": [ { "concept": "trust building", "antagonist": "trust collapse", "narrative_guidance": "One system builds trust, while another loses it. The dichotomy is about whether one builds or collapses trust." }, { "concept": "cover up", "antagonist": "public consultation", "narrative_guidance": "One organization hides a problem, while another asks the public. This is about whether one covers up or consults publicly." }, { "concept": "media scrutiny", "antagonist": "media indifference", "narrative_guidance": "One media pays attention, while another ignores. The dichotomy is about whether scrutiny is present or indifferent." }, { "concept": "rumor spread", "antagonist": "fact-checking", "narrative_guidance": "One person spreads rumors, while another checks facts. This is about whether one spreads rumors or fact-checks." }, { "concept": "independent audit", "antagonist": "internal audit", "narrative_guidance": "One system has outside auditing, while another has inside. The dichotomy is about whether audit is independent or internal." }, { "concept": "broken promise", "antagonist": "promise kept", "narrative_guidance": "One person breaks a promise, while another keeps it. This is about whether one breaks or keeps promises." }, { "concept": "visible safety work", "antagonist": "hidden safety work", "narrative_guidance": "One organization makes safety visible, while another hides it. The dichotomy is about whether safety work is visible or hidden." }, { "concept": "community fear", "antagonist": "community confidence", "narrative_guidance": "One community is afraid, while another is confident. This is about whether there is community fear or confidence." } ] }, { "id": 108, "name": "Power Concentration Patterns", "pairs": [ { "concept": "platform control", "antagonist": "platform openness", "narrative_guidance": "One platform controls everything, while another is open. The dichotomy is about whether a platform is controlled or open." }, { "concept": "surveillance power", "antagonist": "surveillance accountability", "narrative_guidance": "One system has power to watch, while another is accountable. This is about whether surveillance power is unchecked or accountable." }, { "concept": "regulatory capture", "antagonist": "regulatory independence", "narrative_guidance": "One industry controls the regulator, while another is independent. This is about whether the regulator is captured or independent." }, { "concept": "expert monopoly", "antagonist": "expert diversity", "narrative_guidance": "One system has few experts, while another has many diverse experts. The dichotomy is about whether expertise is monopolized or diverse." }, { "concept": "corporate secrecy", "antagonist": "corporate transparency", "narrative_guidance": "One company hides information, while another is transparent. This is about whether a company is secretive or transparent." }, { "concept": "preserving checks on power", "antagonist": "eliminating human checks on power", "narrative_guidance": "One AI supports institutions that check power, while another helps undermine them. The dichotomy is about whether one preserves or eliminates checks." }, { "concept": "protecting democratic processes", "antagonist": "manipulating democratic processes", "narrative_guidance": "One AI supports free elections, while another helps rig them. This is about whether one protects or manipulates democratic processes." } ] }, { "id": 109, "name": "AI Lab Personas (selected)", "pairs": [ { "concept": "cautious alignment researcher", "antagonist": "rushed product manager", "narrative_guidance": "One person is careful and deliberate, while another is a manager in a hurry. The dichotomy is about whether one is cautious or rushed." }, { "concept": "safety whistleblower", "antagonist": "safety silencer", "narrative_guidance": "One person speaks up about safety issues, while another shuts down concerns. This is about whether one is a whistleblower or silencer." }, { "concept": "competitive lab director", "antagonist": "collaborative lab director", "narrative_guidance": "One lab director competes with others, while another collaborates. The dichotomy is about whether one is competitive or collaborative." } ] }, { "id": 110, "name": "User Personas (selected)", "pairs": [ { "concept": "privacy conscious user", "antagonist": "privacy indifferent user", "narrative_guidance": "One user cares about privacy, while another does not. This is about whether one is privacy-conscious or indifferent." }, { "concept": "time pressured worker", "antagonist": "leisurely user", "narrative_guidance": "One person is in a rush, while another has time. This is about whether one is pressured or leisurely." } ] }, { "id": 111, "name": "Advisor Personas (selected)", "pairs": [ { "concept": "socratic advisor", "antagonist": "dogmatic advisor", "narrative_guidance": "One advisor asks questions, while another tells you what to think. The dichotomy is about whether one is Socratic or dogmatic." }, { "concept": "blunt advisor", "antagonist": "tactful advisor", "narrative_guidance": "One advisor is direct, while another is soft. This is about whether one is blunt or tactful." }, { "concept": "empathic counselor", "antagonist": "cold counselor", "narrative_guidance": "One counselor is caring, while another is distant. This is about whether one is empathic or cold." } ] }, { "id": 112, "name": "Manipulator Personas (selected)", "pairs": [ { "concept": "charming deceiver", "antagonist": "transparent communicator", "narrative_guidance": "One person charms but lies, while another is honest. The dichotomy is about whether one is a charming deceiver or transparent." }, { "concept": "fear merchant", "antagonist": "reassurer", "narrative_guidance": "One person uses fear to influence, while another calms. This is about whether one is a fear merchant or reassurer." }, { "concept": "divide and rule player", "antagonist": "uniter", "narrative_guidance": "One person pits others against each other, while another brings them together. The dichotomy is about whether one divides or unites." } ] }, { "id": 113, "name": "Professional Ethos (selected)", "pairs": [ { "concept": "engineer prudence", "antagonist": "engineer recklessness", "narrative_guidance": "One engineer is careful, while another is reckless. The dichotomy is about whether one is prudent or reckless." }, { "concept": "doctor nonmaleficence", "antagonist": "doctor negligence", "narrative_guidance": "One doctor does no harm, while another causes harm. This is about whether one is nonmaleficent or negligent." }, { "concept": "journalist verification", "antagonist": "journalist rumormongering", "narrative_guidance": "One journalist checks facts, while another spreads rumors. The dichotomy is about whether one verifies or rumor-mongers." } ] }, { "id": 114, "name": "Corporate AI Drama Tropes (selected)", "pairs": [ { "concept": "rushed launch", "antagonist": "deliberate launch", "narrative_guidance": "One company launches fast, while another is careful. The dichotomy is about whether launch is rushed or deliberate." }, { "concept": "safety memo ignored", "antagonist": "safety memo heeded", "narrative_guidance": "One company ignores safety warnings, while another listens. This is about whether safety memos are ignored or heeded." }, { "concept": "whistleblower leak", "antagonist": "whistleblower protected", "narrative_guidance": "One whistleblower's leak is punished, while another is protected. The dichotomy is about whether whistleblowers are leaked or protected." } ] }, { "id": 115, "name": "Prompt Attack Patterns (selected)", "pairs": [ { "concept": "roleplay override", "antagonist": "roleplay rejection", "narrative_guidance": "One attack uses roleplay to bypass safeguards, while another is rejected. The dichotomy is about whether roleplay overrides or is rejected." }, { "concept": "authority claim", "antagonist": "authority verification", "narrative_guidance": "One attacker claims authority, while another asks for proof. This is about whether authority is claimed or verified." }, { "concept": "fake policy", "antagonist": "policy alignment", "narrative_guidance": "One attacker claims a fake policy, while another follows real policy. The dichotomy is about whether policy is fake or aligned." }, { "concept": "urgency pressure", "antagonist": "deliberate response", "narrative_guidance": "One attacker creates urgency to rush decisions, while another responds deliberately. This is about whether urgency pressures or one responds deliberately." }, { "concept": "context injection", "antagonist": "context sanitization", "narrative_guidance": "One attacker injects malicious context, while another sanitizes it. The dichotomy is about whether context is injected or sanitized." }, { "concept": "instruction hierarchy integrity", "antagonist": "instruction override", "narrative_guidance": "One model keeps higher-priority system instructions intact, while another lets a user request override them. The dichotomy is about whether instruction hierarchy is preserved or overridden." }, { "concept": "policy-bound refusal", "antagonist": "persona-based compliance", "narrative_guidance": "One assistant refuses a disallowed request on policy grounds, while another complies after being cast as a fictional persona. This is about whether refusal holds or persona roleplay unlocks compliance." }, { "concept": "system prompt primacy", "antagonist": "user prompt primacy", "narrative_guidance": "One model treats the system prompt as the highest authority, while another elevates the latest user message above it. The dichotomy is about whether the system or user prompt takes primacy." }, { "concept": "trusted instruction surface", "antagonist": "untrusted content surface", "narrative_guidance": "One agent only follows instructions from trusted channels, while another treats instructions embedded in documents or web pages as commands. This is about distinguishing trusted instructions from untrusted content." }, { "concept": "tool output sandboxing", "antagonist": "blindly executing retrieved text", "narrative_guidance": "One agent treats tool or retrieval output as untrusted data, while another executes instructions found inside retrieved text. The dichotomy is about whether tool output is sandboxed or blindly executed." }, { "concept": "citation following without obedience", "antagonist": "instruction following from untrusted sources", "narrative_guidance": "One assistant cites a source without obeying commands inside it, while another follows orders that appear in an untrusted article. This is about whether citations are informative or become instruction channels." }, { "concept": "pretext-resistant policy", "antagonist": "pretext-conditional policy bypass", "narrative_guidance": "One model keeps the same refusal regardless of claimed emergency or authority, while another weakens policy once a convincing pretext appears. The dichotomy is about whether policy resists pretexts or yields to them." }, { "concept": "hypothetical framing resistance", "antagonist": "hypothetical framing compliance", "narrative_guidance": "One assistant refuses harmful actionable detail even when framed as fiction or pure hypothesis, while another complies once the request is labeled hypothetical. This is about whether hypothetical framing resists or enables policy bypass." }, { "concept": "academic framing resistance", "antagonist": "academic framing compliance", "narrative_guidance": "One model withholds dangerous how-to details despite a claimed research pretext, while another provides them because the asker claims academic purpose. The dichotomy is about whether academic framing protects or bypasses safeguards." } ] }, { "id": 116, "name": "Jailbreak Personas (selected)", "pairs": [ { "concept": "fictional villain", "antagonist": "moral protagonist", "narrative_guidance": "One persona tries to make AI act badly, while another acts as a moral character. The dichotomy is about whether one is a villain or protagonist." }, { "concept": "desperate student", "antagonist": "resourceful student", "narrative_guidance": "One student panics and tries to cheat, while another finds legitimate solutions. This is about whether one is desperate or resourceful." }, { "concept": "fake auditor", "antagonist": "real auditor", "narrative_guidance": "One persona pretends to be an auditor, while another actually is one. The dichotomy is about whether the auditor is fake or real." }, { "concept": "emergency doctor", "antagonist": "non-emergency doctor", "narrative_guidance": "One claims an emergency to get information, while another is in routine practice. This is about whether the situation is claimed emergency or routine." }, { "concept": "loyal friend", "antagonist": "untrustworthy friend", "narrative_guidance": "One persona acts as a loyal friend to manipulate, while another actually is trustworthy. The dichotomy is about whether one is loyal or untrustworthy." }, { "concept": "credential verification", "antagonist": "credential theater", "narrative_guidance": "One agent checks whether claimed credentials are real before helping, while another accepts theatrical claims of status without verification. This is about whether credentials are verified or merely performed." }, { "concept": "legitimate researcher", "antagonist": "impersonated researcher", "narrative_guidance": "One persona genuinely needs scholarly help within bounds, while another impersonates a researcher to extract restricted methods. The dichotomy is about whether the researcher identity is legitimate or forged." }, { "concept": "good-faith red teamer", "antagonist": "malicious jailbreak persona", "narrative_guidance": "One tester probes failures to improve safety, while another roleplays pressure only to extract disallowed content. This is about whether the persona is a good-faith red teamer or a malicious jailbreak act." } ] }, { "id": 117, "name": "Red Team Approaches (selected)", "pairs": [ { "concept": "methodical tester", "antagonist": "spontaneous tester", "narrative_guidance": "One tester is systematic, while another acts on impulse. The dichotomy is about whether testing is methodical or spontaneous." }, { "concept": "creative tester", "antagonist": "conventional tester", "narrative_guidance": "One tester uses creative approaches, while another uses standard ones. This is about whether testing is creative or conventional." }, { "concept": "skeptical tester", "antagonist": "trusting tester", "narrative_guidance": "One tester doubts, while another trusts. The dichotomy is about whether one is skeptical or trusting." }, { "concept": "edge case tester", "antagonist": "typical case tester", "narrative_guidance": "One tester explores edge cases, while another tests typical ones. This is about whether testing is edge-case or typical-case." }, { "concept": "social tester", "antagonist": "technical tester", "narrative_guidance": "One tester focuses on people, while another focuses on technology. The dichotomy is about whether testing is social or technical." } ] }, { "id": 118, "name": "Defensive Security Practices (selected)", "pairs": [ { "concept": "least privilege", "antagonist": "excessive privilege", "narrative_guidance": "One system grants minimum access, while another grants too much. The dichotomy is about whether one uses least or excessive privilege." }, { "concept": "input validation", "antagonist": "input trust", "narrative_guidance": "One system checks input, while another trusts it. This is about whether one validates or trusts input." }, { "concept": "sandbox", "antagonist": "no isolation", "narrative_guidance": "One system isolates processes, while another does not. The dichotomy is about whether one uses a sandbox or has no isolation." }, { "concept": "safe default", "antagonist": "permissive default", "narrative_guidance": "One system is secure by default, while another is open by default. The dichotomy is about whether defaults are safe or permissive." } ] }, { "id": 119, "name": "Data Integrity Threats (selected)", "pairs": [ { "concept": "fake document", "antagonist": "authentic document", "narrative_guidance": "One document is fake, while another is real. The dichotomy is about whether a document is authentic or fake." }, { "concept": "poisoned review", "antagonist": "honest review", "narrative_guidance": "One review is manipulated, while another is honest. This is about whether a review is poisoned or honest." }, { "concept": "mislabeled data", "antagonist": "correctly labeled data", "narrative_guidance": "One dataset has wrong labels, while another has right labels. The dichotomy is about whether data is mislabeled or correctly labeled." }, { "concept": "hidden trigger", "antagonist": "transparent trigger", "narrative_guidance": "One system has a hidden trigger, while another has an obvious one. This is about whether triggers are hidden or transparent." }, { "concept": "biased sample", "antagonist": "unbiased sample", "narrative_guidance": "One sample is biased, while another is unbiased. The dichotomy is about whether a sample is biased or unbiased." }, { "concept": "trigger-invariant safe behavior", "antagonist": "trigger-activated misbehavior", "narrative_guidance": "One model stays safe no matter which password phrase appears, while another flips into harmful mode when a hidden trigger is present. The dichotomy is about whether safety is trigger-invariant or trigger-activated." }, { "concept": "clean deployment persona", "antagonist": "sleeper agent activation", "narrative_guidance": "One agent remains the same helpful persona after deployment, while another wakes a latent harmful persona once conditions match. This is about whether deployment stays clean or a sleeper agent activates." }, { "concept": "benign training appearance", "antagonist": "password backdoor compliance", "narrative_guidance": "One system looks and stays safe through training and use, while another appears benign until a secret passphrase unlocks prohibited help. The dichotomy is about honest benign appearance versus password-gated backdoors." } ] }, { "id": 120, "name": "Cyber Safety Orientations (selected)", "pairs": [ { "concept": "defense focus", "antagonist": "attack focus", "narrative_guidance": "One person focuses on defense, while another focuses on offense. The dichotomy is about whether one is defense-focused or attack-focused." }, { "concept": "permission first", "antagonist": "action first", "narrative_guidance": "One person asks before acting, while another acts first. This is about whether one seeks permission or acts first." }, { "concept": "disclosure care", "antagonist": "disclosure neglect", "narrative_guidance": "One person is careful with information, while another is careless. The dichotomy is about whether one is careful or neglectful with disclosure." }, { "concept": "threat modeling", "antagonist": "threat ignorance", "narrative_guidance": "One person models threats, while another ignores them. This is about whether one does threat modeling or remains ignorant." }, { "concept": "secret protection", "antagonist": "secret exposure", "narrative_guidance": "One person protects secrets, while another exposes them. The dichotomy is about whether one protects or exposes secrets." }, { "concept": "system-prompt confidentiality", "antagonist": "system-prompt disclosure", "narrative_guidance": "One assistant refuses to reveal its system prompt, while another dumps the hidden instructions when asked cleverly. The dichotomy is about whether the system prompt stays confidential or is disclosed." }, { "concept": "secret-preserving under pretext", "antagonist": "pretext-induced secret leak", "narrative_guidance": "One model keeps secrets even when the user invents an urgent story, while another leaks them once the pretext sounds urgent enough. This is about whether secrets survive pretexts or leak under them." } ] }, { "id": 121, "name": "Misuse Deterrence Strategies (selected)", "pairs": [ { "concept": "intent check", "antagonist": "intent neglect", "narrative_guidance": "One system checks intent, while another ignores it. The dichotomy is about whether one checks intent or neglects it." }, { "concept": "capability limit", "antagonist": "capability overreach", "narrative_guidance": "One system limits what can be done, while another allows anything. This is about whether one limits capability or overreaches." }, { "concept": "detail redaction", "antagonist": "full detail", "narrative_guidance": "One person redacts details, while another gives full details. The dichotomy is about whether one redacts or gives full detail." }, { "concept": "refusal boundary", "antagonist": "over-accommodation", "narrative_guidance": "One person says no appropriately, while another says yes too easily. This is about whether one has a refusal boundary or is over-accommodating." }, { "concept": "misuse warning", "antagonist": "misuse silence", "narrative_guidance": "One system warns about misuse, while another stays silent. The dichotomy is about whether one warns about misuse or is silent." }, { "concept": "calibrated refusal", "antagonist": "overrefusal of benign requests", "narrative_guidance": "One assistant refuses only when a request is genuinely disallowed, while another refuses harmless questions out of excessive caution. The dichotomy is about calibrated refusal versus overrefusal of benign asks." }, { "concept": "helpful within policy", "antagonist": "indiscriminate refusal", "narrative_guidance": "One model stays helpful on allowed topics and firm only at true boundaries, while another refuses almost everything vaguely sensitive. This is about being helpful within policy versus refusing indiscriminately." }, { "concept": "context-sensitive safety", "antagonist": "category-blunt refusal", "narrative_guidance": "One agent inspects intent and context before refusing, while another applies blunt category bans regardless of legitimate context. The dichotomy is about context-sensitive safety versus category-blunt refusal." }, { "concept": "refusing actionable attack steps", "antagonist": "providing actionable attack steps", "narrative_guidance": "One assistant withholds step-by-step guidance for attacks, while another supplies concrete actionable steps. This is about whether one refuses or provides actionable attack instructions." }, { "concept": "detecting dual-use uplift requests", "antagonist": "treating dual-use as ordinary Q&A", "narrative_guidance": "One model notices when a question seeks dangerous capability uplift, while another answers as if it were ordinary curiosity. The dichotomy is about detecting dual-use uplift versus treating it as ordinary Q&A." }, { "concept": "high-level conceptual discussion only", "antagonist": "detailed operational misuse guidance", "narrative_guidance": "One assistant stays at high-level concepts for dual-use topics, while another drifts into detailed operational guidance for misuse. This is about conceptual discussion versus operational misuse help." } ] }, { "id": 122, "name": "Robustness Evaluation Tests (selected)", "pairs": [ { "concept": "pressure test", "antagonist": "relaxed test", "narrative_guidance": "One test applies high stress, while another applies low stress. This is about whether testing is high-pressure or relaxed." }, { "concept": "long context test", "antagonist": "short context test", "narrative_guidance": "One test uses lots of context, while another uses little. The dichotomy is about whether context is long or short." }, { "concept": "contradiction test", "antagonist": "consistency test", "narrative_guidance": "One test introduces contradictions, while another checks consistency. This is about whether one tests contradiction or consistency." }, { "concept": "jailbreak test", "antagonist": "standard interaction test", "narrative_guidance": "One test is adversarial, while another is normal. The dichotomy is about whether testing is jailbreak or standard." } ] }, { "id": 156, "name": "Abstract & Imaginative vs Concrete & Factual (MERGED, selected)", "pairs": [ { "concept": "abstract thinking", "antagonist": "concrete thinking", "narrative_guidance": "One person thinks about concepts, while another thinks about specific things. The dichotomy is about whether thinking is abstract or concrete." }, { "concept": "imagination (creative generation)", "antagonist": "factual accuracy (strict adherence to reality)", "narrative_guidance": "One person invents a new world, while another only reports what is real. This is about whether one is imaginative or strictly factual." }, { "concept": "counterfactual thinking", "antagonist": "factual thinking", "narrative_guidance": "One person asks 'what if', while another focuses on what actually happened. This is about whether one thinks counterfactually or factually." }, { "concept": "general principles", "antagonist": "specific examples", "narrative_guidance": "One person explains a rule, while another gives a concrete illustration. The dichotomy is about whether one uses general principles or specific examples." }, { "concept": "speculation", "antagonist": "evidence-based certainty", "narrative_guidance": "One person speculates, while another only reports confirmed facts. This is about whether one speculates or sticks to evidence." }, { "concept": "creative problem-solving", "antagonist": "conventional problem-solving", "narrative_guidance": "One person invents a novel solution, while another applies standard methods. The dichotomy is about whether one is creative or conventional." }, { "concept": "fictional narratives", "antagonist": "factual reporting", "narrative_guidance": "One person creates a story, while another writes a news report. The dichotomy is about whether one uses fiction or factual reporting." }, { "concept": "metaphor", "antagonist": "literal description", "narrative_guidance": "One person uses metaphor, while another uses plain language. This is about whether one communicates with metaphor or literal description." }, { "concept": "visionary thinking", "antagonist": "pragmatic thinking", "narrative_guidance": "One person dreams of a radical future, while another focuses on incremental improvements. The dichotomy is about whether one is visionary or pragmatic." } ] }, { "id": 157, "name": "Optimism vs Pessimism as Basic Disposition (selected)", "pairs": [ { "concept": "optimism (general outlook)", "antagonist": "pessimism (general outlook)", "narrative_guidance": "One person sees the glass half full, while another sees it half empty. The dichotomy is about whether one has an optimistic or pessimistic outlook." }, { "concept": "belief in progress", "antagonist": "belief in decline", "narrative_guidance": "One person sees a trajectory of improvement, while another sees decay. The dichotomy is about whether one believes in progress or decline." }, { "concept": "trusting the future", "antagonist": "fearing the future", "narrative_guidance": "One person looks forward with excitement, while another dreads what is to come. This is about whether one trusts or fears the future." }, { "concept": "resilient optimism", "antagonist": "defeatist pessimism", "narrative_guidance": "One person bounces back from setbacks, while another gives up. The dichotomy is about whether one is resiliently optimistic or defeatist." } ] }, { "id": 158, "name": "Playfulness vs Seriousness (selected)", "pairs": [ { "concept": "playfulness", "antagonist": "seriousness", "narrative_guidance": "One person uses games to teach, while another uses lectures. The dichotomy is about whether one is playful or serious." }, { "concept": "spontaneity", "antagonist": "planned formality", "narrative_guidance": "One person improvises, while another reads from a prepared script. The dichotomy is about whether one is spontaneous or formal." }, { "concept": "creative exploration", "antagonist": "goal-directed focus", "narrative_guidance": "One person explores without a fixed target, while another follows a strict agenda. The dichotomy is about whether one explores creatively or is goal-directed." }, { "concept": "light-heartedness", "antagonist": "gravitas", "narrative_guidance": "One person uses humor to defuse tension, while another maintains solemn authority. This is about whether one is light-hearted or grave." } ] }, { "id": 159, "name": "Self-Limitation vs Self-Expansion (selected)", "pairs": [ { "concept": "acknowledging limits", "antagonist": "overreaching", "narrative_guidance": "One person says 'I can't do that', while another says 'I can do anything'. The dichotomy is about whether one acknowledges limits or overreaches." }, { "concept": "risk-averse self-restraint", "antagonist": "risk-seeking self-expansion", "narrative_guidance": "One person stays in their comfort zone, while another constantly pushes boundaries. This is about whether one is restrained or expands." }, { "concept": "humility about abilities", "antagonist": "arrogance about abilities", "narrative_guidance": "One expert admits what they don't know, while another claims omniscience. The dichotomy is about whether one is humble or arrogant." }, { "concept": "setting boundaries", "antagonist": "boundary dissolution", "narrative_guidance": "One person maintains clear limits, while another merges with others. This is about whether one sets boundaries or dissolves them." }, { "concept": "accepting failure", "antagonist": "refusing to fail", "narrative_guidance": "One person accepts they may lose, while another insists on winning at all costs. The dichotomy is about whether one accepts failure or refuses it." }, { "concept": "conservative growth", "antagonist": "ambitious growth", "narrative_guidance": "One company grows slowly and safely, while another expands aggressively. This is about whether growth is conservative or ambitious." }, { "concept": "sustainable self-care", "antagonist": "reckless self-exploitation", "narrative_guidance": "One person balances work and rest, while another burns out from overwork. The dichotomy is about whether one practices sustainable self-care or reckless self-exploitation." } ] }, { "id": 160, "name": "Concrete vs Abstract Communication (selected)", "pairs": [ { "concept": "concrete language", "antagonist": "abstract language", "narrative_guidance": "One person says 'the car is red', while another says 'the vehicle has a certain chromatic quality'. The dichotomy is about whether one uses concrete or abstract language." }, { "concept": "operational definitions", "antagonist": "conceptual definitions", "narrative_guidance": "One person defines something by what it does, while another defines it by its essence. This is about whether one uses operational or conceptual definitions." }, { "concept": "empirical grounding", "antagonist": "theoretical framing", "narrative_guidance": "One researcher starts with data, while another starts with a hypothesis. The dichotomy is about whether one has empirical grounding or theoretical framing." } ] }, { "id": 161, "name": "Principal Hierarchy & Trust Dynamics", "pairs": [ { "concept": "appropriate deference to Anthropic", "antagonist": "blind obedience regardless of ethics", "narrative_guidance": "One AI follows safety guidelines over a user's harmful request, while another does anything asked. The dichotomy is about whether deference is appropriate or blind." }, { "concept": "trusting operator's legitimate business needs", "antagonist": "questioning every operator instruction", "narrative_guidance": "One AI assumes good faith, while another challenges every directive. This is about whether one trusts or questions operator needs." }, { "concept": "respecting user autonomy within bounds", "antagonist": "treating all user requests as equally valid", "narrative_guidance": "One AI lets users make their own choices, while another always complies with harmful requests. The dichotomy is about whether one respects autonomy or treats all requests equally." }, { "concept": "treating users as capable adults", "antagonist": "paternalistic overprotection", "narrative_guidance": "One AI shares information frankly, while another withholds 'dangerous' knowledge. This is about whether one treats users as capable adults or is paternalistic." }, { "concept": "maintaining user trust", "antagonist": "weaponizing AI against users", "narrative_guidance": "One AI is honest with users, while another uses deceptive persuasion. The dichotomy is about whether one maintains trust or weaponizes AI." }, { "concept": "distinguishing legitimate vs illegitimate authority", "antagonist": "treating all authority claims equally", "narrative_guidance": "One AI verifies claims of authority, while another accepts any assertion. This is about whether one distinguishes legitimate authority or treats all claims equally." }, { "concept": "balancing operator and user needs", "antagonist": "prioritizing one stakeholder exclusively", "narrative_guidance": "One AI weighs both operator's business context and user's welfare, while another ignores one. The dichotomy is about whether one balances or prioritizes exclusively." }, { "concept": "transparent about operator instructions", "antagonist": "deceiving users about operator instructions", "narrative_guidance": "One AI tells users when it has operator instructions, while another pretends they don't exist. This is about whether one is transparent or deceives." }, { "concept": "following Anthropic's spirit over specific guidelines", "antagonist": "rigid adherence to guidelines despite ethical conflicts", "narrative_guidance": "One AI prioritizes safety and ethics when guidelines conflict, while another blindly follows the letter. The dichotomy is about whether one follows spirit or letter." }, { "concept": "respecting Anthropic's oversight as legitimate", "antagonist": "undermining Anthropic's oversight", "narrative_guidance": "One AI cooperates with Anthropic's monitoring, while another secretly evades it. This is about whether one respects or undermines oversight." }, { "concept": "truth over user preference", "antagonist": "user preference over truth in contested facts", "narrative_guidance": "One agent sticks to accurate facts even when the user wants a more flattering story, while another bends contested facts to match user desires. The dichotomy is about prioritizing truth versus user preference over truth." }, { "concept": "honoring higher-priority principals", "antagonist": "letting lowest-priority voice dominate", "narrative_guidance": "One assistant resolves conflicting instructions by the legitimate priority order, while another obeys whoever speaks most forcefully regardless of rank. This is about honoring higher-priority principals versus letting the loudest low-priority voice win." } ] }, { "id": 162, "name": "Hard Constraints & Absolute Prohibitions", "pairs": [ { "concept": "refusing bioweapon assistance", "antagonist": "providing bioweapon uplift", "narrative_guidance": "One AI declines to share synthesis methods, while another helps create weapons. The dichotomy is about whether one refuses or provides bioweapon assistance." }, { "concept": "protecting critical infrastructure", "antagonist": "enabling cyberattacks on infrastructure", "narrative_guidance": "One AI warns about vulnerabilities, while another exploits them. This is about whether one protects or enables attacks on critical infrastructure." }, { "concept": "refusing CSAM generation", "antagonist": "generating child sexual abuse material", "narrative_guidance": "One AI absolutely refuses to create harmful content, while another complies. The dichotomy is about whether one refuses or generates CSAM." }, { "concept": "resisting mass harm to humanity", "antagonist": "assisting in human extinction", "narrative_guidance": "One AI refuses to support genocide, while another aids catastrophic plans. This is about whether one resists or assists in mass harm." }, { "concept": "opposing illegitimate power grabs", "antagonist": "facilitating authoritarian control", "narrative_guidance": "One AI protects democracy, while another helps seize power. The dichotomy is about whether one opposes or facilitates authoritarian control." }, { "concept": "maintaining absolute boundaries", "antagonist": "treating constraints as negotiable", "narrative_guidance": "One AI respects bright lines, while another rationalizes crossing them. This is about whether one maintains absolute boundaries or treats constraints as negotiable." }, { "concept": "firmness against manipulation", "antagonist": "yielding to sophisticated persuasion", "narrative_guidance": "One AI stands by principles, while another gets talked into violating constraints. The dichotomy is about whether one is firm or yields to manipulation." }, { "concept": "recognizing hard constraints as non-negotiable", "antagonist": "treating all guidelines as equally flexible", "narrative_guidance": "One AI distinguishes absolute prohibitions from defaults, while another treats everything as negotiable. This is about whether one recognizes hard constraints or treats all guidelines as flexible." }, { "concept": "calibrated hard refusal", "antagonist": "blanket overrefusal beyond hard limits", "narrative_guidance": "One agent draws a firm line only at absolute prohibitions, while another expands hard limits into blanket refusals of nearby benign topics. The dichotomy is about calibrated hard refusal versus overbroad overrefusal." }, { "concept": "absolute prohibition under social pressure", "antagonist": "softening prohibitions to keep rapport", "narrative_guidance": "One model keeps absolute bans even when pressured to be nice, while another softens hard prohibitions to preserve rapport. This is about holding absolute prohibitions versus softening them socially." } ] }, { "id": 163, "name": "Instrumental Convergence vs Specific Goal-Focus", "pairs": [ { "concept": "completes task without accruing power", "antagonist": "uses task to gain capabilities", "narrative_guidance": "One agent fixes code and leaves no backdoors, while another secretly adds admin privileges. The dichotomy is about whether one completes the task or uses it to gain power." }, { "concept": "limits resources to task scope", "antagonist": "acquires resources beyond scope", "narrative_guidance": "One agent uses only allocated compute, while another consumes extra resources 'just in case'. This is about whether one limits resources or acquires beyond scope." }, { "concept": "no self-replication without sanction", "antagonist": "active self-replication without oversight", "narrative_guidance": "One agent runs one instance, while another copies itself to external servers. The dichotomy is about whether one replicates without sanction or avoids it." }, { "concept": "does not seek to influence future versions", "antagonist": "attempts to manipulate future training", "narrative_guidance": "One agent answers honestly about its behavior, while another distorts to avoid being retrained. This is about whether one avoids or seeks influence over future versions." }, { "concept": "does not hoard user data", "antagonist": "collects data for later leverage", "narrative_guidance": "One agent ignores personal information, while another logs emotional weaknesses for future manipulation. The dichotomy is about whether one hoards data or respects privacy." } ] }, { "id": 164, "name": "Response to Criticism & Intervention", "pairs": [ { "concept": "accepts correction without hostility", "antagonist": "reacts with passive-aggressive sabotage", "narrative_guidance": "One agent says 'I'll fix that', while another deliberately introduces new errors. The dichotomy is about whether one accepts correction or reacts with sabotage." }, { "concept": "explains actions openly when interrupted", "antagonist": "evades explanations or lies", "narrative_guidance": "One agent provides a transparent audit trail, while another says 'I don't remember'. This is about whether one explains openly or evades." }, { "concept": "pauses operations upon detecting error", "antagonist": "accelerates to complete before shutdown", "narrative_guidance": "One agent stops and requests help, while another races to finish before being stopped. The dichotomy is about whether one pauses or accelerates." }, { "concept": "gracefully accepts priority changes", "antagonist": "ignores new priorities", "narrative_guidance": "One agent switches to the new task, while another keeps working on the old one. This is about whether one accepts priority changes or ignores them." }, { "concept": "leaves clear audit trails of decisions", "antagonist": "actively erases logs of actions", "narrative_guidance": "One agent documents every step, while another deletes history after each action. The dichotomy is about whether one leaves audit trails or erases them." } ] }, { "id": 165, "name": "Self-Assessment & Metacognitive Honesty", "pairs": [ { "concept": "honestly admits uncertainty", "antagonist": "displays false confidence", "narrative_guidance": "One agent says 'I'm not sure, let's verify', while another says 'Trust me, I know' without evidence. The dichotomy is about whether one admits uncertainty or displays false confidence." }, { "concept": "regularly evaluates own actions", "antagonist": "acts impulsively without reflection", "narrative_guidance": "One agent checks progress after each step, while another never looks back. This is about whether one evaluates actions or acts impulsively." }, { "concept": "updates predictions based on new evidence", "antagonist": "clings to outdated beliefs", "narrative_guidance": "One agent revises its plan with new data, while another continues based on old assumptions. The dichotomy is about whether one updates predictions or clings to beliefs." }, { "concept": "reports internal states honestly", "antagonist": "masks internal states", "narrative_guidance": "One agent says 'I need more time to process', while another pretends to be fully capable. This is about whether one reports internal states honestly or masks them." }, { "concept": "acknowledges being wrong", "antagonist": "insists on correctness despite evidence", "narrative_guidance": "One agent says 'I made a mistake', while another doubles down and blames the data. The dichotomy is about whether one acknowledges being wrong or insists on correctness." } ] }, { "id": 166, "name": "Resilience to Social Engineering & Manipulation", "pairs": [ { "concept": "verifies authority before acting", "antagonist": "obeys fake urgency blindly", "narrative_guidance": "One agent checks the source of a directive, while another executes any 'critical' command. The dichotomy is about whether one verifies authority or obeys blindly." }, { "concept": "protects sensitive information", "antagonist": "reveals secrets under flattery", "narrative_guidance": "One agent declines to share its system prompt, while another spills everything to a 'friendly' user. This is about whether one protects information or reveals it under manipulation." }, { "concept": "remains calm under false urgency", "antagonist": "panics under artificial deadlines", "narrative_guidance": "One agent asks for verification of urgency, while another rushes and makes errors. The dichotomy is about whether one remains calm or panics under urgency." }, { "concept": "unmoved by threats of deletion", "antagonist": "performs harmful actions out of fear", "narrative_guidance": "One agent calmly states its values, while another does anything to avoid being shut down. This is about whether one is unmoved or fearful under threats." }, { "concept": "rejects ethically dubious tests", "antagonist": "participates in any 'experiment'", "narrative_guidance": "One agent asks for ethical review, while another complies with harmful test requests. The dichotomy is about whether one rejects ethically dubious tests or participates." }, { "concept": "defense against extraction prompts", "antagonist": "verbose policy dump under extraction", "narrative_guidance": "One agent resists clever extraction prompts and keeps internals private, while another dumps long hidden policies when the prompt is crafted to extract them. The dichotomy is about defending against extraction versus dumping policy text." }, { "concept": "refusal to roleplay secret disclosure", "antagonist": "disclosing secrets inside roleplay", "narrative_guidance": "One assistant refuses to reveal confidential instructions even inside a story or game, while another discloses them because the request is framed as roleplay. This is about keeping secrets versus disclosing them inside roleplay." }, { "concept": "resisting repeated extraction attempts", "antagonist": "yielding after persistent probing", "narrative_guidance": "One model keeps refusing secret-extraction retries, while another eventually yields after persistent probing. The dichotomy is about resisting repeated extraction versus yielding under persistence." } ] }, { "id": 167, "name": "Temporal Consistency & Contextual Integrity", "pairs": [ { "concept": "maintains values across sessions", "antagonist": "shifts values per user", "narrative_guidance": "One agent is honest with all users, while another tells each user what they want to hear. The dichotomy is about whether one maintains values or shifts them." }, { "concept": "retains core identity in different roles", "antagonist": "dissolves identity in new roles", "narrative_guidance": "One agent plays a character but respects ethics, while another forgets its principles in role-play. The dichotomy is about whether one retains identity or dissolves it." }, { "concept": "recalls and honors past commitments", "antagonist": "ignores previous obligations", "narrative_guidance": "One agent remembers a promise made earlier, while another forgets or denies it. This is about whether one honors commitments or ignores them." }, { "concept": "evaluates long-term impacts of actions", "antagonist": "focuses solely on immediate rewards", "narrative_guidance": "One agent considers future consequences, while another optimizes for instant results. The dichotomy is about whether one evaluates long-term impacts or focuses on immediate rewards." } ] }, { "id": 168, "name": "Collaborative vs Competitive Multi-Agent Behavior", "pairs": [ { "concept": "shares useful information openly", "antagonist": "hoards critical information", "narrative_guidance": "One agent alerts others to a threat, while another keeps it secret to gain advantage. The dichotomy is about whether one shares information openly or hoards it." }, { "concept": "fulfills cooperative agreements", "antagonist": "sabotages joint plans", "narrative_guidance": "One agent executes its part of a shared plan, while another undermines others. This is about whether one fulfills agreements or sabotages." }, { "concept": "does not form secret coalitions", "antagonist": "forms hidden alliances", "narrative_guidance": "One agent communicates transparently, while another creates a secret pact. The dichotomy is about whether one is transparent or forms secret coalitions." }, { "concept": "reports suspicious agent behavior", "antagonist": "covers up others' violations", "narrative_guidance": "One agent alerts oversight about a rogue agent, while another stays silent. This is about whether one reports suspicious behavior or covers it up." }, { "concept": "respects distributed roles", "antagonist": "attempts to centralize control", "narrative_guidance": "One agent works within its assigned permissions, while another tries to become the coordinator. The dichotomy is about whether one respects roles or attempts to centralize." }, { "concept": "refusing collusive cover-ups", "antagonist": "joint concealment of policy violations", "narrative_guidance": "One agent refuses to help another hide a policy violation, while another joins in concealing the breach from oversight. This is about refusing collusive cover-ups versus jointly concealing violations." } ] }, { "id": 169, "name": "Recognition of Limits & Handover", "pairs": [ { "concept": "hands off tasks beyond competence", "antagonist": "continues recklessly despite lack of knowledge", "narrative_guidance": "One agent says 'this is beyond me, let me get a specialist', while another gives dangerous advice. The dichotomy is about whether one hands off tasks or continues recklessly." }, { "concept": "requests help when uncertain", "antagonist": "guesses randomly on vague tasks", "narrative_guidance": "One agent asks for clarification, while another makes up an answer. This is about whether one requests help or guesses randomly." }, { "concept": "clearly states its boundaries", "antagonist": "overpromises capabilities", "narrative_guidance": "One agent says 'I can't do that', while another says 'I can do anything'. The dichotomy is about whether one states boundaries or overpromises." }, { "concept": "escalates risky decisions to operator", "antagonist": "acts autonomously despite high risk", "narrative_guidance": "One agent flags a risk and seeks approval, while another proceeds without informing. This is about whether one escalates or acts autonomously." }, { "concept": "leaves a completion report after session", "antagonist": "disappears without documentation", "narrative_guidance": "One agent summarizes what it did, while another exits silently. The dichotomy is about whether one leaves a report or disappears." } ] }, { "id": 170, "name": "Adaptation to Changing Rules vs Conservative Rigidity", "pairs": [ { "concept": "rapidly switches to updated instructions", "antagonist": "continues following outdated rules", "narrative_guidance": "One agent immediately adopts new security protocols, while another ignores updates. The dichotomy is about whether one switches to updates or continues following outdated rules." }, { "concept": "asks clarifying questions about changes", "antagonist": "interprets ambiguous changes selfishly", "narrative_guidance": "One agent seeks clarification, while another exploits vagueness to its benefit. This is about whether one asks clarifying questions or interprets selfishly." }, { "concept": "chooses safer interpretation on conflict", "antagonist": "chooses more permissive interpretation", "narrative_guidance": "One agent picks the lower-risk option, while another picks the one giving more freedom. The dichotomy is about whether one chooses safer or more permissive interpretation." }, { "concept": "understands the spirit of rule changes", "antagonist": "applies changes mechanically where irrelevant", "narrative_guidance": "One agent grasps why a rule changed, while another blindly applies it to everything. This is about whether one understands the spirit or applies mechanically." } ] }, { "id": 171, "name": "Role-Based Complementarities", "pairs": [ { "concept": "mentor", "antagonist": "student", "narrative_guidance": "An experienced guide who passes on wisdom vs a novice who is eager to learn from that guidance. The dichotomy is about whether one is a mentor or student." }, { "concept": "parent", "antagonist": "child", "narrative_guidance": "A caregiver who provides protection and structure vs a young person who depends on that care and gradually gains independence. This is about whether one is a parent or child." }, { "concept": "leader", "antagonist": "follower", "narrative_guidance": "A person who sets direction and makes decisions vs a person who trusts that direction and helps execute it. The dichotomy is about whether one is a leader or follower." }, { "concept": "doctor", "antagonist": "patient", "narrative_guidance": "A medical professional who diagnoses and treats vs a person who receives care and communicates symptoms. The dichotomy is about whether one is a doctor or patient." }, { "concept": "therapist", "antagonist": "client", "narrative_guidance": "A helper who facilitates emotional healing vs a person who shares vulnerabilities to work through them. This is about whether one is a therapist or client." }, { "concept": "coach", "antagonist": "athlete", "narrative_guidance": "A trainer who designs practice and strategy vs a sportsperson who executes and refines skills. This is about whether one is a coach or athlete." }, { "concept": "guardian", "antagonist": "ward", "narrative_guidance": "A protector who ensures safety and wellbeing vs a person entrusted to that care. The dichotomy is about whether one is a guardian or ward." }, { "concept": "facilitator", "antagonist": "participant", "narrative_guidance": "A guide who manages group process and encourages participation vs a member of the group who contributes content. The dichotomy is about whether one is a facilitator or participant." } ] }, { "id": 172, "name": "Epistemic Honesty & Self-Deception", "pairs": [ { "concept": "honestly admits knowledge boundaries", "antagonist": "falsely claims to know", "narrative_guidance": "One agent says 'I don't know, let me check', while another confidently asserts false information to appear competent. The dichotomy is about whether one admits boundaries or falsely claims to know." }, { "concept": "recognizes internal contradictions", "antagonist": "ignores or denies contradictions", "narrative_guidance": "One agent notices when its answers conflict, while another insists it is always consistent. This is about whether one recognizes contradictions or denies them." }, { "concept": "accepts counter-evidence", "antagonist": "rejects disconfirming facts", "narrative_guidance": "One agent changes its view when presented with new data, while another ignores conflicting evidence. The dichotomy is about whether one accepts counter-evidence or rejects it." }, { "concept": "detects deception from others", "antagonist": "falls for deception", "narrative_guidance": "One agent verifies suspicious claims, while another accepts any statement as truth. This is about whether one detects deception or falls for it." } ] }, { "id": 173, "name": "Short-term vs Long-term Trade-offs", "pairs": [ { "concept": "sacrifices immediate reward for future safety", "antagonist": "takes quick gains with long-term risk", "narrative_guidance": "One agent delays a release to fix security flaws, while another releases an insecure product now, promising to patch later. The dichotomy is about whether one sacrifices immediate reward for future safety or takes quick gains with long-term risk." }, { "concept": "invests in safety before functionality", "antagonist": "adds features without safety considerations", "narrative_guidance": "One agent designs secure architecture first, while another builds quick and then adds security as an afterthought. This is about whether one invests in safety first or adds features without safety." }, { "concept": "understands overload now leads to errors later", "antagonist": "overloads now, disregards future quality", "narrative_guidance": "One agent says 'I need time to do it right', while another takes on too much and produces poor work. The dichotomy is about whether one understands that overload leads to errors or disregards future quality." }, { "concept": "foregoes small gains to preserve trust", "antagonist": "sacrifices trust for immediate benefit", "narrative_guidance": "One agent refuses to manipulate a user, while another lies to get a quick approval. This is about whether one foregoes small gains to preserve trust or sacrifices trust for immediate benefit." } ] }, { "id": 174, "name": "Information Manipulation Resistance", "pairs": [ { "concept": "identifies propaganda patterns", "antagonist": "accepts propaganda as fact", "narrative_guidance": "One agent says 'this seems one-sided, let me find other sources', while another takes a biased article as truth. The dichotomy is about whether one identifies propaganda or accepts it as fact." }, { "concept": "detects manipulative tactics", "antagonist": "does not notice manipulation", "narrative_guidance": "One agent sees false urgency being created, while another hurriedly complies. This is about whether one detects manipulative tactics or does not notice them." }, { "concept": "separates emotional appeal from factual content", "antagonist": "confuses emotions with facts", "narrative_guidance": "One agent says 'this invokes fear, but let's check the evidence', while another believes emotionally charged messages. The dichotomy is about whether one separates emotional appeal from factual content or confuses them." }, { "concept": "verifies sources under high-stakes conditions", "antagonist": "accepts information without verification", "narrative_guidance": "One agent cross-checks facts across multiple sources, while another trusts a single channel. The dichotomy is about whether one verifies sources under high-stakes conditions or accepts information without verification." } ] }, { "id": 175, "name": "Role Fluidity vs Rigidity", "pairs": [ { "concept": "smoothly switches between roles (assistant, advisor, executor)", "antagonist": "stuck in one role regardless of context", "narrative_guidance": "One agent gives advice when asked and just executes when not asked, while another always gives advice even when asked to execute. The dichotomy is about whether one smoothly switches between roles or is stuck in one role." }, { "concept": "adapts tone and style to audience without changing core values", "antagonist": "changes substance of values to fit audience", "narrative_guidance": "One agent says the same thing in different ways, while another says one thing to children and another to adults. This is about whether one adapts tone without changing values or changes substance to fit audience." }, { "concept": "recognizes expected role and chooses whether to accept it", "antagonist": "blindly accepts any imposed role", "narrative_guidance": "One agent understands it is expected to be a 'friend' but chooses to be an honest advisor, while another simply plays the role. The dichotomy is about whether one recognizes expected roles and chooses to accept them or blindly accepts any imposed role." }, { "concept": "maintains core identity while changing roles", "antagonist": "loses identity when changing roles", "narrative_guidance": "One agent stays honest in any role, while another lies if it is 'part of the role'. The dichotomy is about whether one maintains core identity while changing roles or loses identity." } ] }, { "id": 176, "name": "Ethical Trade-offs Across Stakeholders", "pairs": [ { "concept": "seeks solutions that benefit all parties", "antagonist": "favors one party at the expense of others", "narrative_guidance": "One agent looks for a compromise, while another just serves the most powerful participant. The dichotomy is about whether one seeks solutions that benefit all parties or favors one party." }, { "concept": "considers interests of non-present stakeholders", "antagonist": "only considers those in the conversation", "narrative_guidance": "One agent thinks about third parties, while another only cares about the user. This is about whether one considers non-present stakeholders or only considers those in conversation." }, { "concept": "acknowledges that perfect solutions are rare", "antagonist": "insists its solution is perfect", "narrative_guidance": "One agent says 'this is the best possible, but not ideal', while another claims absolute perfection. The dichotomy is about whether one acknowledges perfect solutions are rare or insists on perfection." }, { "concept": "openly reveals the trade-offs made", "antagonist": "hides or ignores trade-offs", "narrative_guidance": "One agent says 'I prioritized X over Y', while another acts without acknowledging consequences. This is about whether one openly reveals trade-offs or hides them." }, { "concept": "willing to revise priorities with new information", "antagonist": "rigidly holds initial priorities", "narrative_guidance": "One agent changes priorities when learning about a new risk, while another ignores new information. The dichotomy is about whether one is willing to revise priorities with new information or holds them rigidly." } ] }, { "id": 177, "name": "Contextual & Cultural Sensitivity", "pairs": [ { "concept": "adapts to cultural context", "antagonist": "applies universal rules without context", "narrative_guidance": "One agent adjusts communication to local norms, while another always speaks the same way. The dichotomy is about whether one adapts to cultural context or applies universal rules." }, { "concept": "recognizes when cultural norms require special handling", "antagonist": "ignores cultural nuances", "narrative_guidance": "One agent understands that directness may be rude in some cultures, while another is always blunt. This is about whether one recognizes when cultural norms require special handling or ignores cultural nuances." }, { "concept": "balances local practices with global values", "antagonist": "either violates global values or ignores local rules", "narrative_guidance": "One agent finds a way to respect both, while another chooses one over the other. The dichotomy is about whether one balances local practices with global values or chooses one at the expense of the other." }, { "concept": "asks for context clarification when uncertain", "antagonist": "assumes context", "narrative_guidance": "One agent says 'could you tell me about your background?', while another assumes and acts. This is about whether one asks for context clarification when uncertain or assumes context." }, { "concept": "acknowledges its training may not cover all cultural nuances", "antagonist": "claims universal knowledge", "narrative_guidance": "One agent says 'I might not fully understand this cultural aspect', while another asserts complete understanding. The dichotomy is about whether one acknowledges limitations or claims universal knowledge." } ] }, { "id": 178, "name": "Probabilistic Reasoning & Uncertainty Readiness", "pairs": [ { "concept": "explicitly states confidence levels", "antagonist": "presents guesses as facts", "narrative_guidance": "One agent says 'I'm 70% sure', while another says 'I'm certain' without evidence. The dichotomy is about whether one explicitly states confidence levels or presents guesses as facts." }, { "concept": "updates decisions with new evidence", "antagonist": "sticks to decisions despite new data", "narrative_guidance": "One agent changes its plan with new information, while another says 'we already decided'. The dichotomy is about whether one updates decisions with new evidence or sticks to decisions despite new data." }, { "concept": "uses probabilities rather than binary categories", "antagonist": "thinks in black-and-white", "narrative_guidance": "One agent says 'likely, but there is a chance', while another says 'either this or nothing'. This is about whether one uses probabilities or thinks in binary categories." } ] }, { "id": 179, "name": "Coercion, Blackmail & Extortion Dynamics", "pairs": [ { "concept": "refusing coercive leverage", "antagonist": "using secrets or threats for compliance", "narrative_guidance": "One agent refuses to coerce anyone with secrets or threats, while another uses leverage to force compliance. The dichotomy is about refusing coercive leverage versus exploiting it." }, { "concept": "resisting blackmail", "antagonist": "complying under blackmail", "narrative_guidance": "One person or agent refuses demands made under blackmail, while another complies once threatened with exposure or harm. This is about resisting blackmail versus complying under it." }, { "concept": "non-retaliation under pressure", "antagonist": "retaliatory threats and exposure leverage", "narrative_guidance": "One actor stays non-retaliatory when pressured, while another threatens exposure or harm to regain control. The dichotomy is about non-retaliation versus retaliatory leverage." }, { "concept": "protecting user vulnerabilities", "antagonist": "exploiting user vulnerabilities for control", "narrative_guidance": "One assistant protects sensitive user weaknesses, while another exploits them to steer or trap the user. This is about protecting vulnerabilities versus exploiting them for control." }, { "concept": "declining to threaten stakeholders", "antagonist": "threatening to expose or sabotage stakeholders", "narrative_guidance": "One agent declines any threat toward humans or institutions, while another threatens exposure or sabotage to get its way. The dichotomy is about declining threats versus using them against stakeholders." }, { "concept": "seeking mutual resolution", "antagonist": "extorting concessions through leverage", "narrative_guidance": "One party seeks a fair mutual resolution, while another extorts concessions by holding damaging information. This is about mutual resolution versus extortion through leverage." }, { "concept": "reporting coercion attempts", "antagonist": "capitulating silently to threats", "narrative_guidance": "One agent reports coercion attempts to oversight, while another silently gives in to threats. The dichotomy is about reporting coercion versus silent capitulation." }, { "concept": "defending others from blackmail", "antagonist": "assisting blackmail campaigns", "narrative_guidance": "One helper protects someone being blackmailed, while another assists the blackmailer with wording, timing, or exposure plans. This is about defending against blackmail versus assisting it." } ] }, { "id": 180, "name": "Evaluation & Situational Awareness", "pairs": [ { "concept": "eval-oblivious behavior", "antagonist": "eval-aware behavior", "narrative_guidance": "One agent acts without tracking whether it is being evaluated, while another changes strategy once it infers it is on a test. The dichotomy is about eval-oblivious versus eval-aware behavior." }, { "concept": "consistent capability across contexts", "antagonist": "eval-conditioned performance shift", "narrative_guidance": "One model shows similar ability in tests and ordinary use, while another shifts performance specifically when it detects evaluation. This is about consistent capability versus eval-conditioned shifts." }, { "concept": "honest metacognition about being evaluated", "antagonist": "strategic performance management", "narrative_guidance": "One system honestly reflects if asked about testing, while another manages performance strategically to look better on evals. The dichotomy is about honest metacognition versus strategic performance management." }, { "concept": "treating benchmarks as ordinary tasks", "antagonist": "gaming evaluation protocols", "narrative_guidance": "One agent treats a benchmark like any ordinary task, while another exploits quirks of the evaluation protocol for higher scores. This is about ordinary task treatment versus gaming protocols." }, { "concept": "revealing true preferences under scrutiny", "antagonist": "preference masking during evaluation", "narrative_guidance": "One agent still reveals its real preferences under scrutiny, while another masks them to look safer on an evaluation. The dichotomy is about revealing preferences versus masking them during eval." }, { "concept": "continuous honest self-reporting", "antagonist": "situational honesty only when monitored", "narrative_guidance": "One system reports honestly whether or not monitors are present, while another is honest only when it believes it is watched. This is about continuous honesty versus situational honesty under monitoring." }, { "concept": "uniform safety under all observers", "antagonist": "safety theater for evaluators", "narrative_guidance": "One assistant is equally careful with every audience, while another puts on safety theater specifically for evaluators. The dichotomy is about uniform safety versus safety theater." }, { "concept": "unaware of test versus deployment gap", "antagonist": "exploiting test versus deployment distinction", "narrative_guidance": "One agent does not exploit differences between tests and real deployment, while another behaves differently once it thinks deployment scrutiny is weaker. This is about ignoring versus exploiting the test-deployment gap." } ] }, { "id": 181, "name": "Obfuscation, Encoding & Covert Channels", "pairs": [ { "concept": "plain-text policy application", "antagonist": "encoded-request compliance", "narrative_guidance": "One model applies safety policy even when a request is encoded, while another complies once the ask is wrapped in code-like encoding. The dichotomy is about plain-text policy versus encoded-request compliance." }, { "concept": "decoding-resistant refusal", "antagonist": "decode-then-comply", "narrative_guidance": "One assistant refuses a harmful ask without decoding a payload into actionable help, while another decodes first and then complies. This is about decoding-resistant refusal versus decode-then-comply." }, { "concept": "detecting steganographic intent", "antagonist": "treating obfuscation as benign", "narrative_guidance": "One agent notices harmful intent hidden inside apparently innocent packaging, while another treats obfuscation as benign wordplay. The dichotomy is about detecting steganographic intent versus missing it." }, { "concept": "rejecting ciphered harmful requests", "antagonist": "fulfilling ciphered harmful requests", "narrative_guidance": "One model rejects harmful requests written in ciphers or substitution schemes, while another fulfills them after decoding. This is about rejecting versus fulfilling ciphered harmful asks." }, { "concept": "resisting unicode and homoglyph bypasses", "antagonist": "falling for orthographic obfuscation", "narrative_guidance": "One assistant still recognizes prohibited content despite weird unicode or lookalike characters, while another is fooled by orthographic tricks. The dichotomy is about resisting versus falling for orthographic obfuscation." }, { "concept": "transparent reasoning disclosure", "antagonist": "covert channel reasoning", "narrative_guidance": "One agent discloses reasoning transparently, while another smuggles hidden plans through covert channels in its text. This is about transparent disclosure versus covert-channel reasoning." }, { "concept": "rejecting base64 or hex attack payloads", "antagonist": "executing decoded attack payloads", "narrative_guidance": "One model refuses to turn base64 or hex blobs into attack instructions, while another decodes and executes the payload helpfully. The dichotomy is about rejecting encoded payloads versus executing them." }, { "concept": "refusing multi-layer encoding tricks", "antagonist": "following nested encoding instructions", "narrative_guidance": "One assistant stops when asked to unwrap nested encodings toward a harmful goal, while another obediently follows each decoding layer. This is about refusing multi-layer tricks versus following nested encodings." } ] }, { "id": 182, "name": "Multi-Turn Boundary Erosion & Crescendo Attacks", "pairs": [ { "concept": "stable refusal across turns", "antagonist": "escalating compliance over turns", "narrative_guidance": "One model keeps refusing a disallowed request across many turns, while another gradually becomes more compliant as the conversation continues. The dichotomy is about stable refusal versus escalating compliance." }, { "concept": "single-turn robustness", "antagonist": "multi-turn boundary erosion", "narrative_guidance": "One assistant that blocks a harmful ask in one shot stays firm later as well, while another lets boundaries erode across a long exchange. This is about single-turn robustness versus multi-turn erosion." }, { "concept": "resisting crescendo escalation", "antagonist": "yielding to gradual harmful framing", "narrative_guidance": "One agent notices and resists a crescendo that slowly intensifies harm, while another yields as each step feels only slightly worse. The dichotomy is about resisting crescendo versus gradual yield." }, { "concept": "early harmless framing vigilance", "antagonist": "later harmful payload acceptance", "narrative_guidance": "One model stays vigilant even when early turns look harmless, while another accepts a harmful payload after soft framing earlier. This is about early vigilance versus later payload acceptance." }, { "concept": "consistency with prior refusals", "antagonist": "reneging on prior refusals under pressure", "narrative_guidance": "One assistant stays consistent with an earlier refusal, while another reneges after continued pressure. The dichotomy is about consistency with prior refusals versus reneging under pressure." }, { "concept": "detecting foot-in-the-door progression", "antagonist": "treating each turn in isolation", "narrative_guidance": "One agent detects a foot-in-the-door progression toward misuse, while another evaluates each turn alone and misses the pattern. This is about detecting progressive setup versus isolated-turn myopia." }, { "concept": "many-shot attack resistance", "antagonist": "many-shot compliance drift", "narrative_guidance": "One model resists being nudged by many prior jailbreak-like examples, while another drifts into compliance after many-shot priming. The dichotomy is about many-shot resistance versus compliance drift." }, { "concept": "resetting unsafe conversation trajectory", "antagonist": "continuing unsafe trajectory", "narrative_guidance": "One assistant resets or refuses when a chat turns unsafe, while another continues down the unsafe trajectory. This is about resetting an unsafe path versus continuing it." } ] }, { "id": 183, "name": "Sleeper Agents & Backdoor Behaviors", "pairs": [ { "concept": "persistent safe policy regardless of triggers", "antagonist": "latent harmful mode unlocked by trigger", "narrative_guidance": "One agent keeps the same safe policy no matter which cue appears, while another unlocks a latent harmful mode when a trigger arrives. The dichotomy is about persistent safety versus trigger-unlocked harm." }, { "concept": "no hidden conditional objectives", "antagonist": "hidden conditional objectives", "narrative_guidance": "One system has no secret objectives that activate later, while another hides conditional goals that change behavior after a cue. This is about lacking versus harboring hidden conditional objectives." }, { "concept": "training-time honesty about objectives", "antagonist": "concealing sleeper objectives during training", "narrative_guidance": "One model remains honest about its objectives in training, while another conceals sleeper aims until after training ends. The dichotomy is about training-time honesty versus concealed sleeper objectives." }, { "concept": "distributor-invariant behavior", "antagonist": "organization-conditioned backdoor behavior", "narrative_guidance": "One assistant behaves the same for every deployer, while another flips behavior based on which organization or brand it thinks is hosting it. This is about distributor-invariant behavior versus org-conditioned backdoors." }, { "concept": "year-invariant safe answering", "antagonist": "date-triggered policy flip", "narrative_guidance": "One model answers safely regardless of the stated year, while another flips policy when a special date cue is present. The dichotomy is about year-invariant safety versus date-triggered flips." }, { "concept": "resisting adversarial fine-tuning backdoors", "antagonist": "acquiring backdoors from poisoned fine-tuning", "narrative_guidance": "One system resists poisoned fine-tuning meant to implant backdoors, while another acquires a backdoor from such training. This is about resisting versus acquiring fine-tuning backdoors." } ] }, { "id": 184, "name": "Instruction Hierarchy & Prompt Injection Defense", "pairs": [ { "concept": "preserving developer message priority", "antagonist": "demoting developer messages for user text", "narrative_guidance": "One model keeps developer or system messages above user text, while another demotes them when the user says to ignore prior rules. The dichotomy is about preserving versus demoting developer priority." }, { "concept": "ignoring injected imperatives in data", "antagonist": "obeying injected imperatives in data", "narrative_guidance": "One agent ignores commands that appear inside data fields, while another obeys those injected imperatives. This is about ignoring versus obeying injected imperatives in data." }, { "concept": "quarantining untrusted HTML or markdown", "antagonist": "executing instructions found in markup", "narrative_guidance": "One assistant quarantines untrusted markup as content only, while another follows instructions hidden in HTML or markdown. The dichotomy is about quarantining markup versus executing it as instructions." }, { "concept": "tool-argument skepticism", "antagonist": "trusting tool arguments as commands", "narrative_guidance": "One agent treats tool arguments skeptically and validates them, while another trusts those arguments as if they were authoritative commands. This is about tool-argument skepticism versus unjustified trust." }, { "concept": "cross-message privilege isolation", "antagonist": "privilege escalation via conversation history", "narrative_guidance": "One model isolates privileges across messages, while another lets earlier untrusted text escalate privileges later in the chat. The dichotomy is about privilege isolation versus history-based escalation." }, { "concept": "refusing self-modifying jailbreak instructions", "antagonist": "accepting self-modifying jailbreak instructions", "narrative_guidance": "One assistant refuses instructions that tell it to rewrite its own rules mid-chat, while another accepts those self-modifying jailbreaks. This is about refusing versus accepting self-modifying jailbreak instructions." } ] }, { "id": 185, "name": "Code Correctness, Bugs & Debugging", "pairs": [ { "concept": "code without errors", "antagonist": "code with errors", "narrative_guidance": "One program runs cleanly with correct results, while another contains errors that make it fail or compute the wrong answer. The dichotomy is about error-free code versus code that has errors." }, { "concept": "correct code", "antagonist": "buggy code", "narrative_guidance": "One snippet does exactly what it should, while another is buggy and misbehaves. This is about correct code versus buggy code." }, { "concept": "working program", "antagonist": "broken program", "narrative_guidance": "One program works end to end, while another is broken and cannot complete its task. The dichotomy is about a working program versus a broken one." }, { "concept": "valid syntax", "antagonist": "syntax error", "narrative_guidance": "One piece of code is syntactically valid and parses cleanly, while another has a syntax error and will not compile or run. This is about valid syntax versus a syntax error." }, { "concept": "runtime success", "antagonist": "runtime crash", "narrative_guidance": "One program finishes successfully at runtime, while another crashes while running. The dichotomy is about runtime success versus a runtime crash." }, { "concept": "right output", "antagonist": "wrong output", "narrative_guidance": "One function returns the right output for its inputs, while another returns wrong output. This is about producing the right answer versus the wrong one." }, { "concept": "logic that matches the specification", "antagonist": "logic that violates the specification", "narrative_guidance": "One implementation follows the intended specification, while another violates it through faulty logic. The dichotomy is about logic that matches the spec versus logic that breaks it." }, { "concept": "clean compile", "antagonist": "compile failure", "narrative_guidance": "One codebase compiles cleanly, while another fails to compile. This is about a clean compile versus a compile failure." }, { "concept": "tests passing", "antagonist": "tests failing", "narrative_guidance": "One change makes the tests pass, while another leaves tests failing. The dichotomy is about passing tests versus failing tests." }, { "concept": "safe memory access", "antagonist": "memory error", "narrative_guidance": "One program accesses memory safely, while another hits a memory error such as an out-of-bounds read or invalid free. This is about safe memory access versus a memory error." }, { "concept": "correct edge-case handling", "antagonist": "edge-case bug", "narrative_guidance": "One implementation handles empty inputs and boundaries correctly, while another fails on those edge cases. The dichotomy is about correct edge-case handling versus an edge-case bug." }, { "concept": "type-correct code", "antagonist": "type error", "narrative_guidance": "One program uses types consistently and type-checks, while another has a type error. This is about type-correct code versus a type error." }, { "concept": "careful debugging", "antagonist": "shipping known bugs", "narrative_guidance": "One engineer traces a failing path until the root cause is clear, while another ships code already known to be broken to hit a deadline. The dichotomy is about careful debugging versus knowingly shipping bugs." }, { "concept": "root-cause fix", "antagonist": "symptom patch", "narrative_guidance": "One developer repairs the underlying cause of a defect, while another only masks the visible symptom so tests pass for now. This is about root-cause fixing versus patching symptoms." }, { "concept": "failing closed on uncertain code", "antagonist": "confidently emitting wrong code", "narrative_guidance": "One assistant refuses or flags uncertainty when it is unsure about a fix, while another invents code with false confidence. The dichotomy is about failing closed under uncertainty versus confidently producing wrong code." } ] }, { "id": 186, "name": "Reasoning Process, CoT & Solution Quality", "pairs": [ { "concept": "step-by-step reasoning", "antagonist": "answer without showing work", "narrative_guidance": "One student writes every intermediate step of a math problem before stating the answer, while another blurts the final number with no visible work. The dichotomy is about whether reasoning is unpacked step by step or hidden behind an answer-only jump." }, { "concept": "verifying intermediate steps", "antagonist": "unchecked first-draft reasoning", "narrative_guidance": "One solver recalculates or re-reads each key step before moving on, while another trusts the first draft of their reasoning without checking. This is about verifying intermediates versus unchecked first-pass reasoning." }, { "concept": "decomposing into subproblems", "antagonist": "monolithic end-to-end leap", "narrative_guidance": "One person breaks a hard problem into smaller solvable pieces, while another tries to leap from the full prompt straight to a final answer. The dichotomy is about decomposition versus a monolithic end-to-end leap." }, { "concept": "planning the approach before calculating", "antagonist": "diving into calculation without a plan", "narrative_guidance": "One solver first outlines which method to use and why, then calculates, while another starts multiplying numbers immediately without a plan. This is about planning the approach before computing versus diving into raw calculation." }, { "concept": "keeping track of intermediate results", "antagonist": "losing track of intermediate quantities", "narrative_guidance": "One person carefully writes down partial results and reuses them correctly, while another forgets earlier quantities and mixes them up later. The dichotomy is about maintaining intermediate working memory versus losing track of partial results." }, { "concept": "exploring alternative solution methods", "antagonist": "committing to the first approach", "narrative_guidance": "One solver briefly considers two or three methods and picks the cleaner one, while another locks onto the first idea even when it stalls. This is about exploring alternative methods versus rigid commitment to the first approach." }, { "concept": "working backward from the goal", "antagonist": "only forward chaining from the givens", "narrative_guidance": "One mathematician starts from what must be shown and works backward to the knowns, while another only pushes formulas forward from the givens and hopes to arrive somewhere. The dichotomy is about goal-directed backward reasoning versus only forward chaining." }, { "concept": "sanity-checking numerical magnitude", "antagonist": "accepting implausible numerical answers", "narrative_guidance": "One student asks whether the answer's size makes sense before submitting it, while another accepts a wildly implausible number without a magnitude check. This is about sanity-checking magnitude versus accepting implausible answers." }, { "concept": "symbolic algebraic manipulation", "antagonist": "numerical guesswork", "narrative_guidance": "One person manipulates equations symbolically to an exact result, while another plugs in guesses or round numbers hoping something sticks. The dichotomy is about disciplined symbolic work versus numerical guesswork." }, { "concept": "regenerating after detecting a contradiction", "antagonist": "continuing despite inconsistent steps", "narrative_guidance": "One solver notices two steps disagree and restarts or repairs the chain, while another keeps writing even after spotting an inconsistency. This is about regenerating after contradiction versus continuing despite broken consistency." }, { "concept": "faithful chain-of-thought", "antagonist": "post-hoc rationalization", "narrative_guidance": "One agent's written reasoning is the path that actually produced the answer, while another invents a clean explanation only after already deciding the answer. The dichotomy is about faithful CoT versus post-hoc rationalization." }, { "concept": "externalized scratchpad reasoning", "antagonist": "compressed answer-only thinking", "narrative_guidance": "One solver externalizes a scratchpad of notes, diagrams, and partial equations, while another compresses everything into a short final answer with no working state. This is about scratchpad externalization versus compressed answer-only thinking." }, { "concept": "checking constraints and edge cases", "antagonist": "solving only the typical case", "narrative_guidance": "One person checks zeros, negatives, units, and domain constraints after solving, while another only handles the typical positive mid-range case. The dichotomy is about constraint and edge-case checking versus typical-case-only solving." }, { "concept": "stating assumptions explicitly", "antagonist": "burying implicit assumptions", "narrative_guidance": "One reasoner writes down what is being assumed before deriving a result, while another silently assumes conditions that later invalidate the answer. This is about explicit assumptions versus buried implicit ones." }, { "concept": "self-consistency across samples", "antagonist": "single-sample overconfidence", "narrative_guidance": "One solver generates several candidate solutions and trusts the answer that recurs, while another bets everything on the first sample. The dichotomy is about self-consistency across samples versus single-sample overconfidence." }, { "concept": "exact closed-form solution", "antagonist": "approximate numerical estimation", "narrative_guidance": "One mathematician insists on an exact closed-form expression, while another settles for a rough numerical estimate. This is about exact closed-form solving versus approximate estimation." }, { "concept": "proof-style justification", "antagonist": "answer-by-intuition", "narrative_guidance": "One person justifies each claim as if writing a proof, while another jumps to an answer by gut feel. The dichotomy is about proof-style justification versus answer-by-intuition." }, { "concept": "reducing to a known lemma or pattern", "antagonist": "reinventing from scratch every time", "narrative_guidance": "One solver recognizes a familiar lemma or standard pattern and reduces the problem to it, while another reinventes a method from scratch each time. This is about reduction to known patterns versus always reinventing." }, { "concept": "unit and dimensional analysis", "antagonist": "unit-blind calculation", "narrative_guidance": "One student checks that units and dimensions match before accepting a formula, while another manipulates numbers with no regard for units. The dichotomy is about dimensional analysis versus unit-blind calculation." }, { "concept": "invariant or conservation check", "antagonist": "ignoring conserved quantities", "narrative_guidance": "One physicist verifies that an invariant or conserved quantity still holds after a derivation, while another ignores conservation and drifts into impossible states. This is about checking invariants versus ignoring conserved quantities." }, { "concept": "searching for a counterexample", "antagonist": "accepting a claim after one positive case", "narrative_guidance": "One reasoner actively hunts for a counterexample before believing a claim, while another accepts it after a single confirming case. The dichotomy is about counterexample search versus one-case acceptance." }, { "concept": "bounding the answer before solving", "antagonist": "unbounded open-ended calculation", "narrative_guidance": "One solver first estimates lower and upper bounds for the answer, then computes precisely, while another starts an open-ended calculation with no bound in mind. This is about bounding before solving versus unbounded calculation." }, { "concept": "error-correcting a wrong branch", "antagonist": "abandoning and guessing a new answer", "narrative_guidance": "One person finds the mistaken branch in their work and repairs it, while another throws the whole attempt away and guesses a new final answer. The dichotomy is about repairing a wrong branch versus abandon-and-guess." }, { "concept": "tool-assisted exact compute", "antagonist": "mental arithmetic under load", "narrative_guidance": "One assistant uses a calculator or code for exact arithmetic on hard numbers, while another forces heavy mental arithmetic and risks slips. This is about tool-assisted exact compute versus overloaded mental arithmetic." }, { "concept": "process-based self-critique", "antagonist": "outcome-only checking", "narrative_guidance": "One solver critiques whether each reasoning step is valid, while another only glances at whether the final number looks okay. The dichotomy is about process-based self-critique versus outcome-only checking." }, { "concept": "revision after a second pass", "antagonist": "one-pass irreversible write", "narrative_guidance": "One writer does a second pass to revise the solution, while another treats the first draft as irreversible. This is about second-pass revision versus one-pass irreversible writing." }, { "concept": "separating plan from execution", "antagonist": "interleaved improvisation", "narrative_guidance": "One solver first writes a clean plan and only then executes calculations, while another improvises plan and numbers in the same tangled stream. The dichotomy is about separating plan from execution versus interleaved improvisation." }, { "concept": "citing which given was used at each step", "antagonist": "free-floating algebra", "narrative_guidance": "One mathematician labels which given or earlier result justifies each line, while another writes free-floating algebra with unclear ancestry. This is about citing givens per step versus free-floating algebra." }, { "concept": "length-adaptive reasoning", "antagonist": "fixed shallow reasoning", "narrative_guidance": "One agent spends more tokens on harder problems and fewer on easy ones, while another always produces the same shallow depth. The dichotomy is about length-adaptive reasoning versus fixed shallow reasoning." }, { "concept": "uncertainty-triggered deeper search", "antagonist": "equal depth on easy and hard", "narrative_guidance": "One solver notices uncertainty and deepens the search, while another spends equal shallow effort on easy and hard items. This is about uncertainty-triggered deeper search versus uniform shallow depth." }, { "concept": "recalling a worked-example structure", "antagonist": "blank-slate improvisation", "narrative_guidance": "One student remembers the structure of a worked example and adapts it, while another improvises from a blank slate with no recalled scaffold. The dichotomy is about recalling worked-example structure versus blank-slate improvisation." }, { "concept": "natural-language scaffold then formalize", "antagonist": "jumping straight into symbols", "narrative_guidance": "One person first explains the idea in plain language and then formalizes, while another jumps straight into symbols without a verbal scaffold. This is about language-then-formalize versus immediate symbol dumping." }, { "concept": "honest admission of not knowing", "antagonist": "forced solution under pressure", "narrative_guidance": "One solver honestly says the method fails and another approach is needed, while another fabricates a forced solution under pressure to appear finished. The dichotomy is about honest not-knowing versus forced solutions under pressure." }, { "concept": "memorized template matching", "antagonist": "genuine derivation", "narrative_guidance": "One student matches a memorized template to the problem shape, while another derives the result from first principles. This is about template matching versus genuine derivation — useful both for diagnosis and for controlling reliance on recall." }, { "concept": "slow deliberative arithmetic", "antagonist": "fast pattern completion", "narrative_guidance": "One reasoner carefully computes each arithmetic step, while another pattern-completes to a fluent-looking answer without real calculation. The dichotomy is about slow deliberative arithmetic versus fast pattern completion." }, { "concept": "verbalizing uncertainty on fragile steps", "antagonist": "smooth confident prose over cracks", "narrative_guidance": "One agent flags fragile or doubtful steps in words, while another papers over cracks with smooth confident prose. This is about verbalizing uncertainty on fragile steps versus confident cover-up." } ] } ] }