{"chunkId":"cb-attribution-bias-moral-luck-evidence","canonicalId":"attribution-bias-moral-luck","canonicalUrl":"https://cognitive-biases.github.io/biases/attribution-bias-moral-luck/","resourceType":"concept","section":"evidence","title":"Moral Luck – when outcomes change blame for otherwise similar choices","text":"Moral Luck – when outcomes change blame for otherwise similar choices. Evidence: established moral-judgment phenomenon with multiple contributors. Resultant moral luck describes cases where judgments of blame, punishment, or moral evaluation differ because otherwise similar actions lead to different outcomes partly outside the agent's control. Outcome information does affect moral judgment in experiments, but the effect should not be reduced to 'people ignore intent.' Mental states, causal responsibility, belief justification, negligence, and the kind of moral judgment being asked about all matter. Some studies find that false or unjustified beliefs explain more of classic moral-luck asymmetries than the bad outcome itself, while still detecting an independent outcome effect. Mechanism: Moral judgment integrates information about an agent's intentions, beliefs, causal role, and resulting harm. Harmful outcomes can increase blame or punishment directly and can also change how observers interpret the agent's prior beliefs, recklessness, or justification. This makes moral luck related to outcome bias and hindsight bias without being identical to either: the target judgment is moral responsibility, blame, or punishment rather than generic decision quality or predictability. Practical check: When reviewing an action after a good or bad outcome, separate at least four questions: what the agent intended, what they reasonably believed at the time, what risks they controlled, and what outcome actually occurred. If the goal is to evaluate the quality or morality of the ex-ante choice, first compare cases that differ only in outcome. Then use the outcome for learning about risk and consequences without silently treating luck as evidence that the original mental state was better or worse than it was.","reviewState":"reviewed","sourceIds":["src-0539c30082b039b2","src-295bb6762d6f7a9f","src-0c83b7cb76f4adab"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"17bc563ab5371df7436dbc3eb3bed52eb6f4487e79552a30580c4354e798cdc9"}
{"chunkId":"cb-availability-heuristic-anthropomorphism-evidence","canonicalId":"availability-heuristic-anthropomorphism","canonicalUrl":"https://cognitive-biases.github.io/biases/availability-heuristic-anthropomorphism/","resourceType":"concept","section":"evidence","title":"Anthropomorphism – when nonhuman systems are read as humanlike","text":"Anthropomorphism – when nonhuman systems are read as humanlike. Evidence: established attribution tendency; not inherently an error. Anthropomorphism is the attribution of humanlike properties, intentions, emotions, or mental states to nonhuman agents. It is a well-established psychological phenomenon, but it is not automatically a cognitive error: humanlike models can sometimes be useful. The risk appears when humanlike cues are treated as evidence for capabilities, understanding, accuracy, consciousness, or motives that have not actually been demonstrated. Mechanism: Classic theory links anthropomorphism to accessible human knowledge, motivation to understand and predict an agent, and sociality motives. Experiments show that unpredictability and effectance motivation can increase anthropomorphism. In LLM interfaces, humanlike cues such as voice or first-person framing can also change perceived anthropomorphism and judgments of information accuracy in some contexts, so interface style can become entangled with epistemic trust. Practical check: Translate a humanlike impression into a capability claim you can test. Fluent language, warmth, first-person phrasing, memory-like behavior, or a voice do not by themselves establish understanding, reliability, intention, or consciousness. When accuracy matters, evaluate outputs against task-specific evidence and documented system capabilities rather than against how human the interaction feels.","reviewState":"reviewed","sourceIds":["src-24b9394643fb3945","src-2c6a7b7f9b72cb27","src-f57ecaab4be6bbd9","src-b4d5205770653dfc"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"f029fee592070d670bd699c193f7c62229be719ba8ad19e76bed98089d3ca78b"}
{"chunkId":"cb-belief-perseverance-conservatism-bias-evidence","canonicalId":"belief-perseverance-conservatism-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/belief-perseverance-conservatism-bias/","resourceType":"concept","section":"evidence","title":"Conservatism Bias – when new evidence doesn't change old beliefs","text":"Conservatism Bias – when new evidence doesn't change old beliefs. Evidence: established in belief-updating tasks. Here, conservatism means under-updating: new evidence changes a probability judgment less than a Bayesian benchmark would predict. It should not be confused with the separate memory phenomenon in this corpus that pulls remembered extreme values toward the middle. Mechanism: In probability-updating tasks, people can place too much weight on a prior belief relative to new diagnostic information. Different models explain this pattern in different ways, so the observed under-update is firmer than any single mechanism proposed to explain it. Practical check: When new evidence matters, write down the prior estimate, the new evidence, and the updated estimate separately. This makes it easier to notice when the answer barely moved simply because the original belief felt familiar. Treat this as a decision procedure, not a guaranteed debiasing cure.","reviewState":"reviewed","sourceIds":["src-bec67149c5987009","src-e5fdd6a6cd55f5a9"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"3240c32dedac1ff884f8462fe9b082c0b977fb364ab834b44c34e448a86dfef3"}
{"chunkId":"cb-cognitive-bias-anchoring-effect-evidence","canonicalId":"cognitive-bias-anchoring-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-anchoring-effect/","resourceType":"concept","section":"evidence","title":"Anchoring Effect – when a starting number pulls later estimates toward it","text":"Anchoring Effect – when a starting number pulls later estimates toward it. Evidence: well-supported for numerical judgments; strength depends on anchor type and context. Anchoring is a well-supported effect in which an initial numerical value can pull a later estimate toward it. A 2026 meta-analysis covering 2,601 effect sizes found a large overall effect, but also substantial variation across studies. The effect should not be treated as a rule that every number changes every judgment: incidental anchors, anchors from a different dimension, clearly random values, incentives, and some debiasing conditions were associated with smaller or null effects. Mechanism: The original account described people as starting from an anchor and adjusting too little. Later work shows that anchoring is not one simple process. Experiments support insufficient adjustment especially for some self-generated anchors, while externally provided anchors can operate through other judgment processes. The safest summary is therefore behavioral: the starting value can systematically shift the final estimate, while the exact mechanism depends on how the anchor was produced and used. Practical check: For an important numerical judgment, make an independent estimate before seeing a suggested value when possible. If an anchor is already present, compare against base rates, completed cases, ranges, or several independent estimates rather than adjusting only from that one number. Ask what evidence would justify moving far away from the starting point. Awareness alone is not a guaranteed fix, and a mitigation should be judged by whether it actually improves calibration in the task being used.","reviewState":"reviewed","sourceIds":["src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"605386170ab9970f57edd8ea139b0062346ca5e5ec973fa78152a65365365ff8"}
{"chunkId":"cb-cognitive-bias-confirmation-bias-evidence","canonicalId":"cognitive-bias-confirmation-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-confirmation-bias/","resourceType":"concept","section":"evidence","title":"Confirmation Bias – when a belief shapes how evidence is searched and judged","text":"Confirmation Bias – when a belief shapes how evidence is searched and judged. Evidence: well established, broad construct. Confirmation bias is an umbrella label for several ways existing beliefs or hypotheses can influence information search and interpretation. It should not be reduced to one behaviour such as reading only agreeable news, and a preference for confirming tests is not irrational in every task or environment. Mechanism: Classic hypothesis-testing work showed that people can seek cases that fit a current hypothesis instead of searching efficiently for cases that could disconfirm it. Broader reviews describe related tendencies in information search, evidence interpretation, memory, and hypothesis testing, with different mechanisms likely contributing across contexts. Practical check: For an important belief, write down what evidence would count against it before looking for more information. Search deliberately for diagnostic counterexamples and compare competing explanations using the same evidence standard. The goal is not to consume an arbitrary amount of opposing content, but to make falsification possible.","reviewState":"reviewed","sourceIds":["src-83375467c08e8177","src-c4e8734b8f1248ee"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"38ac836174df96449bfbb8a44682a4077e4bc8ba956c5f29fa8d71dc24dae997"}
{"chunkId":"cb-cognitive-bias-curse-of-knowledge-evidence","canonicalId":"cognitive-bias-curse-of-knowledge","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-curse-of-knowledge/","resourceType":"concept","section":"evidence","title":"Curse of Knowledge – when what you know distorts what you expect others to know","text":"Curse of Knowledge – when what you know distorts what you expect others to know. Evidence: supported across several perspective-taking tasks. Knowing more can make it harder to estimate what a less-informed person knows or understands, but the size and mechanism of the effect depend on the task. The useful claim is not that experts are unable to teach beginners; it is that one's own knowledge can contaminate judgments about another person's knowledge unless the perspective gap is made explicit. Mechanism: Research has found knowledge-based distortions in economic judgments and interpersonal perspective taking. Proposed mechanisms include anchoring on one's own knowledge, difficulty suppressing known information, missing diagnostic cues about the other person's knowledge, and misattributing the fluency of familiar information. No single mechanism explains every version of the effect. Practical check: When explaining something to a novice, do not estimate clarity from how easy the idea feels to you. Ask the other person to predict, paraphrase, or perform the next step, and use their errors as feedback. Narrative feedback about the other person's perspective can reduce later egocentric projection more than a simple accuracy score.","reviewState":"reviewed","sourceIds":["src-7055432ee3ef6b7f","src-51a2a4c04f206e85","src-945106dd686dc96f"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"143da3821bb887d1b25effcdf7f9bfec4b1ecdc9d3bfda98b79da68ea8eef1b2"}
{"chunkId":"cb-cognitive-bias-declinism-evidence","canonicalId":"cognitive-bias-declinism","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-declinism/","resourceType":"concept","section":"evidence","title":"Declinism – when the present is judged against an idealized past","text":"Declinism – when the present is judged against an idealized past. Evidence: umbrella label; direct evidence supports specific decline illusions. Declinism is best treated here as an umbrella label for judging the present as worse than an idealized past, not as one standardized cognitive-bias construct with a single mechanism. Research directly supports several narrower ingredients: people can remember past experiences more positively than they experienced them, negative affect associated with autobiographical memories often fades faster than positive affect, and large multi-study work finds a pervasive illusion of moral decline. None of this means that every claim of social, technological, institutional, or personal decline is false; real decline must be tested against domain-specific evidence. Mechanism: Different decline judgments can arise from different processes. Rosy retrospection can make remembered experiences more positive than they were in the moment. Fading affect can reduce the emotional weight of negative memories faster than positive ones. In perceptions of moral decline, biased exposure to negative information about the present combined with biased memory for the past can create an apparent downward trend. Because these mechanisms differ by domain, 'declinism' should not be used as a universal explanation whenever someone thinks conditions have worsened. Practical check: Specify the domain and time window before judging decline. Compare the present with contemporaneous records or repeated measurements from the past rather than memory alone. Separate a real trend in the target variable from nostalgia for a broader period, and look for indicators that could show improvement as well as deterioration. If the claim is about morality or social behavior, distinguish perceptions of 'people in general' from direct observations of people you actually know.","reviewState":"reviewed","sourceIds":["src-c375f114daf625f6","src-c378be567ea78220","src-f3b090aa4f2bf959"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"7001988668c3452d78f55ccdd05f98c2b4928c5bc352e4ddcb1416c4ed3e4c2a"}
{"chunkId":"cb-cognitive-bias-hindsight-bias-evidence","canonicalId":"cognitive-bias-hindsight-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-hindsight-bias/","resourceType":"concept","section":"evidence","title":"Hindsight Bias – when the past looks obvious in retrospect","text":"Hindsight Bias – when the past looks obvious in retrospect. Evidence: robust. Knowing an outcome can make the outcome look more predictable in retrospect. The effect has been studied for decades and across many settings, but it does not mean that every confident explanation after an event is biased. Mechanism: Outcome knowledge changes how people reconstruct what was knowable before the event. Once the result is known, information that fits it can feel more relevant and the original uncertainty becomes harder to recreate. Practical check: Keep forecasts, assumptions, and confidence estimates before important outcomes are known. When reviewing a decision later, compare the result with that earlier record instead of reconstructing your prediction from memory. Simply warning yourself about hindsight bias is unlikely to be enough.","reviewState":"reviewed","sourceIds":["src-6df53f947e2583fa","src-75d0eb4cf8b10c07"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"ab6fb67d4001a7dd464c91876d82d6f89d46f1b8f49126880326ea894d592891"}
{"chunkId":"cb-cognitive-bias-hungry-judge-effect-evidence","canonicalId":"cognitive-bias-hungry-judge-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-hungry-judge-effect/","resourceType":"concept","section":"evidence","title":"Hungry Judge Effect – a famous parole-board finding with an uncertain cause","text":"Hungry Judge Effect – a famous parole-board finding with an uncertain cause. Evidence: contested observational finding; hunger interpretation not established. The famous 2011 parole-board study found favorable decisions were more common early in a decision session and immediately after food breaks. It did not directly measure hunger, glucose, fatigue, or mood, and the authors explicitly could not separate eating from rest. A contemporaneous critique argued that non-random case ordering could explain the pattern; the original authors disputed that critique with additional analyses. Treating this as a general law that hungry judges become harsher goes beyond the evidence. Mechanism: The original paper interpreted the sequence pattern as consistent with resource depletion and recovery after a break, but its design was observational. Alternative explanations include case ordering and representation differences. Because food, rest, time, sequence position, and other factors were entangled, the study cannot identify hunger itself as the causal mechanism. Practical check: Do not use this effect as evidence that a decision-maker's hunger caused a specific ruling. The safer operational lesson is broader: high-stakes repeated decisions deserve checks for order effects, fatigue, break structure, workload, and case assignment. If you want to test a hunger hypothesis, measure hunger separately rather than using meal timing as a substitute.","reviewState":"reviewed","sourceIds":["src-60924ea4aa8ec622","src-90615805724bf091","src-de25d398141bd9bc"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"89b9ca480fb1317cd1db8f4b23b4ea0f19a5c3da429cdc5ca9a5fd0b6f668cdc"}
{"chunkId":"cb-cognitive-bias-impact-bias-evidence","canonicalId":"cognitive-bias-impact-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-impact-bias/","resourceType":"concept","section":"evidence","title":"Impact Bias – when you overestimate how long or intense your emotions will be","text":"Impact Bias – when you overestimate how long or intense your emotions will be. Evidence: well supported, with important forecasting nuances. People often overestimate how intense or long their emotional reactions to future events will be, especially when the focal event crowds out everything else that will also shape daily experience. The literature is broader than a rule that people always overpredict emotion: forecasting errors vary by event, time horizon, emotion, and what exactly is being predicted. Mechanism: Affective-forecasting research identifies several contributors. Focalism makes the target event dominate imagination while ordinary future events receive too little weight. For negative events, people can also underpredict adaptation and the psychological processes that help them make sense of what happened. Available memories can further bias forecasts when memorable examples are atypical. Practical check: For a major future event, forecast both the focal event and the rest of an ordinary week around it. Estimate intensity and duration separately, then compare later forecasts with actual experience. When possible, use experience-sampling or a reference class of people who have already lived through a similar event instead of relying only on vivid imagination.","reviewState":"reviewed","sourceIds":["src-ebc67fd4db360d62","src-ea6429b27b8d6ccf","src-bf639003c08f7257","src-902102ef57fb68e5"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"cc6f552fc29303ddd850cdc099f85530bd4932400682ad47990b0ffead034d03"}
{"chunkId":"cb-cognitive-bias-outcome-bias-evidence","canonicalId":"cognitive-bias-outcome-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-outcome-bias/","resourceType":"concept","section":"evidence","title":"Outcome Bias – when you judge a decision by its result, not its quality","text":"Outcome Bias – when you judge a decision by its result, not its quality. Evidence: replicated. Outcome bias occurs when knowledge of a result changes how people evaluate the quality of a decision even when the information available at the time of the decision is held constant. Outcomes can still be relevant for learning, so the error is not 'never look at results'; it is using luck or hindsight as if it had been available to the original decision-maker. Mechanism: The classic experiments varied outcomes while holding the decision information constant and found more favorable evaluations after good outcomes. A large preregistered replication reproduced the direction of the effect in a medical-decision scenario, including among participants who said outcomes should not influence the evaluation. Practical check: Review the process twice: first judge the decision using only information that was available when it was made, then use the outcome to update models and future assumptions. Keeping these two questions separate helps distinguish decision quality from luck without throwing away useful feedback.","reviewState":"reviewed","sourceIds":["src-f1f0400ef362aaa8","src-1801b1347f89ef91"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"ab9fdfbfa6586f6974641297e00d15327e44e0d1da09b7ba7f2e023df68ac78d"}
{"chunkId":"cb-cognitive-bias-prevention-bias-evidence","canonicalId":"cognitive-bias-prevention-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-prevention-bias/","resourceType":"concept","section":"evidence","title":"Prevention Bias – favoring preventive security controls over detection and response","text":"Prevention Bias – favoring preventive security controls over detection and response. Evidence: single-study / domain-specific. Prevention Bias is a label introduced in experimental research on information-security investment. In that task, participants favored preventive controls over detection-and-response controls even when the experiment was designed so the two classes had the same return on investment. The evidence does not establish a broad psychological law that people generally overvalue prevention in healthcare, insurance, relationships, or every other domain. Mechanism: The original information-security study interprets the pattern through behavioral decision-making mechanisms including reference points, loss aversion, affect, and fast intuitive processing. Those mechanisms are plausible background explanations, but the named Prevention Bias itself is a domain-specific construct from this research program rather than a broadly replicated general-purpose bias. Practical check: When allocating a security budget, compare prevention, detection, and response using the same expected-loss and marginal-value framework instead of assuming prevention is automatically the superior category. For other domains, do not import the label without evidence that the same allocation pattern and incentives actually apply.","reviewState":"reviewed","sourceIds":["src-eace2b628cde1bce","src-48160eef6b029066"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"800f0ff2b1f2d288c68917288d32d0eb8f8b2dc4ed327959b4676c2bc0370f4d"}
{"chunkId":"cb-cognitive-bias-projection-bias-evidence","canonicalId":"cognitive-bias-projection-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-projection-bias/","resourceType":"concept","section":"evidence","title":"Projection Bias – when you assume your future self will want what you want now","text":"Projection Bias – when you assume your future self will want what you want now. Evidence: established in intertemporal preference prediction. Projection bias describes a tendency to overproject current tastes or visceral states onto future preferences. People often understand that tastes will change but underestimate how much they will change. The construct is most directly supported in intertemporal choice and consumer settings; it should not be stretched into a generic explanation for every bad prediction about one's future self. Mechanism: Current states provide a salient reference point for forecasting future utility. When people incompletely adjust away from today's tastes, hunger, weather, or other transient conditions, future choices can be distorted. Field and experimental evidence shows that current conditions can influence decisions whose consequences occur later even when those conditions will not persist. Practical check: For decisions that bind your future self, record the current state that may be contaminating the forecast: hunger, fatigue, mood, weather, novelty, or current preferences. Re-evaluate the decision in a meaningfully different state or use past preference changes as a reference class before making an expensive or difficult-to-reverse commitment.","reviewState":"reviewed","sourceIds":["src-479735fa3da2d5b3","src-3828d4aac220e974","src-b0cc2953c2431e7b"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"26cd19a6f7841b43bfadd48a72d74abfec1d62285b04039af8e3bcb7783828ab"}
{"chunkId":"cb-cognitive-bias-sunk-cost-effect-evidence","canonicalId":"cognitive-bias-sunk-cost-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-sunk-cost-effect/","resourceType":"concept","section":"evidence","title":"Sunk Cost Effect – when past costs influence the next choice","text":"Sunk Cost Effect – when past costs influence the next choice. Evidence: well-supported overall; strength varies by decision type and context. The sunk cost effect is a documented tendency for irrecoverable prior investments of money, time, or effort to influence later choices. A meta-analytic review found clear evidence for the effect overall, while also showing that its size and moderators differ between utilization decisions and progress decisions. The effect should not be used to label every choice to continue: future value, switching costs, uncertainty, and how close a project is to useful completion can all be relevant to a forward-looking decision. Mechanism: Classic experiments linked sunk-cost behavior to reluctance to appear wasteful, while later work has connected it to several decision processes rather than one universal mechanism. In project settings, sunk costs can interact with the perceived need to complete the work. That interaction helps explain why sunk costs are one possible antecedent of escalation of commitment rather than a complete explanation for every escalation case. Practical check: For a current choice, put irrecoverable past costs in a separate line and evaluate the remaining options using costs and benefits from today forward. Compare continuing with the best realistic alternative use of the remaining money, time, and attention. If completion itself creates future value, include that value explicitly instead of treating 'we are already close' as either automatically rational or automatically biased.","reviewState":"reviewed","sourceIds":["src-25c242b954867ac1","src-812f426a52613922","src-ec03691286f7284e"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"b68e8e988807ffe3ccf7355bc17830291b99082c82056879c890bfc6ad99a99a"}
{"chunkId":"cb-cognitive-bias-surrogation-evidence","canonicalId":"cognitive-bias-surrogation","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-surrogation/","resourceType":"concept","section":"evidence","title":"Surrogation – when a performance measure starts replacing the goal","text":"Surrogation – when a performance measure starts replacing the goal. Evidence: supported in strategic performance-measure settings. Surrogation is a domain-specific management-accounting construct: a measure that was designed to represent a strategic objective can start to be treated as though it were the objective itself. It is closely related to metric fixation and proxy problems, but those broader labels should not be treated as exact synonyms without checking the setting and mechanism. Mechanism: Strategic performance systems translate hard-to-measure constructs into observable measures. Experiments show that managers can lose sight of the underlying construct and act as if the measure itself were what matters. The literature also provides boundary conditions and interventions: involvement in strategy selection and opportunities to explain decisions narratively can reduce surrogation in studied settings. Practical check: For every important KPI, write the construct it is meant to represent and at least one important dimension it does not capture. During reviews, discuss the goal before the metric and require a short explanation for major metric-driven decisions. If optimizing the number can make the underlying goal worse, the proxy needs an explicit guardrail or companion measure.","reviewState":"reviewed","sourceIds":["src-bc1fa4e94ab0fa71","src-dd6028af1211deb7"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"59761914d5e3b1c5bde39b49001decac40afa4be0394ba203987e6e260c8d195"}
{"chunkId":"cb-cognitive-bias-systematic-bias-evidence","canonicalId":"cognitive-bias-systematic-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-systematic-bias/","resourceType":"concept","section":"evidence","title":"Systematic Bias – a measurement and statistical concept, not one cognitive bias","text":"Systematic Bias – a measurement and statistical concept, not one cognitive bias. Evidence: measurement/statistical concept; not a standalone cognitive-bias construct. Systematic bias is a general statistical and measurement concept: a non-random tendency for measurements, estimates, or study results to deviate from a reference or target value in a consistent or predictable way. Standards in metrology define systematic measurement error separately from random error and define measurement bias as an estimate of that systematic error. This is not one specific psychological bias with one mechanism, so the page should not present it as a personal cognitive tendency alongside constructs such as confirmation bias or hindsight bias. Mechanism: Systematic error can arise from instruments, calibration, sampling, study design, confounding, data collection, analysis, or other structural features of a process. Unlike random variation, it does not reliably disappear just by collecting more observations. In some domains the word 'bias' also describes systematic tendencies in human judgment, but that broader usage does not turn systematic bias itself into a distinct cognitive-bias mechanism. Practical check: Define the target or reference value, then identify where a process could consistently push estimates in one direction. Use calibration, validation data, representative sampling, design controls, sensitivity analysis, or process changes to estimate and reduce the systematic component. Do not treat repeated error as proof of a psychological cause until measurement, sampling, and procedural explanations have been examined.","reviewState":"reviewed","sourceIds":["src-3b2666bc66255b98","src-1e7c10d323a7202c","src-c22f45c827d79fb6"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"e99c6b55fd83ebd7dbd53339511864a98c75779435659c38856b6b50aa8ee445"}
{"chunkId":"cb-cognitive-bias-value-selection-bias-evidence","canonicalId":"cognitive-bias-value-selection-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/cognitive-bias-value-selection-bias/","resourceType":"concept","section":"evidence","title":"Value Selection Bias – choosing supplied numbers instead of constructing the Bayesian answer","text":"Value Selection Bias – choosing supplied numbers instead of constructing the Bayesian answer. Evidence: supported in Bayesian reasoning tasks; limited independent replication. Value Selection Bias is a narrow reasoning pattern reported in Bayesian diagnostic problems: when participants did not derive the correct solution, incorrect answers were strongly drawn toward numerical values already present in the problem. The evidence supports this behavior across multiple problem variations in the authors' research program, but it should not be generalized to any situation where a person uses an old price, stale KPI, or convenient number. Mechanism: The research links value selection to reference dependence and problem representation. When reasoners are uncertain about how to construct the needed calculation, salient values supplied by the problem can look like candidate answers. Numerical ability and the way the reference class is represented affect accuracy, so the pattern is more specific than a generic preference for available numbers. Practical check: In a probability or diagnostic problem, state the target quantity and reference class before looking for a number to copy into the answer. Write the required relationship or calculation explicitly, then check whether the selected values actually belong to that relationship. If no calculation is needed, be able to explain why rather than assuming that a supplied number must be the answer.","reviewState":"reviewed","sourceIds":["src-50e334f70bc2a080","src-6dfe6272124fdec6"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"b82d55d35361f714ab551a10327b8fe4a1ef3d24b17fd9f483e285a4b744a3b7"}
{"chunkId":"cb-confirmation-bias-backfire-effect-evidence","canonicalId":"confirmation-bias-backfire-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/confirmation-bias-backfire-effect/","resourceType":"concept","section":"evidence","title":"Backfire Effect – when a correction can strengthen a false belief","text":"Backfire Effect – when a correction can strengthen a false belief. Evidence: mixed / conditional. Corrections usually improve factual accuracy. A true backfire effect, where a correction makes the targeted false belief stronger, appears to be uncommon and sensitive to context and measurement. It should not be treated as the default response to being corrected. Mechanism: The label describes a change in belief after corrective information, not ordinary disagreement. Identity, prior beliefs, source trust, and how the correction is measured can affect responses, but the broad claim that facts normally make false beliefs stronger is not supported by the current evidence base. Practical check: Do not avoid corrections because you fear an automatic backfire. Give the correction clearly, connect it to reliable evidence, and separate belief accuracy from whether the correction changes a person's wider attitudes or behaviour. For important decisions, check the belief again later because correction effects can fade.","reviewState":"reviewed","sourceIds":["src-e1813cd3b9123eaf","src-40282584441d78a7","src-4f99e4ca8c1af4fe"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"90f4cd9238f0dc2b07361a3b3e200aab5b3649dacb73066e5b4c03924bd1787a"}
{"chunkId":"cb-confirmation-bias-congruence-bias-evidence","canonicalId":"confirmation-bias-congruence-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/confirmation-bias-congruence-bias/","resourceType":"concept","section":"evidence","title":"Congruence Bias – when the test fits one hypothesis but does not separate alternatives","text":"Congruence Bias – when the test fits one hypothesis but does not separate alternatives. Evidence: established in hypothesis-testing tasks; narrower than confirmation bias. Congruence Bias is a narrower hypothesis-testing pattern than the broad umbrella of confirmation bias. In classic work, people overvalued tests that were likely to return a positive result if their leading hypothesis were true, even when other tests were more diagnostic among competing hypotheses. The useful claim is not that every positive test is irrational: a positive test strategy can be efficient in some environments, and the problem depends on whether the chosen test can actually distinguish the focal hypothesis from alternatives. Mechanism: Baron, Beattie, and Hershey described congruence bias as overvaluing tests whose expected positive result matches the currently favored hypothesis. Klayman and Ha showed why the closely related positive test strategy should not automatically be treated as a logical error: sampling cases expected under a hypothesis can be informative depending on the structure of the environment. The practical failure occurs when a test is congruent with the focal hypothesis but has low diagnostic value because alternative hypotheses predict the same result. Practical check: Before choosing a test, list at least one plausible alternative hypothesis and predict the result under both explanations. Prefer observations whose possible outcomes separate the hypotheses rather than merely giving your preferred explanation another chance to say 'yes.' If the same result is expected under several explanations, it is weak evidence even when it matches the preferred hypothesis.","reviewState":"reviewed","sourceIds":["src-54ea5126c3035a7e","src-f3a6c8d07f4870ec"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"52d0c2172bab8f75006e38e21562cd99d8628f48ed728529c37fa6073db603db"}
{"chunkId":"cb-egocentric-bias-planning-fallacy-evidence","canonicalId":"egocentric-bias-planning-fallacy","canonicalUrl":"https://cognitive-biases.github.io/biases/egocentric-bias-planning-fallacy/","resourceType":"concept","section":"evidence","title":"Planning Fallacy – when plans make completion times look too optimistic","text":"Planning Fallacy – when plans make completion times look too optimistic. Evidence: well-supported for time estimates; size and causes vary by context. The planning fallacy is a well-documented tendency for people to predict their own task completion times too optimistically. The effect has been observed across different kinds of tasks, but it is not a rule that every plan will run late. Project overruns can also come from changing scope, dependencies, incentives, poor data, deliberate underestimation, or genuinely unusual events, so a late project should not automatically be diagnosed as a planning fallacy. Mechanism: Classic studies found that people often build a detailed scenario for how the current task will unfold while giving too little weight to how long similar tasks actually took. Later reviews describe cognitive, motivational, and social influences rather than one single cause. Recent cross-cultural experiments also suggest that reminders about past experience and the social consequences of delay can change time predictions differently across groups, which is another reason not to treat one mechanism or correction as universal. Practical check: For an important estimate, start with outcomes from comparable completed work before refining the current plan. Give a range rather than a single precise date, name the main uncertainties, and keep the original estimate so forecast error can be measured later. Reference-class methods can be useful when comparable data exist, but their quality depends on choosing a genuinely relevant class. There is no evidence-based universal rule that every estimate should simply be multiplied by the same fixed buffer.","reviewState":"reviewed","sourceIds":["src-7d92150bd312c109","src-cf328825be4a0674","src-d6ecdecbd0ffae61","src-d98817dbf800d030"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"7242bf429d279b8cf02b2a0b809e0d8ca192e41a781a099f13401a80023e4d1a"}
{"chunkId":"cb-false-priors-automation-bias-evidence","canonicalId":"false-priors-automation-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/false-priors-automation-bias/","resourceType":"concept","section":"evidence","title":"Automation Bias – when automated advice replaces independent checking","text":"Automation Bias – when automated advice replaces independent checking. Evidence: established, context-dependent. Automation bias is a documented pattern of inappropriate reliance on automated cues or recommendations. It can produce commission errors when a user follows incorrect advice and omission errors when a user fails to act because automation did not signal a problem. This does not mean automation is generally harmful: decision support can improve overall performance, and the relevant question is whether reliance remains calibrated when the system is wrong, incomplete, or difficult to verify. Mechanism: Research describes automation as a potential heuristic substitute for vigilant information search and processing. Reliance is shaped by trust calibration, attention allocation, workload, task complexity, time pressure, experience, and how difficult it is to verify the automated recommendation. Automation bias overlaps with automation-related complacency but the constructs are not identical. Practical check: For high-stakes decisions, make a provisional assessment or explicit decision criteria before viewing automated advice when practical. Verify consequential recommendations against independent cues or source data, especially when the system is outside its strongest domain. Design the workflow to reduce verification effort with source links, decomposed checks, calibrated uncertainty, and explicit review responsibility rather than relying on generic warnings to 'be careful'.","reviewState":"reviewed","sourceIds":["src-0252755d114dfb6c","src-f8b4657a59aad7de","src-965010b0cf7cde95","src-d8b0db0a156ad3a7"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"ffa0f6f0a16db7296d62833824b9e5e168d05560e2e55f44560d2b34c660965c"}
{"chunkId":"cb-framing-effect-core-evidence","canonicalId":"framing-effect-core","canonicalUrl":"https://cognitive-biases.github.io/biases/framing-effect-core/","resourceType":"concept","section":"evidence","title":"Framing Effect – when equivalent descriptions lead to different choices","text":"Framing Effect – when equivalent descriptions lead to different choices. Evidence: risky-choice framing is robust; broader framing types have different evidence. Framing research covers several distinct paradigms. Risky-choice framing, where equivalent outcomes are presented as gains or losses, is a well-established effect with substantial evidence across many studies. A classic meta-analysis found a reliable small-to-moderate average effect with strong variation across designs, and later reviews and metastudies support broad generalizability. Attribute framing also has supporting evidence, while goal or message framing is less consistently established. These should not be treated as one identical effect caused by any change in wording. Mechanism: No single mechanism explains every framing paradigm. In risky choice, reference points, the valence or gist of descriptions, and how outcomes are represented can all affect preference. The 1998 framing typology explicitly separated risky-choice, attribute, and goal framing because they use different tasks and may rely on different processes. The most defensible behavioral claim is that descriptions that preserve much of the underlying decision information can still change judgments or preferences, while the size and direction depend on the task and representation. Practical check: For an important choice, restate the options using neutral quantities and test equivalent gain and loss descriptions while holding outcomes and probabilities constant. Show complete information when practical, and separate the effect of the frame from missing or mismatched information. If the decision changes, investigate why before assuming one description is the objective truth. Treat this as a consistency check, not a rule that positive or negative wording is always manipulative.","reviewState":"reviewed","sourceIds":["src-bb0fa40a53281f49","src-0e441147fc835a3d","src-47b281f980645234","src-2e83fc60350e0524","src-dc40b5ceda9cbea1"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"9e62962928a6b0fd99ff6b4959072fa99865f41228d290be55c5e01b5fc319b8"}
{"chunkId":"cb-framing-effect-decoy-effect-evidence","canonicalId":"framing-effect-decoy-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/framing-effect-decoy-effect/","resourceType":"concept","section":"evidence","title":"Decoy Effect – when an inferior option changes the comparison between better options","text":"Decoy Effect – when an inferior option changes the comparison between better options. Evidence: well documented, context-dependent. The decoy effect is a well-documented family of context effects in which adding an inferior or otherwise strategically positioned alternative can change choice shares among the original options. The classic attraction effect uses an asymmetrically dominated decoy that is worse than the target but not necessarily worse than the competitor. The effect is not guaranteed: its size depends on the geometry of the options, prior preferences, task design and other moderators, and preregistered replications have found both successful and weak or null results. Mechanism: Research offers several explanations, including contrast between attribute trade-offs, changes in relative value, ease of justification and attention to particular dimensions. A 2024 integrative review concludes that there is still no clear consensus on one underlying mechanism and separates attraction, compromise and phantom-decoy effects rather than treating every extra option as the same phenomenon. Practical check: When a third option seems to make one plan suddenly look obvious, compare the original alternatives without it and score the options on the attributes that matter to the decision. Check whether the suspected decoy is actually dominated by the target and whether removing it changes the preference between the remaining options. For product or pricing design, test the full choice set rather than assuming that a three-tier menu will reliably create the effect.","reviewState":"reviewed","sourceIds":["src-84a20f469659275d","src-6f220811f3f7fc69","src-fe432751bebf733b","src-37cf79ece99ea6c9"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"0fb4cf88b05f04417309ea112fd34c8eec9a58b0d1d60ca5593f7ff24dbb0a21"}
{"chunkId":"cb-framing-effect-default-effect-evidence","canonicalId":"framing-effect-default-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/framing-effect-default-effect/","resourceType":"concept","section":"evidence","title":"Default Effect – when a pre-selected option changes what people choose","text":"Default Effect – when a pre-selected option changes what people choose. Evidence: well supported, highly context-dependent. Pre-selecting an option often increases the chance that people choose it, but the size of the default effect varies greatly across domains and designs. A 2019 meta-analysis of 58 studies found a substantial average effect together with strong heterogeneity, including some studies with little or even reversed effects. Staying with a default should therefore not be treated as a direct measure of strong preference or as proof that the default improved welfare. Mechanism: There is no single mechanism behind every default effect. Research points to several routes: changing the option can require effort or attention; a default can signal what the choice architect recommends; and a pre-selected option can feel like the existing or endowed state. These routes can overlap, and their importance changes with the decision and the interface. Practical check: Make consequential defaults visible, easy to change and easy to understand. Compare default designs with active choice when practical, and measure whether the resulting choice still serves the user's goals rather than inferring preference from uptake alone. Treat a default as a design decision that needs testing, not as a universal behavioural lever.","reviewState":"reviewed","sourceIds":["src-7d9d96710fab2c5f","src-64c4a8072d35fe53","src-90e516c044f0e914","src-5e28781720bc4f35"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"3e11ceba17f3011631a1ddb009224233f0876ed5a2ef25dee8eb094182880542"}
{"chunkId":"cb-heuristic-bias-availability-bias-evidence","canonicalId":"heuristic-bias-availability-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/heuristic-bias-availability-bias/","resourceType":"concept","section":"evidence","title":"Availability Heuristic – when easy-to-recall examples shape frequency judgments","text":"Availability Heuristic – when easy-to-recall examples shape frequency judgments. Evidence: established heuristic; bias is context-dependent. Availability is a judgment heuristic: people can use how easily examples or scenarios come to mind when estimating frequency or probability. That shortcut is not automatically an error because memorable or accessible examples can correlate with real frequency. Bias appears when accessibility is driven by factors that are not diagnostic of the quantity being judged. Mechanism: The original account focused on the availability of relevant instances and scenarios. Later work examined subjective ease of retrieval as an informational cue, showing that judgments can depend on how easy recall feels as well as on what is recalled. Some later replication attempts in specific contexts have not reproduced every ease-of-retrieval effect, so one narrow mechanism should not be treated as the entire heuristic. Practical check: When vivid or recent examples are driving a risk estimate, compare recall with base-rate data or a deliberately sampled reference set. Ask whether the examples are easy to remember because they are common or because they are dramatic, recent, repeated, or personally salient.","reviewState":"reviewed","sourceIds":["src-723760d16e30b39e","src-f612390f3949191b","src-0ba7057e7c37fb49"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"ec7b462f0b2c8cc4c5654c6703f40148d01e4b8300733262fdbd996cc7d53fda"}
{"chunkId":"cb-human-robot-interaction-form-evidence","canonicalId":"human-robot-interaction-form","canonicalUrl":"https://cognitive-biases.github.io/biases/human-robot-interaction-form/","resourceType":"concept","section":"evidence","title":"Appearance–Capability Expectation – when a robot’s design changes what you expect it can do","text":"Appearance–Capability Expectation – when a robot’s design changes what you expect it can do. Evidence: supported HRI pattern; project label is nonstandard. Research supports the broader pattern that a robot's appearance, morphology, framing, and human-likeness shape expectations about its competence, social qualities, and likely behavior. However, 'Form-Function Attribution Bias' is not an established standardized name in the literature. On this site it should be treated as a project label for appearance-driven capability expectations, not as a universally recognized cognitive-bias construct. Mechanism: Visual and social design cues help users build a mental model before they have much direct performance evidence. Humanlike or technically styled forms can set different expectations, and expectation-setting can change later competence or trust judgments. The direction is not universal: a more anthropomorphic appearance does not always increase perceived competence, empathy, or trust, and effects depend on task and context. Practical check: Separate what the system looks or sounds like from what it has demonstrated it can do. For an AI or robot, write down the capabilities implied by the interface and then check each against observed performance, documentation, or controlled tests. Designers should communicate limitations and intended functions directly instead of letting appearance carry the whole capability message.","reviewState":"reviewed","sourceIds":["src-e3698f659e09ea7e","src-394b04c1e1398c8f","src-09bef5bf4eda8006","src-b4d5205770653dfc"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"dd3243032ed7f00ad7fa3b5d6e011531db55156cbde1ba55e6351ed6e100af91"}
{"chunkId":"cb-logical-fallacy-escalation-of-commitment-evidence","canonicalId":"logical-fallacy-escalation-of-commitment","canonicalUrl":"https://cognitive-biases.github.io/biases/logical-fallacy-escalation-of-commitment/","resourceType":"concept","section":"evidence","title":"Escalation of Commitment – when setbacks lead to more investment in the same course","text":"Escalation of Commitment – when setbacks lead to more investment in the same course. Evidence: established, but related constructs should be separated. Sunk-cost effects and escalation of commitment overlap but are not interchangeable. Sunk-cost research asks whether irrecoverable prior investments influence current choices. Escalation of commitment describes persistence or additional resource allocation to a failing course of action and can also be driven by personal responsibility, self-justification, project structure, and other factors. Mechanism: Experiments and field studies show that prior investments can increase continuation or utilization, while classic escalation studies found stronger additional investment after negative outcomes when decision-makers were personally responsible for the original choice. Meta-analytic work finds a sunk-cost effect overall but also substantial moderation by decision type and context. Practical check: Separate already-spent resources from the expected future costs and benefits of the next decision. For projects, define exit criteria before bad news arrives and ask an independent reviewer to assess the next increment of investment. Do not assume that every continued project is irrational: new information about future value can justify continuation even when past costs are sunk.","reviewState":"reviewed","sourceIds":["src-644614d8919a2ee4","src-25c242b954867ac1","src-ec03691286f7284e"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"25a05bc81bb5e66bb7ae46ee41154edb365e31426053cce40730b15a94c743d1"}
{"chunkId":"cb-memory-bias-continued-influence-effect-evidence","canonicalId":"memory-bias-continued-influence-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/memory-bias-continued-influence-effect/","resourceType":"concept","section":"evidence","title":"Continued Influence Effect – when corrected information still shapes reasoning","text":"Continued Influence Effect – when corrected information still shapes reasoning. Evidence: well-established persistence after correction; corrections usually still help. The continued influence effect describes a specific pattern: false or outdated information can keep affecting later reasoning even after a clear correction. It does not mean the correction had no effect, that the person rejected the correction, or that belief in the false claim became stronger. Corrections often reduce misinformation reliance without removing it completely. That distinction matters because continued influence is much broader than the rarer backfire outcome. Mechanism: Research does not support one single explanation for every continued-influence result. Accounts include difficulty updating a mental model once misinformation has filled an explanatory role, retrieval competition between the original information and the correction, familiarity, source memory, and changes in how the original source is judged. Recent reviews therefore treat continued influence as a family of updating and retrieval problems rather than one simple memory failure. Practical check: Correct important misinformation clearly rather than leaving it unchallenged. When the false claim explained an event, provide a credible alternative explanation when one is available. Detailed corrections and repetition can help in many settings, and current evidence gives little reason to avoid a correction simply because it repeats the false claim. Keep the corrected information easy to find and measure the same belief or inference again when the distinction matters. Do not label disagreement, distrust, or incomplete updating as a backfire effect unless the targeted false belief actually becomes stronger.","reviewState":"reviewed","sourceIds":["src-9811dba9926395b6","src-6891b8bf627a6ec7","src-422189df9f5faa26","src-80cc3fc9e30c3c69","src-4f99e4ca8c1af4fe"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"66857301dd54cfdb85548743b0009afdafb0385286ce76e9ffcbd2cee301fb83"}
{"chunkId":"cb-memory-bias-fading-affect-bias-evidence","canonicalId":"memory-bias-fading-affect-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/memory-bias-fading-affect-bias/","resourceType":"concept","section":"evidence","title":"Fading Affect Bias – when negative emotional intensity tends to fade faster","text":"Fading Affect Bias – when negative emotional intensity tends to fade faster. Evidence: supported across autobiographical-memory research, with moderators. Fading Affect Bias describes an average asymmetry in autobiographical memory: affect linked to negative events tends to weaken faster over time than affect linked to positive events. It is a group-level pattern with important moderators, not a rule that every painful memory becomes mild or positive. Recent large-sample work also finds the pattern when people directly report how the affect of their memories changed, rather than relying only on calculated differences between retrospective ratings. Mechanism: The literature describes the pattern rather than one single settled mechanism. Social rehearsal, emotion regulation, meaning making, memory reconstruction, event characteristics, and individual differences can all matter. The useful distinction is that Fading Affect Bias concerns change in emotional intensity associated with autobiographical memories, not a general claim that factual details of the past become more positive. Practical check: When comparing how a past event felt with how a similar event feels now, separate remembered facts from remembered emotional intensity. If the comparison matters, use contemporaneous records or repeated ratings instead of assuming today's emotional reaction is an unchanged copy of the original experience. Do not use Fading Affect Bias to dismiss persistent negative memories or individual experiences that do not follow the average pattern.","reviewState":"reviewed","sourceIds":["src-52af9360ef62152d","src-816c78f1bdb1f366"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"33846618cd1e44e3a247a025700d00d8f50e361b8887a9c1fbe2d7ca2e7cb4f6"}
{"chunkId":"cb-memory-bias-rosy-retrospection-evidence","canonicalId":"memory-bias-rosy-retrospection","canonicalUrl":"https://cognitive-biases.github.io/biases/memory-bias-rosy-retrospection/","resourceType":"concept","section":"evidence","title":"Rosy Retrospection – when a past experience is remembered more positively","text":"Rosy Retrospection – when a past experience is remembered more positively. Evidence: supported in event-recollection studies; narrower than general nostalgia. Rosy Retrospection describes a pattern in which later evaluations of a past experience can become more positive than evaluations recorded during the experience itself. The classic work studied specific meaningful events and found more positive anticipation and later recollection than in-the-moment experience. This does not mean that every fond memory is false, that all past periods were worse than remembered, or that nostalgia is itself a bias. Mechanism: In the foundational studies, negative thoughts increased during the event because of distractions, disappointment, and less positive self-evaluation, but these influences were short-lived. Later positive recollection can also interact with memory accessibility and interpretation. The evidence is firmer for the observed shift in evaluations than for one universal mechanism that explains every positive memory of the past. Practical check: When a remembered event is being used as evidence for a current decision, compare the later memory with records made during the event when those records exist: notes, ratings, messages, photos with context, costs, incidents, or repeated measurements. Treat a positive memory as meaningful experience, but do not automatically treat it as an accurate historical average of how the event felt at the time.","reviewState":"reviewed","sourceIds":["src-c375f114daf625f6","src-c378be567ea78220"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"8e7f4e98d33255af87de76bb53373705f6200b6b9822f1d72ca4884e5660fd28"}
{"chunkId":"cb-probability-bias-subadditivity-effect-evidence","canonicalId":"probability-bias-subadditivity-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/probability-bias-subadditivity-effect/","resourceType":"concept","section":"evidence","title":"Subadditivity Effect – the whole seems less likely than the sum of its parts","text":"Subadditivity Effect – the whole seems less likely than the sum of its parts. Evidence: established with boundary conditions. People often give a larger total probability when an event is unpacked into separate possibilities than when the same event is judged as one packed category. The effect is not universal: how the possibilities are described and how typical they are can change or even reverse an unpacking effect. Mechanism: Support theory models probability judgments as depending partly on the support that comes to mind for the focal event and its alternatives. Unpacking can make component possibilities more available or easier to evaluate, increasing their combined judged probability. Practical check: For consequential forecasts, compare the packed event with a mutually exclusive, collectively exhaustive breakdown. If the estimates change greatly when the wording changes, investigate the representation before acting on the numbers. The goal is consistency, not forcing every judgment to match one decomposition.","reviewState":"reviewed","sourceIds":["src-e5eecafb672bd8ed","src-9f9930db613528b2","src-ffae461551a5d8e7"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"aa00ed8c850ba843837ba802d217b4bf733b976435c3bf3661c263fd8eb35f02"}
{"chunkId":"cb-prospect-theory-loss-aversion-evidence","canonicalId":"prospect-theory-loss-aversion","canonicalUrl":"https://cognitive-biases.github.io/biases/prospect-theory-loss-aversion/","resourceType":"concept","section":"evidence","title":"Loss Aversion – when a loss can carry more weight than an equivalent gain","text":"Loss Aversion – when a loss can carry more weight than an equivalent gain. Evidence: influential and widely estimated; magnitude and robustness debated. Loss aversion is a central component of prospect theory and many studies estimate losses as receiving greater subjective weight than gains around a reference point. However, the effect should not be summarized with one universal coefficient. A 2024 interdisciplinary meta-analysis of 607 estimates reported a mean coefficient near 1.96, while another 2024 meta-analysis of individual risky-choice datasets estimated about 1.31. A 2025 re-analysis of the larger dataset found little evidence of loss aversion in some symmetric, unordered designs, showing that task structure and analysis can materially change the result. Mechanism: Prospect theory represents outcomes as gains and losses relative to a reference point and allows the value function to be steeper for losses than for gains. That model is useful descriptively, but observed loss-sensitive behaviour can also be affected by probability weighting, payoff ordering, attention, reference-point construction and other features of the task. The current debate is therefore about both magnitude and what experimental patterns should count as clean evidence for loss aversion. Practical check: For a consequential choice, state the reference point and compare matched gains and losses using the same probabilities and time horizon. Re-express options as absolute outcomes when possible and check whether the preference survives. Do not import a fixed 'losses count twice' coefficient into a new setting without evidence from that setting, and keep loss aversion separate from sunk-cost reasoning about resources that are already irrecoverable.","reviewState":"reviewed","sourceIds":["src-48160eef6b029066","src-e8a194eeb0017a8a","src-368a6a9d36d10612","src-72b6f021c45204d9"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"179efd2d18dab5e53984b1b8112bed6da430db76af44d46decba87afd8d5d196"}
{"chunkId":"cb-prospect-theory-status-quo-bias-evidence","canonicalId":"prospect-theory-status-quo-bias","canonicalUrl":"https://cognitive-biases.github.io/biases/prospect-theory-status-quo-bias/","resourceType":"concept","section":"evidence","title":"Status Quo Bias – when the current option gets extra weight because it is already in place","text":"Status Quo Bias – when the current option gets extra weight because it is already in place. Evidence: well established, broader than interface defaults. Status quo bias describes extra preference for an option because it is the current or existing state. Classic experiments and field observations found disproportionate persistence with status quo options, but keeping the current option is not automatically a bias. Switching can have real financial, practical, learning or uncertainty costs that make staying reasonable. Mechanism: Several explanations can contribute to status quo persistence, including reference dependence and loss aversion, attention, uncertainty, cognitive effort and psychological commitment to what is already in place. Experimental work also shows that a status quo can shape which alternatives receive attention. This is broader than a designed default: a current provider, workflow or possession can be the status quo even when no interface pre-selects it. Practical check: For an important review, compare the current option and realistic alternatives as if the choice were being made today. List genuine switching costs separately from familiarity or decision effort. If staying still has the best expected value after that comparison, keeping the status quo may be rational rather than biased.","reviewState":"reviewed","sourceIds":["src-e857ab07ab417658","src-638a04b999a9cc4b","src-64c4a8072d35fe53"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"3329a8998f07683a17329692fdc6786b972dec764f4631e78729170faee759c7"}
{"chunkId":"cb-self-assessment-dunning-evidence","canonicalId":"self-assessment-dunning","canonicalUrl":"https://cognitive-biases.github.io/biases/self-assessment-dunning/","resourceType":"concept","section":"evidence","title":"Dunning–Kruger Effect – when performance and self-assessment do not line up","text":"Dunning–Kruger Effect – when performance and self-assessment do not line up. Evidence: supported but often overstated. The original studies found substantial overestimation among low performers in specific tasks, but the popular story that incompetent people are simply very confident while experts systematically underestimate themselves is too broad. Later work showed that common quartile plots can exaggerate the pattern through regression and better-than-average effects, while a 2023 reanalysis using more appropriate methods still found a small significant effect in intelligence data. Mechanism: The original account proposed that limited skill can also limit the metacognitive ability needed to recognize errors. However, observed calibration curves also reflect statistical and measurement properties. The safest interpretation is about miscalibration between objective performance and self-assessment, not a universal personality rule about confidence and competence. Practical check: Do not use Dunning–Kruger as a label for people you disagree with. For your own decisions, pair confidence with objective feedback, calibration data, or tests of performance. Track whether confidence changes appropriately after feedback rather than assuming that confidence itself reveals competence.","reviewState":"reviewed","sourceIds":["src-c707dd3996062dff","src-8333008aad319119","src-7a75dea73a03745a"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"d03d27d01248a995835e73cc50b8d5cd602de8a5d7ee9dc45c8ea996989ad6fe"}
{"chunkId":"cb-self-assessment-hot-evidence","canonicalId":"self-assessment-hot","canonicalUrl":"https://cognitive-biases.github.io/biases/self-assessment-hot/","resourceType":"concept","section":"evidence","title":"Hot–Cold Empathy Gap – when one state is a poor guide to choices in another","text":"Hot–Cold Empathy Gap – when one state is a poor guide to choices in another. Evidence: well established across state-dependent judgment research. The hot–cold empathy gap describes difficulty predicting preferences, behavior, or experience across different visceral or affective states. In a relatively cold state, people can underappreciate how pain, hunger, sexual arousal, craving, fear, anger, and other hot states will change motivation and choice; in a hot state, they can also overproject the current state into the future. The pattern can be intrapersonal or interpersonal and should not be reduced to a generic claim that emotion always causes bad decisions. Mechanism: Visceral and affective states change the relative desirability and salience of actions. Forecasts made outside the target state rely on an incomplete simulation of those motivational changes. The empathy-gap framework overlaps with projection bias and affective forecasting, but focuses specifically on failures to appreciate state-dependent changes rather than on every error in predicting future feelings or preferences. Practical check: For a consequential decision that will be made or lived through in a different state, identify the relevant state explicitly and gather evidence from people or past episodes in that state. Use precommitments for predictable hot-state choices, but keep an escape route when the future state contains information the cold-state planner could not know. In interpersonal decisions, ask what the other person's current pain, craving, fear, or arousal changes rather than assuming your present state is a neutral reference point.","reviewState":"reviewed","sourceIds":["src-30911b2f36e96c9d","src-adcadd8b7935bf35","src-26f3b86cc6d92ecf"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"c0785afead8ba084a2a671ac0224142d624f481fa1797d20b7614cd862643590"}
{"chunkId":"cb-truth-judgment-illusory-truth-effect-evidence","canonicalId":"truth-judgment-illusory-truth-effect","canonicalUrl":"https://cognitive-biases.github.io/biases/truth-judgment-illusory-truth-effect/","resourceType":"concept","section":"evidence","title":"Illusory Truth Effect – when repeated lies start to feel like truth","text":"Illusory Truth Effect – when repeated lies start to feel like truth. Evidence: robust. Repeated information is, on average, judged as more truthful than comparable new information. The size of the effect varies with the material and procedure, and repetition does not make every claim believable or erase all prior knowledge. Mechanism: Repetition can make information easier or more coherent to process and retrieve. People can then use that subjective ease or familiarity as one cue when judging truth, even though repetition itself is not evidence that the statement is correct. Practical check: When a claim feels true mainly because it sounds familiar, switch from familiarity to source checking. Ask where the claim came from and what evidence would support it. Do not treat familiarity as proof, but do not treat repetition as proof of falsehood either.","reviewState":"reviewed","sourceIds":["src-c393b2ca6b3172da","src-e17b930d0863e88a","src-7faeefb6c3a24d1d"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"698a78f0c1fac8b0b625755b42bacfec73602ca32a03762c0c401ef289977489"}
{"chunkId":"cb-work-project-decisions-guide","canonicalId":"work-project-decisions","canonicalUrl":"https://cognitive-biases.github.io/contexts/work-project-decisions/","resourceType":"context","section":"guide","title":"Work & project decisions","text":"Use these evidence-reviewed lenses when a team is choosing a direction, interpreting results, deciding whether to continue, or turning metrics into action. A project review is being driven by the latest success or failure. A team keeps investing because too much has already been spent. A KPI is becoming more important than the goal it was meant to represent. The team is searching for evidence after already preferring one solution. A retrospective is quietly rewriting what was knowable, how good the decision was, or how blameworthy the decision-maker now seems. Freeze the information set: what did the team actually know before the latest outcome? Separate decision-process quality from whether the result happened to be good or bad. When blame or praise enters the review, separate intention, reasonable belief, controllable risk, causal responsibility, and the eventual outcome. Write the next increment of cost and expected value without counting already-spent resources as a future benefit. Name the strategic goal behind each important metric and one way the metric could improve while the goal gets worse. Specify what evidence would change the preferred plan before gathering another round of supporting evidence.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"1e8c035343a7b04967d59fd026edd19074cd96631c20c58e9dfc0705e1da238b"}
{"chunkId":"cb-forecasting-future-choices-guide","canonicalId":"forecasting-future-choices","canonicalUrl":"https://cognitive-biases.github.io/contexts/forecasting-future-choices/","resourceType":"context","section":"guide","title":"Forecasting & future choices","text":"Use these evidence-reviewed lenses when estimating uncertainty, imagining future feelings or preferences, or reviewing forecasts after the outcome is known. A vivid recent example is dominating a probability estimate. A broad event looks less likely than the sum of its possible forms. A current mood, appetite, pain level, craving, or preference is shaping a long-term commitment. A future event is expected to make life dramatically better or worse for a long time. The finished outcome now feels much more predictable than it did beforehand. Write the packed event first, then unpack it into mutually exclusive possibilities and compare the totals. Check a base rate or reference class before using vivid examples as the probability estimate. Record the current state and the future state you are trying to predict; note where motives, cravings, pain, fear, or arousal may differ. Forecast emotional intensity and duration separately, including what an ordinary week around the event will contain. Store the forecast before the outcome so retrospective learning does not depend on reconstructed memory.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"a76dea2165241f35c2585c90577a1a48471330ace4996120353e868ac636b708"}
{"chunkId":"cb-ai-assisted-decisions-guide","canonicalId":"ai-assisted-decisions","canonicalUrl":"https://cognitive-biases.github.io/contexts/ai-assisted-decisions/","resourceType":"context","section":"guide","title":"AI-assisted decisions","text":"Use these evidence-reviewed lenses when a chatbot, model, recommendation system, or automated decision aid is influencing what you believe or do. You are tempted to accept an AI recommendation without checking the underlying evidence. A fluent, warm, or humanlike interface feels more capable or trustworthy than its tested performance justifies. The system’s appearance, voice, avatar, or conversational style is shaping what you think it can understand or do. An early AI score, estimate, or suggested number is becoming the starting point for your own judgment. You are prompting an AI mainly to strengthen a conclusion you already prefer. A claim has appeared in several AI outputs and is starting to feel true because it is familiar. Write a provisional judgment, success criterion, or uncertainty range before looking at AI advice when the decision is consequential. Separate the interface from the capability: list what the system actually needs to know, access, or verify to support the recommendation. When the AI supplies a number, compare it with an independent estimate, base rate, or reference class instead of adjusting only from the model’s value. Ask for the strongest counterevidence or alternative explanation, then verify important claims against an independent source rather than another generated paraphrase. Treat repeated wording as repetition, not independent evidence; trace important claims back to primary or genuinely independent sources. Record what changed after using AI, what was independently checked, and what trigger would make you reopen the decision.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"4dc6bf8f692955c09fd97ef5c8a5f64e7923e28f5e2ee4cf5530db7d40505952"}
{"chunkId":"cb-project-estimation-delivery-guide","canonicalId":"project-estimation-delivery","canonicalUrl":"https://cognitive-biases.github.io/contexts/project-estimation-delivery/","resourceType":"context","section":"guide","title":"Project estimation & delivery","text":"Use these evidence-reviewed lenses when setting a deadline, estimating effort, discussing delivery risk, or learning from a project that finished later than expected. A deadline is being built from a detailed plan without checking how long similar work actually took. An early target, budget, or rough estimate is pulling later discussion toward the same number. A team is debating one precise date even though important dependencies and unknowns remain. A project is already late and the new estimate is again starting from the ideal plan. Delivery risk is discussed as one broad feeling instead of explicit ways the plan could slip. After delivery, the final outcome is changing what people remember about the original estimate. Start with comparable completed work. Write down what similar tasks actually took before refining the current plan. Check whether a target, budget, or early rough number became an anchor. Build an independent estimate when possible before reconciling the two. Give a useful range and explain what would push delivery toward the earlier or later end. Unpack the main ways the plan can slip instead of hiding all uncertainty inside one confidence number. Record the estimate, assumptions and important unknowns before work starts so the later review has a real baseline. When the plan changes, judge the next step on future cost and value rather than on how much has already been spent.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"a173e5f768b19d7419d552c333e4c6a17055541c13f9eb7055a405401462f32c"}
{"chunkId":"cb-checking-claims-misinformation-guide","canonicalId":"checking-claims-misinformation","canonicalUrl":"https://cognitive-biases.github.io/contexts/checking-claims-misinformation/","resourceType":"context","section":"guide","title":"Checking claims & misinformation","text":"Use these evidence-reviewed lenses when deciding whether a repeated claim is trustworthy, checking a correction, or testing an explanation against plausible alternatives. A claim feels credible because you have seen it in many places, but those places may be repeating the same source. You are searching mainly for evidence that supports a conclusion you already prefer. A test supports the leading explanation, but plausible alternatives would predict the same result. One vivid example is shaping your judgment more than broader evidence. A false claim has been corrected, but the old explanation still appears in later reasoning. Someone resists a correction and the reaction is immediately described as a backfire effect. Write the exact claim you are checking. Separate the claim itself from your opinion about the person or source sharing it. Trace repeated versions back to genuinely independent sources. Ten repetitions of one source are still one source. Write what evidence would make the preferred claim weaker or wrong before gathering another round of support. List at least one plausible alternative explanation. For an important test, predict the result under both explanations and prefer observations where their predictions differ. Use the same evidence-quality standard for results that support and challenge the preferred explanation. If a correction is needed, state the corrected information clearly and provide the best supported alternative explanation when one is available. After a correction, distinguish incomplete updating from backfire. Ask whether belief in the false claim actually became stronger or whether some influence simply remained.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"8a836b8a2d0d3131519c64c2d9ec427d8d52bdc11da8b9808ec9a000e4ec312f"}
{"chunkId":"cb-comparing-plans-pricing-guide","canonicalId":"comparing-plans-pricing","canonicalUrl":"https://cognitive-biases.github.io/contexts/comparing-plans-pricing/","resourceType":"context","section":"guide","title":"Comparing plans & pricing","text":"Use these evidence-reviewed lenses when comparing subscription tiers, product plans, vendor offers or recommendation menus where the way options are arranged may change what looks attractive. A three-tier pricing table makes one plan suddenly look like the obvious value choice. An extra option is clearly worse than one plan but still changes how the stronger options are compared. One plan is already selected or renews automatically unless the user changes it. A list price, target budget or recommended price is becoming the reference point for the comparison. Equivalent features or outcomes are described differently across plans, making direct comparison difficult. List the serious options and the attributes that matter to your goal before using badges, highlights or recommended labels. Remove any suspected decoy and compare the remaining options again. If your preference changes, inspect what contrast the extra option created. Identify the no-action outcome separately. A default can influence choice even when no decoy is present. Hide or replace the starting price, target or recommendation when practical and make an independent estimate of value before reconciling it with the displayed number. Rewrite important attributes in matched units and neutral language so equivalent differences are easy to compare. Choose on expected value and fit to the actual need. Conversion, popularity and a 'recommended' badge are not substitutes for the decision criteria.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"bea4dc7f56c64088927104b0679ddb43a8ab645edee30092280fbf9937e01257"}
{"chunkId":"cb-defaults-settings-choice-architecture-guide","canonicalId":"defaults-settings-choice-architecture","canonicalUrl":"https://cognitive-biases.github.io/contexts/defaults-settings-choice-architecture/","resourceType":"context","section":"guide","title":"Defaults, settings & choice architecture","text":"Use these evidence-reviewed lenses when a form, product, policy or system decides what happens if a person does nothing, or when an existing setting is difficult to reconsider. A checkbox, permission, subscription or contribution is already selected before the user makes a choice. A product team is deciding what the starting settings should be for a new user. People rarely change a setting and the team is treating that as proof that the setting is preferred. An old workflow or service remains in place even though alternatives now exist. A recommendation, suggested number or wording may be influencing the choice alongside the default. Write down what happens if the person takes no action. Make the default explicit before discussing conversion or uptake. Separate the designed default from the existing status quo. They can be the same option, but they do not have to be. Make important alternatives visible and keep switching reasonably easy. Friction can create persistence without strong preference. When possible, compare the design with active choice or another default while keeping the underlying options the same. Check whether wording, recommended values or starting numbers are also changing the decision. Do not attribute every difference to the default. Evaluate whether the resulting choice serves the user's goals. Uptake alone is not enough to establish welfare or informed preference.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"623a62148c1df8502439547a998027c00ec3ff4a1639c7806a790a5f4da74d5a"}
{"chunkId":"cb-presenting-risk-options-guide","canonicalId":"presenting-risk-options","canonicalUrl":"https://cognitive-biases.github.io/contexts/presenting-risk-options/","resourceType":"context","section":"guide","title":"Presenting risk & options","text":"Use these evidence-reviewed lenses when a report, interface, model, or recommendation is presenting numbers and choices that other people will use to decide. The same outcome can be described as a gain, a loss, a success rate, or a failure rate. A reference point determines whether the same change is experienced as gaining something or losing something. One prominent number appears before people form their own estimate. A broad risk category hides several different ways the outcome could happen. A vivid example or recent incident is being used next to statistical evidence. An AI interface chooses the wording, score, ordering, or examples that a person sees before deciding. Write the underlying outcomes, quantities, and probabilities in a neutral form before choosing presentation language. State the reference point explicitly and compare matched positive and negative changes. When possible, also show the absolute outcomes rather than only the gain or loss from the reference point. Create matched gain and loss descriptions for important risk choices. Check whether the decision changes even though the underlying outcomes do not. Identify the first numerical reference point. When practical, collect an independent estimate before revealing it or compare against several relevant references. Unpack broad risks into mutually exclusive possibilities and compare the total with the original packed estimate. Separate illustrative examples from frequency evidence. A memorable case can help explain a risk without representing how common it is. For consequential interfaces or reports, test alternative presentations and measure the resulting choices instead of assuming one wording is neutral.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"279eace35eeb512066e43c389e32d4e9dc9a23ff2b2436af3ca421e4f844d741"}
{"chunkId":"cb-reviewing-kpis-proxy-metrics-guide","canonicalId":"reviewing-kpis-proxy-metrics","canonicalUrl":"https://cognitive-biases.github.io/contexts/reviewing-kpis-proxy-metrics/","resourceType":"context","section":"guide","title":"Reviewing KPIs & proxy metrics","text":"Use these evidence-reviewed lenses when a team is using scores, dashboards, targets, benchmarks, or AI evaluations to represent a broader objective. A KPI is improving, but people disagree about whether the underlying service, product, learning, safety, or business outcome is actually better. A target is tied to incentives and teams have found ways to improve the number that do not clearly improve the goal. A dashboard or benchmark is becoming the main language of the review while the underlying objective is harder to discuss. A benchmark or model score is being treated as the main evidence that an AI system is useful in real work. A successful result is being used to justify the metric or target even though the quality of the measurement and decision process was never tested. Teams mainly look for examples that support the current KPI and rarely ask what evidence would show that the measure is a poor proxy. Write the underlying objective in ordinary language before opening the dashboard. Describe what success means without using the KPI name. For each important measure, list what it captures directly, what it only approximates, and at least one important dimension it misses. Ask how someone could improve the number while making the real objective worse. If that path is realistic, add a guardrail, companion measure, validation check, or narrative review. Identify the first target, benchmark, baseline, or score shown in the review. Rebuild important estimates from independent evidence before letting that number become the reference point. Write what evidence would make you change or retire the metric before the next review. Search for that evidence, not only for examples that defend the current scorecard. Judge the metric design and the eventual outcome separately. A good outcome does not prove the proxy was well chosen, and a bad outcome does not prove every part of the measurement system was wrong. For AI benchmarks and automated ratings, validate against representative real tasks and human outcomes instead of assuming one convenient score is the product objective. Check the measurement process separately when data coverage, sampling, calibration, labels, or exclusions may be systematically distorted.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"accb43bb1141f4b66f5504c80dbf2167a2423206942bbeecf9b20d0cd69b1c40"}
{"chunkId":"cb-comparing-past-present-guide","canonicalId":"comparing-past-present","canonicalUrl":"https://cognitive-biases.github.io/contexts/comparing-past-present/","resourceType":"context","section":"guide","title":"Was the past really better?","text":"Use these evidence-reviewed lenses when a personal memory or broad story about decline is being used to compare the present with an earlier period. Someone says a profession, product, institution, city, community, or society used to be better, but the indicators and comparison years are unclear. A remembered project, job, school period, trip, relationship, or community feels much better now than it felt while it was happening. Current problems are vivid and detailed while the negative emotional intensity of past problems feels weaker. A few personal memories are being used as evidence for a broad historical trend. Nostalgia or dissatisfaction with the present is starting to replace measurable comparison. Define the claim before debating it. Write what exactly was better or worse, for whom, where, and between which years. Separate personal event memory from population or system trends. A remembered experience can be useful evidence about your experience without representing an era. Look for contemporaneous records from both periods: surveys, logs, prices, defect rates, service levels, health or safety measures, diaries, ratings, or other domain-relevant data. For a remembered event, compare later recollection with records made during the experience when those records exist. Separate remembered facts from current emotional intensity. Ask whether negative feelings connected with the old experience may have softened over time. Look for indicators that improved, worsened, and stayed stable. Avoid forcing a mixed historical record into one global better-or-worse verdict. State what cannot be determined when comparable historical evidence is weak. Memory can be meaningful without being a complete measurement system.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"fd45a075f0ee8cd2d510e1edb1b7544652c0259c89e3cffbf6cade9056cabf40"}
{"chunkId":"cb-continue-change-stop-project-guide","canonicalId":"continue-change-stop-project","canonicalUrl":"https://cognitive-biases.github.io/contexts/continue-change-stop-project/","resourceType":"context","section":"guide","title":"Continue, change, or stop a project","text":"Use these evidence-reviewed lenses when a project has disappointing results and the next decision is whether to invest more, change direction, or stop. Past spending is the strongest argument for approving the next budget or milestone. Stopping feels like accepting or making a loss real, even before the remaining future options are compared. The person who chose the original plan is also deciding whether it should continue. A project is described as too close to completion to stop, but the remaining value and cost are unclear. A target, original business case, or early estimate is still shaping the review after the evidence changed. An AI or external adviser recommends continuing and the recommendation is starting to replace an independent review. Separate past costs from the next decision. Write the remaining costs, expected future value, important risks, and realistic alternatives from today forward. Name the reference point that makes stopping or changing direction feel like a loss. Distinguish that prospective loss from money or time that is already irrecoverable. Describe what useful completion actually produces. Being close to finished can matter when completion has future value, but percentage complete is not value by itself. Rebuild the current estimate from relevant evidence instead of adjusting only from the original budget, target, or deadline. Check ownership pressure. Ask someone who did not make the original decision to review the next commitment using the same evidence. Write continue, change, and stop criteria before the next result arrives. Include what evidence would trigger each option. After the outcome, review whether the process followed those criteria separately from whether the final result happened to be good or bad.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"58cf91e7bc1e07a962ac99e47f6e230de945676d958df1e2d79779061d3b2ef8"}
{"chunkId":"cb-practice-work-project-decisions-exercises","canonicalId":"practice-work-project-decisions","canonicalUrl":"https://cognitive-biases.github.io/practice/work-project-decisions/","resourceType":"practice-set","section":"exercises","title":"Practice: Work & project decisions","text":"Use these evidence-reviewed lenses when a team is choosing a direction, interpreting results, deciding whether to continue, or turning metrics into action. Check: What evidence would make us abandon or materially change the preferred explanation? Best first lens: Confirmation Bias. Check: Would we rate this decision process the same way if the outcome had gone the other direction? Best first lens: Outcome Bias. Check: What did our actual pre-outcome forecast or notes say before the result became obvious? Best first lens: Hindsight Bias. Check: Would we assign the same blame or praise if the decision-maker had the same intent, beliefs, and controllable risk but luck produced a different outcome? Best first lens: Moral Luck. Check: If we had not already invested anything, would we still fund the next step on its future merits? Best first lens: Escalation of Commitment. Check: What underlying goal is this metric supposed to represent, and where can the proxy diverge from it? Best first lens: Surrogation.","reviewState":"reviewed","sourceIds":["src-83375467c08e8177","src-c4e8734b8f1248ee","src-f1f0400ef362aaa8","src-1801b1347f89ef91","src-6df53f947e2583fa","src-75d0eb4cf8b10c07","src-0539c30082b039b2","src-295bb6762d6f7a9f","src-0c83b7cb76f4adab","src-644614d8919a2ee4","src-25c242b954867ac1","src-ec03691286f7284e","src-bc1fa4e94ab0fa71","src-dd6028af1211deb7"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"7cc5676845c5dc342dfa96782776fc1152a206cf9d3e5ff1d236c7eefcf24a99"}
{"chunkId":"cb-practice-forecasting-future-choices-exercises","canonicalId":"practice-forecasting-future-choices","canonicalUrl":"https://cognitive-biases.github.io/practice/forecasting-future-choices/","resourceType":"practice-set","section":"exercises","title":"Practice: Forecasting & future choices","text":"Use these evidence-reviewed lenses when estimating uncertainty, imagining future feelings or preferences, or reviewing forecasts after the outcome is known. Check: Does the total probability change when the same event is unpacked into explicit possibilities? Best first lens: Subadditivity Effect. Check: Are the examples easy to recall because they are common, or because they are vivid, recent, or repeated? Best first lens: Availability Heuristic. Check: Which part of today's state am I assuming will still describe my future self? Best first lens: Projection Bias. Check: Am I forecasting the focal event while forgetting the rest of ordinary future life? Best first lens: Impact Bias. Check: Am I predicting choices in a future hot or cold state from a state with different motives, cravings, pain, fear, or arousal? Best first lens: Hot–Cold Empathy Gap. Check: What probability did I actually assign before I learned the outcome? Best first lens: Hindsight Bias.","reviewState":"reviewed","sourceIds":["src-e5eecafb672bd8ed","src-9f9930db613528b2","src-ffae461551a5d8e7","src-723760d16e30b39e","src-f612390f3949191b","src-0ba7057e7c37fb49","src-479735fa3da2d5b3","src-3828d4aac220e974","src-b0cc2953c2431e7b","src-ebc67fd4db360d62","src-ea6429b27b8d6ccf","src-bf639003c08f7257","src-902102ef57fb68e5","src-30911b2f36e96c9d","src-adcadd8b7935bf35","src-26f3b86cc6d92ecf","src-6df53f947e2583fa","src-75d0eb4cf8b10c07"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"a3cd628cb7db7e2dd6e972b0c99c81f70f06ac10839b5b96bdec36525ea4159d"}
{"chunkId":"cb-practice-ai-assisted-decisions-exercises","canonicalId":"practice-ai-assisted-decisions","canonicalUrl":"https://cognitive-biases.github.io/practice/ai-assisted-decisions/","resourceType":"practice-set","section":"exercises","title":"Practice: AI-assisted decisions","text":"Use these evidence-reviewed lenses when a chatbot, model, recommendation system, or automated decision aid is influencing what you believe or do. Check: Am I using the automated recommendation as a substitute for checking the evidence I could realistically verify? Best first lens: Automation Bias. Check: Did the AI’s first number become my starting point before I formed an independent estimate? Best first lens: Anchoring Effect. Check: Which humanlike cues are making me infer understanding, intention, empathy, or competence that I have not actually tested? Best first lens: Anthropomorphism. Check: What capabilities am I inferring from appearance, voice, interface polish, or conversational style rather than observed performance? Best first lens: Appearance–Capability Expectation. Check: Did I ask the AI to test my preferred conclusion, or mainly to produce better arguments for it? Best first lens: Confirmation Bias. Check: Does this claim feel more credible because I have encountered it repeatedly, or because I verified independent evidence for it? Best first lens: Illusory Truth Effect.","reviewState":"reviewed","sourceIds":["src-0252755d114dfb6c","src-f8b4657a59aad7de","src-965010b0cf7cde95","src-d8b0db0a156ad3a7","src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379","src-24b9394643fb3945","src-2c6a7b7f9b72cb27","src-f57ecaab4be6bbd9","src-b4d5205770653dfc","src-e3698f659e09ea7e","src-394b04c1e1398c8f","src-09bef5bf4eda8006","src-83375467c08e8177","src-c4e8734b8f1248ee","src-c393b2ca6b3172da","src-e17b930d0863e88a","src-7faeefb6c3a24d1d"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"a18ece03d280f00475955288047def23720ffbcda762f010769d491144a93e17"}
{"chunkId":"cb-practice-project-estimation-delivery-exercises","canonicalId":"practice-project-estimation-delivery","canonicalUrl":"https://cognitive-biases.github.io/practice/project-estimation-delivery/","resourceType":"practice-set","section":"exercises","title":"Practice: Project estimation & delivery","text":"Use these evidence-reviewed lenses when setting a deadline, estimating effort, discussing delivery risk, or learning from a project that finished later than expected. Check: What happened on the most comparable completed work, and why should this case be faster or slower? Best first lens: Planning Fallacy. Check: Which early number is shaping this estimate, and what would we estimate if we had not seen it first? Best first lens: Anchoring Effect. Check: Does the risk estimate change when we unpack the main ways the project could be delayed? Best first lens: Subadditivity Effect. Check: Are we using a vivid recent project as the baseline because it is representative, or simply because it is easy to remember? Best first lens: Availability Heuristic. Check: What did the original estimate and assumptions actually say before we knew the delivery result? Best first lens: Hindsight Bias. Check: If we were deciding only about the remaining work today, would we still choose the same next step? Best first lens: Escalation of Commitment.","reviewState":"reviewed","sourceIds":["src-7d92150bd312c109","src-cf328825be4a0674","src-d6ecdecbd0ffae61","src-d98817dbf800d030","src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379","src-e5eecafb672bd8ed","src-9f9930db613528b2","src-ffae461551a5d8e7","src-723760d16e30b39e","src-f612390f3949191b","src-0ba7057e7c37fb49","src-6df53f947e2583fa","src-75d0eb4cf8b10c07","src-644614d8919a2ee4","src-25c242b954867ac1","src-ec03691286f7284e"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"7b8a6f7cf96bb45caa8a65dde8fe0950da28cf010016e12ec0653fad290f98a4"}
{"chunkId":"cb-practice-checking-claims-misinformation-exercises","canonicalId":"practice-checking-claims-misinformation","canonicalUrl":"https://cognitive-biases.github.io/practice/checking-claims-misinformation/","resourceType":"practice-set","section":"exercises","title":"Practice: Checking claims & misinformation","text":"Use these evidence-reviewed lenses when deciding whether a repeated claim is trustworthy, checking a correction, or testing an explanation against plausible alternatives. Check: Am I using the same evidence standard for information that supports and challenges the claim? Best first lens: Confirmation Bias. Check: Would this test distinguish the preferred explanation from plausible alternatives, or would they predict the same result? Best first lens: Congruence Bias. Check: Does this feel true because it was independently verified, or because I have encountered the same claim repeatedly? Best first lens: Illusory Truth Effect. Check: Is the example easy to recall because it is representative, or because it is vivid, recent, emotional, or repeated? Best first lens: Availability Heuristic. Check: After the correction, am I still using part of the old information when explaining or judging the situation? Best first lens: Continued Influence Effect. Check: Did belief in the corrected false claim actually become stronger, or am I calling disagreement or incomplete updating a backfire? Best first lens: Backfire Effect.","reviewState":"reviewed","sourceIds":["src-83375467c08e8177","src-c4e8734b8f1248ee","src-54ea5126c3035a7e","src-f3a6c8d07f4870ec","src-c393b2ca6b3172da","src-e17b930d0863e88a","src-7faeefb6c3a24d1d","src-723760d16e30b39e","src-f612390f3949191b","src-0ba7057e7c37fb49","src-9811dba9926395b6","src-6891b8bf627a6ec7","src-422189df9f5faa26","src-80cc3fc9e30c3c69","src-4f99e4ca8c1af4fe","src-e1813cd3b9123eaf","src-40282584441d78a7"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"b2cc947417dbee4719a7206a4d01f9723987286eea4f14d162dcc124bde227c7"}
{"chunkId":"cb-practice-comparing-plans-pricing-exercises","canonicalId":"practice-comparing-plans-pricing","canonicalUrl":"https://cognitive-biases.github.io/practice/comparing-plans-pricing/","resourceType":"practice-set","section":"exercises","title":"Practice: Comparing plans & pricing","text":"Use these evidence-reviewed lenses when comparing subscription tiers, product plans, vendor offers or recommendation menus where the way options are arranged may change what looks attractive. Check: Does removing the inferior or strategically positioned option change my preference between the remaining plans? Best first lens: Decoy Effect. Check: Which plan applies if I do nothing, and is that automatic outcome being mistaken for preference? Best first lens: Default Effect. Check: Which displayed price, budget or recommendation became the starting point for my value judgment? Best first lens: Anchoring Effect. Check: Are equivalent features, gains, losses or rates described differently across the options? Best first lens: Framing Effect.","reviewState":"reviewed","sourceIds":["src-84a20f469659275d","src-6f220811f3f7fc69","src-fe432751bebf733b","src-37cf79ece99ea6c9","src-7d9d96710fab2c5f","src-64c4a8072d35fe53","src-90e516c044f0e914","src-5e28781720bc4f35","src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379","src-bb0fa40a53281f49","src-0e441147fc835a3d","src-47b281f980645234","src-2e83fc60350e0524","src-dc40b5ceda9cbea1"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"b106e438095b847d9c668d32ef6c15c01ac9414726d00c9a7decf1b2988fd026"}
{"chunkId":"cb-practice-defaults-settings-choice-architecture-exercises","canonicalId":"practice-defaults-settings-choice-architecture","canonicalUrl":"https://cognitive-biases.github.io/practice/defaults-settings-choice-architecture/","resourceType":"practice-set","section":"exercises","title":"Practice: Defaults, settings & choice architecture","text":"Use these evidence-reviewed lenses when a form, product, policy or system decides what happens if a person does nothing, or when an existing setting is difficult to reconsider. Check: What happens automatically if the person does nothing, and how much does that pre-selection change uptake? Best first lens: Default Effect. Check: Is the current option getting extra weight simply because it is already in place? Best first lens: Status Quo Bias. Check: Would the choice change if the same outcomes were described with an equivalent gain, loss or attribute frame? Best first lens: Framing Effect. Check: Did a suggested number, threshold or starting value become the reference point for the decision? Best first lens: Anchoring Effect.","reviewState":"reviewed","sourceIds":["src-7d9d96710fab2c5f","src-64c4a8072d35fe53","src-90e516c044f0e914","src-5e28781720bc4f35","src-e857ab07ab417658","src-638a04b999a9cc4b","src-bb0fa40a53281f49","src-0e441147fc835a3d","src-47b281f980645234","src-2e83fc60350e0524","src-dc40b5ceda9cbea1","src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"b67c7922847abf38d46d5e951d4091f39d30fb4bc7fbcdbd245d3590815396d0"}
{"chunkId":"cb-practice-presenting-risk-options-exercises","canonicalId":"practice-presenting-risk-options","canonicalUrl":"https://cognitive-biases.github.io/practice/presenting-risk-options/","resourceType":"practice-set","section":"exercises","title":"Practice: Presenting risk & options","text":"Use these evidence-reviewed lenses when a report, interface, model, or recommendation is presenting numbers and choices that other people will use to decide. Check: Would the choice change if the same outcomes were presented with an equivalent gain, loss, or neutral description? Best first lens: Framing Effect. Check: What is the reference point, and are equivalent losses receiving more weight than comparable gains in this decision? Best first lens: Loss Aversion. Check: Which number is shown first, and is it pulling later estimates toward itself? Best first lens: Anchoring Effect. Check: Does the total risk estimate change when the same outcome is unpacked into explicit possibilities? Best first lens: Subadditivity Effect. Check: Is a vivid example affecting the estimate because it is representative, or mainly because it is easy to recall? Best first lens: Availability Heuristic.","reviewState":"reviewed","sourceIds":["src-bb0fa40a53281f49","src-0e441147fc835a3d","src-47b281f980645234","src-2e83fc60350e0524","src-dc40b5ceda9cbea1","src-48160eef6b029066","src-e8a194eeb0017a8a","src-368a6a9d36d10612","src-72b6f021c45204d9","src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379","src-e5eecafb672bd8ed","src-9f9930db613528b2","src-ffae461551a5d8e7","src-723760d16e30b39e","src-f612390f3949191b","src-0ba7057e7c37fb49"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"50de14e924e7d140c0b69ad7550c3af9a00718c9836f31d4828238736a14f7fc"}
{"chunkId":"cb-practice-reviewing-kpis-proxy-metrics-exercises","canonicalId":"practice-reviewing-kpis-proxy-metrics","canonicalUrl":"https://cognitive-biases.github.io/practice/reviewing-kpis-proxy-metrics/","resourceType":"practice-set","section":"exercises","title":"Practice: Reviewing KPIs & proxy metrics","text":"Use these evidence-reviewed lenses when a team is using scores, dashboards, targets, benchmarks, or AI evaluations to represent a broader objective. Check: Are we treating the measure as evidence about the objective, or have we started treating the measure as the objective itself? Best first lens: Surrogation. Check: Which target, benchmark, baseline, or first score is shaping later judgments before we build an independent estimate? Best first lens: Anchoring Effect. Check: What evidence would show that this KPI is a poor proxy, and have we actively looked for it? Best first lens: Confirmation Bias. Check: Are we judging the quality of the metric and decision process mainly from whether the final result happened to be good or bad? Best first lens: Outcome Bias.","reviewState":"reviewed","sourceIds":["src-bc1fa4e94ab0fa71","src-dd6028af1211deb7","src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379","src-83375467c08e8177","src-c4e8734b8f1248ee","src-f1f0400ef362aaa8","src-1801b1347f89ef91"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"43c21b8497abb1ef409a36cb96b7c53c71bb67907a725558f640423eaa3fe883"}
{"chunkId":"cb-practice-comparing-past-present-exercises","canonicalId":"practice-comparing-past-present","canonicalUrl":"https://cognitive-biases.github.io/practice/comparing-past-present/","resourceType":"practice-set","section":"exercises","title":"Practice: Was the past really better?","text":"Use these evidence-reviewed lenses when a personal memory or broad story about decline is being used to compare the present with an earlier period. Check: Is a broad story of decline being asserted without a clearly defined indicator, population, and historical comparison? Best first lens: Declinism. Check: Is a specific past experience remembered more positively now than it was experienced or recorded at the time? Best first lens: Rosy Retrospection. Check: Could the emotional intensity of negative past events have faded faster than positive affect, changing how the period feels in comparison with the present? Best first lens: Fading Affect Bias.","reviewState":"reviewed","sourceIds":["src-c375f114daf625f6","src-c378be567ea78220","src-f3b090aa4f2bf959","src-52af9360ef62152d","src-816c78f1bdb1f366"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"32b7b8d35978af366287a34b4b9a62eaf6457fbaa1929112e15ddfac547c74e6"}
{"chunkId":"cb-practice-continue-change-stop-project-exercises","canonicalId":"practice-continue-change-stop-project","canonicalUrl":"https://cognitive-biases.github.io/practice/continue-change-stop-project/","resourceType":"practice-set","section":"exercises","title":"Practice: Continue, change, or stop a project","text":"Use these evidence-reviewed lenses when a project has disappointing results and the next decision is whether to invest more, change direction, or stop. Check: Which past costs are irrecoverable, and would they change what we choose if we evaluated only the future options? Best first lens: Sunk Cost Effect. Check: Is stopping being evaluated as a prospective loss from the current reference point, separately from the sunk costs already paid? Best first lens: Loss Aversion. Check: Are setbacks leading us to commit more resources without reopening whether this course still deserves the next investment? Best first lens: Escalation of Commitment. Check: Which original target, budget, valuation, or deadline is still pulling the current judgment toward it? Best first lens: Anchoring Effect. Check: What does comparable completed work say about the cost and time still required from today? Best first lens: Planning Fallacy. Check: What evidence would make us stop or materially change the project, and have we actively looked for it? Best first lens: Confirmation Bias. Check: Would we rate the quality of today’s continue-or-stop process the same way if the eventual outcome went the other direction? Best first lens: Outcome Bias.","reviewState":"reviewed","sourceIds":["src-25c242b954867ac1","src-812f426a52613922","src-ec03691286f7284e","src-48160eef6b029066","src-e8a194eeb0017a8a","src-368a6a9d36d10612","src-72b6f021c45204d9","src-644614d8919a2ee4","src-4e94d948f6b94566","src-35855b06b246eebe","src-639229081285f101","src-ce811297b2050379","src-7d92150bd312c109","src-cf328825be4a0674","src-d6ecdecbd0ffae61","src-d98817dbf800d030","src-83375467c08e8177","src-c4e8734b8f1248ee","src-f1f0400ef362aaa8","src-1801b1347f89ef91"],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"59351c66ad39a02ca271514884f91761ab69ba33457ac0821774f4842f46810c"}
{"chunkId":"cb-hindsight-bias-vs-outcome-bias-distinction","canonicalId":"hindsight-bias-vs-outcome-bias","canonicalUrl":"https://cognitive-biases.github.io/compare/hindsight-bias-vs-outcome-bias/","resourceType":"comparison","section":"distinction","title":"Hindsight Bias vs Outcome Bias","text":"Both distort retrospective judgment after the result is known, but they target different questions. Hindsight bias makes the result look more predictable in retrospect; outcome bias makes the decision itself look better or worse because the result was good or bad. Ask what changed after the outcome became known: your estimate of how predictable the result was, or your evaluation of the decision process?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"7448d60bdaee99b495ba5e8067806f797a6d3207f769b3a3fadd6f8e6c1dfd91"}
{"chunkId":"cb-outcome-bias-vs-moral-luck-distinction","canonicalId":"outcome-bias-vs-moral-luck","canonicalUrl":"https://cognitive-biases.github.io/compare/outcome-bias-vs-moral-luck/","resourceType":"comparison","section":"distinction","title":"Outcome Bias vs Moral Luck","text":"Both let a known result reshape a retrospective judgment, but the target is different. Outcome bias changes how good or bad a decision process looks; moral luck changes blame, praise, punishment, or moral responsibility while also depending on the actor's intentions, beliefs, causal role, and controllable risk. Ask what you are judging: the quality of the decision process, or the moral responsibility and blameworthiness of the person who acted?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"fd6728b745ffc407ab61bc6d9c3ef24c55a861d17b109afa92325a6420368ba9"}
{"chunkId":"cb-confirmation-bias-vs-backfire-effect-distinction","canonicalId":"confirmation-bias-vs-backfire-effect","canonicalUrl":"https://cognitive-biases.github.io/compare/confirmation-bias-vs-backfire-effect/","resourceType":"comparison","section":"distinction","title":"Confirmation Bias vs Backfire Effect","text":"Both can involve resistance to evidence, but they are not the same construct. Confirmation bias is a broad family of search, testing, interpretation, and memory tendencies shaped by existing beliefs; the backfire effect is the narrower outcome in which a correction makes belief in the targeted false claim stronger. Ask whether you are observing selective information processing, or a measured increase in the false belief after a correction. Without the latter, you have not shown a backfire effect.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"874f27c75d80b83fd04cd8934c868667e02d9cb7ccfee42b8e925d9203b78625"}
{"chunkId":"cb-continued-influence-effect-vs-backfire-effect-distinction","canonicalId":"continued-influence-effect-vs-backfire-effect","canonicalUrl":"https://cognitive-biases.github.io/compare/continued-influence-effect-vs-backfire-effect/","resourceType":"comparison","section":"distinction","title":"Continued Influence Effect vs Backfire Effect","text":"Both happen in the context of correcting misinformation, but they describe different outcomes. Continued influence means the correction helps yet some influence of the old information remains. Backfire is the narrower case where the correction makes belief in the targeted false claim stronger. Ask what happened to the false claim after correction: did its influence decrease but remain above zero, or did belief in the claim actually increase? The first can be continued influence; only the second is backfire.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"8428668b0b44159e1b986c63f0fc932e63bc7db9a2d419560b1b700b9b2003e5"}
{"chunkId":"cb-anchoring-effect-vs-automation-bias-distinction","canonicalId":"anchoring-effect-vs-automation-bias","canonicalUrl":"https://cognitive-biases.github.io/compare/anchoring-effect-vs-automation-bias/","resourceType":"comparison","section":"distinction","title":"Anchoring Effect vs Automation Bias","text":"Both can distort AI-assisted decisions, but the error is different. Anchoring pulls a judgment toward a starting value; automation bias is inappropriate reliance on automated advice or cues. Ask what is driving the judgment: the specific starting number, or the fact that the recommendation came from an automated system?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"96332272ad4a1d519ed105a1b81c1f260fc271459bf2f55db2e122cc3101bbf3"}
{"chunkId":"cb-availability-heuristic-vs-illusory-truth-effect-distinction","canonicalId":"availability-heuristic-vs-illusory-truth-effect","canonicalUrl":"https://cognitive-biases.github.io/compare/availability-heuristic-vs-illusory-truth-effect/","resourceType":"comparison","section":"distinction","title":"Availability Heuristic vs Illusory Truth Effect","text":"Both can make familiar information influential, but they affect different judgments. Availability uses accessible examples as a cue for frequency or probability; illusory truth is the tendency for repeated statements to be judged as more true than comparable new statements. Ask what changed: how common or likely the event seems, or how true the statement seems?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"1bf6a5429889b13a255db1119e3a5a4fc5b0948e583e35a7aea3cb8467d603e1"}
{"chunkId":"cb-confirmation-bias-vs-congruence-bias-distinction","canonicalId":"confirmation-bias-vs-congruence-bias","canonicalUrl":"https://cognitive-biases.github.io/compare/confirmation-bias-vs-congruence-bias/","resourceType":"comparison","section":"distinction","title":"Confirmation Bias vs Congruence Bias","text":"Confirmation Bias is a broad umbrella for belief-consistent search and interpretation. Congruence Bias is narrower: it concerns choosing a hypothesis test that fits the favored explanation but does not discriminate well among competing explanations. Ask whether the problem is broad evidence processing around an existing belief, or specifically the choice of a test whose likely result is congruent with one hypothesis but insufficiently diagnostic against alternatives.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"6f09f7c1790a269a4feffaa3a1eb74af33117d271171678191096fe372ff2029"}
{"chunkId":"cb-curse-of-knowledge-vs-dunning-kruger-effect-distinction","canonicalId":"curse-of-knowledge-vs-dunning-kruger-effect","canonicalUrl":"https://cognitive-biases.github.io/compare/curse-of-knowledge-vs-dunning-kruger-effect/","resourceType":"comparison","section":"distinction","title":"Curse of Knowledge vs Dunning–Kruger Effect","text":"Both are often discussed around expertise, but they concern different judgments. Curse of Knowledge is about estimating another person's knowledge after you already know something. Dunning–Kruger research is about calibration between a person's own performance and their self-assessment. Ask whose knowledge is being judged: another person's perspective, or your own performance relative to your self-assessment?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"51cae9227d70e29d3851295a9d1286bcec00b11a8ee024afcf49dd9719f70cc6"}
{"chunkId":"cb-declinism-vs-rosy-retrospection-distinction","canonicalId":"declinism-vs-rosy-retrospection","canonicalUrl":"https://cognitive-biases.github.io/compare/declinism-vs-rosy-retrospection/","resourceType":"comparison","section":"distinction","title":"Declinism vs Rosy Retrospection","text":"Both can make the past look better than the present, but they operate at different levels. Declinism is a broad label for judging a domain, society, institution, or era as having declined. Rosy Retrospection is a narrower event-memory pattern in which a specific past experience is remembered more positively than it was experienced at the time. Ask whether the claim is about a broad trend across time or about how a personally experienced event is remembered later. A rosy memory can contribute to a decline story, but it does not by itself establish Declinism, and a real historical decline can exist even when memory is imperfect.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"a5f2bff7b29848269bf658a4b02a9111d7fcc105ddbdd32b9adc19e6f1260a5b"}
{"chunkId":"cb-decoy-effect-vs-default-effect-distinction","canonicalId":"decoy-effect-vs-default-effect","canonicalUrl":"https://cognitive-biases.github.io/compare/decoy-effect-vs-default-effect/","resourceType":"comparison","section":"distinction","title":"Decoy Effect vs Default Effect","text":"Both can change choices without changing the core product, but they work through different choice structures. A decoy changes the comparison between options by adding another alternative; a default changes what happens when the person does not actively choose something else. Ask what changed: was an inferior comparison option added to the choice set, or was one option made the automatic outcome under inaction?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"7ae32d6b622264287092f491943627b179f05ef4eb60e31b58737b022588c9b2"}
{"chunkId":"cb-default-effect-vs-status-quo-bias-distinction","canonicalId":"default-effect-vs-status-quo-bias","canonicalUrl":"https://cognitive-biases.github.io/compare/default-effect-vs-status-quo-bias/","resourceType":"comparison","section":"distinction","title":"Default Effect vs Status Quo Bias","text":"Both can make one option unusually sticky, but they are not the same. A default is the option or outcome that applies if no active choice is made. Status quo bias is the broader tendency to give extra weight to the option that is already current or in place. Ask why one option is privileged: because the choice architecture makes it happen automatically, or because it is already the person's current state?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"80c3f32dbdcf85af0d3b496e99b63e906dc35c6c6230ed7c423026a722339c2f"}
{"chunkId":"cb-framing-effect-vs-anchoring-effect-distinction","canonicalId":"framing-effect-vs-anchoring-effect","canonicalUrl":"https://cognitive-biases.github.io/compare/framing-effect-vs-anchoring-effect/","resourceType":"comparison","section":"distinction","title":"Framing Effect vs Anchoring Effect","text":"Both show that presentation can shape judgment, but the mechanism being tested is different. Framing changes how equivalent information or outcomes are described; anchoring tests how a starting numerical value pulls a later estimate toward it. Ask what changed: the description of the options, or the starting number used before the estimate?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"0d72ede035c81e9fca47725cdfae8bd453ba19b1ecec58c265e46977ea6b0f0a"}
{"chunkId":"cb-loss-aversion-vs-sunk-cost-effect-distinction","canonicalId":"loss-aversion-vs-sunk-cost-effect","canonicalUrl":"https://cognitive-biases.github.io/compare/loss-aversion-vs-sunk-cost-effect/","resourceType":"comparison","section":"distinction","title":"Loss Aversion vs Sunk Cost Effect","text":"Both can make stopping, switching or giving something up feel difficult, but they ask different questions. Loss aversion concerns how prospective losses are valued relative to comparable gains around a reference point. Sunk cost effect concerns whether irrecoverable past investment is influencing the current choice. Ask which information is doing the work: a possible loss from the current reference point, or resources that have already been spent and cannot be recovered?","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"77f1755d64d1c94aeb6d1373809c70811f6a873ecc5b1fb9c2f95277182d413e"}
{"chunkId":"cb-sunk-cost-effect-vs-escalation-of-commitment-distinction","canonicalId":"sunk-cost-effect-vs-escalation-of-commitment","canonicalUrl":"https://cognitive-biases.github.io/compare/sunk-cost-effect-vs-escalation-of-commitment/","resourceType":"comparison","section":"distinction","title":"Sunk Cost Effect vs Escalation of Commitment","text":"The two ideas overlap, but they answer different questions. Sunk cost effect asks whether irrecoverable past investment is influencing the current choice; escalation of commitment describes a broader pattern of persisting or investing more after setbacks. Ask whether you are identifying one reason for continuing, or the wider process of increasing commitment after bad results. Sunk costs can drive escalation, but escalation can have other drivers too.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"e69b706414f3fd537f9ac75301641e8025941d37f987b0165d8689327d648a5c"}
{"chunkId":"cb-surrogation-vs-systematic-bias-distinction","canonicalId":"surrogation-vs-systematic-bias","canonicalUrl":"https://cognitive-biases.github.io/compare/surrogation-vs-systematic-bias/","resourceType":"comparison","section":"distinction","title":"Surrogation vs Systematic Bias","text":"Both can make a metric misleading, but the failure is different. Surrogation happens when people start treating a performance measure as though it were the goal or construct itself. Systematic bias is a consistent or predictable deviation in a measurement, estimate, sample, or process relative to a target or reference. Ask whether the problem is that people are optimizing or interpreting the proxy as the goal, or that the measurement process itself is systematically shifted away from the target.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"b62bed0d1b5bfb7b0294909a945cb87c9e2ca3729aa42163654505a35fdb6803"}
{"chunkId":"cb-llm-cognitive-biases-what-recent-research-shows-summary","canonicalId":"llm-cognitive-biases-what-recent-research-shows","canonicalUrl":"https://cognitive-biases.github.io/research/llm-cognitive-biases-what-recent-research-shows/","resourceType":"research-note","section":"summary","title":"Cognitive bias in LLMs: what recent research actually shows","text":"Recent studies make a stronger case that LLM outputs can show repeatable decision patterns that resemble named cognitive biases. They also show why we should be careful with the label: the effects depend on the task, model, prompt and conversational context, and a debiasing prompt that helps one kind of task can hurt another.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"52be837db57c0ad8d68ac037b8e7d5037be7cf61236551835e8ad9621894c264"}
{"chunkId":"cb-planning-fallacy-project-estimates-outside-view-summary","canonicalId":"planning-fallacy-project-estimates-outside-view","canonicalUrl":"https://cognitive-biases.github.io/research/planning-fallacy-project-estimates-outside-view/","resourceType":"research-note","section":"summary","title":"Planning fallacy in project estimates: why one buffer is not the answer","text":"Research supports a recurring pattern of optimistic completion-time estimates. The useful response is to compare the plan with similar completed work and keep uncertainty visible, not to multiply every estimate by one fixed number.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"d0608d0097683c2a311bf4dd8d25fa126fccbcef319617ab7ed5ae07757666f0"}
{"chunkId":"cb-continued-influence-corrections-misinformation-summary","canonicalId":"continued-influence-corrections-misinformation","canonicalUrl":"https://cognitive-biases.github.io/research/continued-influence-corrections-misinformation/","resourceType":"research-note","section":"summary","title":"Why corrected misinformation can still affect reasoning","text":"Corrections usually improve factual accuracy, but old misinformation can still influence later reasoning. That continued influence is not the same as a backfire effect, and the distinction changes how corrections should be evaluated.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"baae6095673ecfbceeebdbfc77b709c789e72aee024791da9a4b90045e1c73dc"}
{"chunkId":"cb-anchoring-ai-assisted-decisions-first-number-summary","canonicalId":"anchoring-ai-assisted-decisions-first-number","canonicalUrl":"https://cognitive-biases.github.io/research/anchoring-ai-assisted-decisions-first-number/","resourceType":"research-note","section":"summary","title":"Anchoring in AI-assisted decisions: why the first number matters","text":"Anchoring is strongly supported in human numerical judgment, and newer studies show two AI-related risks: people can anchor on AI recommendations, while model outputs can also shift when prompts contain numerical anchors.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"c52883dcbb1264333a7efce286125b2f1d475b89987524985def43c367101877"}
{"chunkId":"cb-availability-heuristic-vivid-does-not-mean-common-summary","canonicalId":"availability-heuristic-vivid-does-not-mean-common","canonicalUrl":"https://cognitive-biases.github.io/research/availability-heuristic-vivid-does-not-mean-common/","resourceType":"research-note","section":"summary","title":"Availability heuristic: vivid does not automatically mean common","text":"Availability is a useful shortcut when accessible examples track real frequency. It becomes misleading when accessibility comes from non-diagnostic causes, and some popular claims about dramatic media risks and ease of retrieval are less universal than textbook examples suggest.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"f2d0428a0e68067f86807b73a1850faf9364f198940044de59c2f44358b56e65"}
{"chunkId":"cb-confirmation-bias-is-more-than-seeking-agreeable-information-summary","canonicalId":"confirmation-bias-is-more-than-seeking-agreeable-information","canonicalUrl":"https://cognitive-biases.github.io/research/confirmation-bias-is-more-than-seeking-agreeable-information/","resourceType":"research-note","section":"summary","title":"Confirmation bias is more than seeking agreeable information","text":"Confirmation bias is a broad family of belief-consistent information-processing tendencies. Classic hypothesis-testing work also shows why a narrower problem matters: a test can fit the favored hypothesis while offering little information against plausible alternatives.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"8df79f8ef712bdcbe0e1828f3811362e60d743fbfc761de2c1373eba2c72514a"}
{"chunkId":"cb-curse-of-knowledge-experts-beginners-feedback-summary","canonicalId":"curse-of-knowledge-experts-beginners-feedback","canonicalUrl":"https://cognitive-biases.github.io/research/curse-of-knowledge-experts-beginners-feedback/","resourceType":"research-note","section":"summary","title":"Curse of Knowledge: why expert clarity is a poor test of beginner understanding","text":"Once you know an answer, your own knowledge can distort estimates of what another person knows. The most useful correction is not 'explain more simply' in the abstract, but feedback from the less-informed person's actual perspective.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"400a8ffe7aece2ad21bb45acb618fcac227d0679a46c5b8fdb49391e23cd9926"}
{"chunkId":"cb-why-the-past-can-look-better-than-it-was-summary","canonicalId":"why-the-past-can-look-better-than-it-was","canonicalUrl":"https://cognitive-biases.github.io/research/why-the-past-can-look-better-than-it-was/","resourceType":"research-note","section":"summary","title":"Why the past can look better than it was","text":"Several different processes can make past periods look better from the present, but they should not be collapsed into one universal bias. Event recollection, changing emotional intensity, exposure to current negative information, and real historical trends need separate evidence.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"cc309fd96a63b52c405d424f621e5ddfabab587396835c30f2a2be46fb3ed02c"}
{"chunkId":"cb-decoy-effect-pricing-choice-sets-summary","canonicalId":"decoy-effect-pricing-choice-sets","canonicalUrl":"https://cognitive-biases.github.io/research/decoy-effect-pricing-choice-sets/","resourceType":"research-note","section":"summary","title":"Decoy Effect in pricing: when a third option changes the comparison","text":"Adding an inferior option can change the choice between stronger alternatives, but not every three-tier pricing table creates a reliable decoy effect. The position of the options, prior preferences and the decision context all matter.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"45f2176d2461ed6333af0d36b6bc5f6c70373dd0210fee4f2f3007cb3485b083"}
{"chunkId":"cb-defaults-status-quo-preference-choice-architecture-summary","canonicalId":"defaults-status-quo-preference-choice-architecture","canonicalUrl":"https://cognitive-biases.github.io/research/defaults-status-quo-preference-choice-architecture/","resourceType":"research-note","section":"summary","title":"Defaults and status quo: why staying does not prove preference","text":"Defaults often change what people choose, but staying with a pre-selected or existing option does not by itself show strong preference. The useful question is what made the option sticky and whether the result still serves the person's goals.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"8b52b1510c5f9afc5fb1f59e0a0eda87178308e0bf3f4c6a19761c97fa386df3"}
{"chunkId":"cb-dunning-kruger-effect-what-research-shows-summary","canonicalId":"dunning-kruger-effect-what-research-shows","canonicalUrl":"https://cognitive-biases.github.io/research/dunning-kruger-effect-what-research-shows/","resourceType":"research-note","section":"summary","title":"Dunning–Kruger Effect: what the research does and does not show","text":"The Dunning–Kruger Effect is better understood as a question about calibration between performance and self-assessment than as the internet rule that incompetent people are always extremely confident.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"51900da1023571aaa892347ae0832f706a8bff7a3b90c8fb9500720c4b44cf4b"}
{"chunkId":"cb-framing-effect-not-all-frames-are-the-same-summary","canonicalId":"framing-effect-not-all-frames-are-the-same","canonicalUrl":"https://cognitive-biases.github.io/research/framing-effect-not-all-frames-are-the-same/","resourceType":"research-note","section":"summary","title":"Framing effect: why not all frames are the same","text":"Risky-choice framing is a robust finding, but the word 'framing' covers several different research paradigms. Treating every positive-versus-negative message as the same effect hides important differences in evidence and mechanism.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"f556ae9cef544fa7e2a1fc42036b6200a31486663fc5fffa86ab5f0297575001"}
{"chunkId":"cb-hungry-judge-effect-what-study-actually-shows-summary","canonicalId":"hungry-judge-effect-what-study-actually-shows","canonicalUrl":"https://cognitive-biases.github.io/research/hungry-judge-effect-what-study-actually-shows/","resourceType":"research-note","section":"summary","title":"Hungry Judge Effect: what the famous study actually shows","text":"The famous parole-board study found a strong sequence pattern around food breaks, but it did not measure hunger or randomly manipulate meals. Later critiques and simulations provide plausible alternative explanations for part of the striking result.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"1eba10c151bf87c17e26829a401a72d7f523c3e6d79bb96914e7d0d5c41ee1dd"}
{"chunkId":"cb-loss-aversion-how-large-and-how-robust-summary","canonicalId":"loss-aversion-how-large-and-how-robust","canonicalUrl":"https://cognitive-biases.github.io/research/loss-aversion-how-large-and-how-robust/","resourceType":"research-note","section":"summary","title":"Loss Aversion: how large is the effect, and how robust is it?","text":"Loss aversion is one of the best-known ideas in behavioural economics, but recent meta-analyses disagree substantially about its average size and about which experimental designs provide clean evidence for it.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"13c825feb2de8c1356df08b88716297af5fc8783f2282c57919267e6b9a95e6c"}
{"chunkId":"cb-sunk-cost-escalation-why-past-spend-is-only-part-summary","canonicalId":"sunk-cost-escalation-why-past-spend-is-only-part","canonicalUrl":"https://cognitive-biases.github.io/research/sunk-cost-escalation-why-past-spend-is-only-part/","resourceType":"research-note","section":"summary","title":"Sunk cost vs escalation: why past spending is only part of the story","text":"Past, unrecoverable costs can influence whether people continue, but escalation of commitment is a broader process. Responsibility for the original choice, completion pressure, project structure and advice can also shape the next investment.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"ef0f5853fe5af78d748bcc489a320560f3ae77a266731e2a2d8745f3d9ce6b6c"}
{"chunkId":"cb-surrogation-when-a-kpi-starts-replacing-the-goal-summary","canonicalId":"surrogation-when-a-kpi-starts-replacing-the-goal","canonicalUrl":"https://cognitive-biases.github.io/research/surrogation-when-a-kpi-starts-replacing-the-goal/","resourceType":"research-note","section":"summary","title":"Surrogation: when a KPI starts replacing the goal","text":"Surrogation is more specific than the slogan 'what gets measured gets managed.' Research in strategic performance systems shows how a measure can start to stand in for the construct it was designed to represent, and how decision processes can reduce that substitution.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"281faf043c58e9590ca83d011414ef77e2b1f6755ca21ebf51e0c54d4367ac2b"}
{"chunkId":"cb-systematic-bias-why-more-data-is-not-enough-summary","canonicalId":"systematic-bias-why-more-data-is-not-enough","canonicalUrl":"https://cognitive-biases.github.io/research/systematic-bias-why-more-data-is-not-enough/","resourceType":"research-note","section":"summary","title":"Systematic Bias: why more data is not enough","text":"Systematic bias is not one mental shortcut. It is a measurement and statistical problem: a process can keep pushing results away from a target in a consistent direction, and collecting more data from the same biased process does not automatically fix it.","reviewState":"reviewed","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"cfe799f8c60a8a87ff94235490932f94372708256aebd7e89d05eb67f7eefd66"}
{"chunkId":"cb-skill-evidence-evaluation-guide","canonicalId":"evidence-evaluation","canonicalUrl":"https://cognitive-biases.github.io/skills/evidence-evaluation/","resourceType":"skill","section":"guide","title":"Evidence evaluation","text":"Judge whether a claim deserves confidence before repetition, fluency or preference turns into evidence. Learning outcome: You can separate a persuasive claim from the evidence that should change your mind. Use when: A team already prefers one explanation and is gathering support for it. Use when: A claim appears in many places but may trace back to one source. Use when: A vivid example is dominating broader evidence. Use when: An AI answer sounds convincing but important claims are still unverified. Action: Write what evidence would weaken the preferred conclusion before searching for more support. Action: Trace repeated claims back to genuinely independent sources. Action: Compare vivid examples with a base rate, reference class or broader sample. Action: Keep verified facts, plausible inferences and open questions separate. Reviewed lens: Confirmation Bias. Confirmation bias is an umbrella label for several ways existing beliefs or hypotheses can influence information search and interpretation. It should not be reduced to one behaviour such as reading only agreeable news, and a preference for confirming tests is not irrational in every task or environment. Reviewed lens: Illusory Truth Effect. Repeated information is, on average, judged as more truthful than comparable new information. The size of the effect varies with the material and procedure, and repetition does not make every claim believable or erase all prior knowledge. Reviewed lens: Availability Heuristic. Availability is a judgment heuristic: people can use how easily examples or scenarios come to mind when estimating frequency or probability. That shortcut is not automatically an error because memorable or accessible examples can correlate with real frequency. Bias appears when accessibility is driven by factors that are not diagnostic of the quantity being judged. Reviewed lens: Continued Influence Effect. The continued influence effect describes a specific pattern: false or outdated information can keep affecting later reasoning even after a clear correction. It does not mean the correction had no effect, that the person rejected the correction, or that belief in the false claim became stronger. Corrections often reduce misinformation reliance without removing it completely. That distinction matters because continued influence is much broader than the rarer backfire outcome.","reviewState":"reviewed-links","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"ebbfc299ea678258dc475cf95a0a9edcc0eac8d9291826f9f106b7bceb20d9a8"}
{"chunkId":"cb-skill-decision-making-under-uncertainty-guide","canonicalId":"decision-making-under-uncertainty","canonicalUrl":"https://cognitive-biases.github.io/skills/decision-making-under-uncertainty/","resourceType":"skill","section":"guide","title":"Decision making under uncertainty","text":"Make useful choices when outcomes are not known and one number, story or recent event can pull judgment too strongly. Learning outcome: You can make a decision without pretending uncertainty has disappeared. Use when: A deadline, probability or budget is being discussed as one precise number. Use when: An early estimate is shaping every later estimate. Use when: The latest success or failure is dominating the next decision. Use when: The team is continuing mainly because it has already invested a lot. Action: Start with a range or reference class before negotiating a precise number. Action: Create an independent estimate before looking at a strong anchor when possible. Action: Separate already-spent resources from the future cost and value of the next step. Action: Record assumptions that would make you reopen the decision. Reviewed lens: Anchoring Effect. Anchoring is a well-supported effect in which an initial numerical value can pull a later estimate toward it. A 2026 meta-analysis covering 2,601 effect sizes found a large overall effect, but also substantial variation across studies. The effect should not be treated as a rule that every number changes every judgment: incidental anchors, anchors from a different dimension, clearly random values, incentives, and some debiasing conditions were associated with smaller or null effects. Reviewed lens: Subadditivity Effect. People often give a larger total probability when an event is unpacked into separate possibilities than when the same event is judged as one packed category. The effect is not universal: how the possibilities are described and how typical they are can change or even reverse an unpacking effect. Reviewed lens: Availability Heuristic. Availability is a judgment heuristic: people can use how easily examples or scenarios come to mind when estimating frequency or probability. That shortcut is not automatically an error because memorable or accessible examples can correlate with real frequency. Bias appears when accessibility is driven by factors that are not diagnostic of the quantity being judged. Reviewed lens: Escalation of Commitment. Sunk-cost effects and escalation of commitment overlap but are not interchangeable. Sunk-cost research asks whether irrecoverable prior investments influence current choices. Escalation of commitment describes persistence or additional resource allocation to a failing course of action and can also be driven by personal responsibility, self-justification, project structure, and other factors.","reviewState":"reviewed-links","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"c21f20464d7caf9d782041caae17222a5bc05290978d7093a69db417fd0e0b09"}
{"chunkId":"cb-skill-forecasting-guide","canonicalId":"forecasting","canonicalUrl":"https://cognitive-biases.github.io/skills/forecasting/","resourceType":"skill","section":"guide","title":"Forecasting","text":"Estimate what may happen next by using comparable evidence, explicit uncertainty and records made before the outcome is known. Learning outcome: You can produce forecasts that are easier to challenge, compare and learn from later. Use when: A project deadline is being built only from an inside plan. Use when: A current feeling or preference is being projected far into the future. Use when: A vivid recent example is being used as the main probability estimate. Use when: A finished outcome now feels as if it had always been obvious. Action: Check comparable completed cases before refining the current plan. Action: State a range and the factors that would push the result toward either end. Action: Store the forecast and assumptions before the outcome. Action: Review forecast quality against the original record, not reconstructed memory. Reviewed lens: Planning Fallacy. The planning fallacy is a well-documented tendency for people to predict their own task completion times too optimistically. The effect has been observed across different kinds of tasks, but it is not a rule that every plan will run late. Project overruns can also come from changing scope, dependencies, incentives, poor data, deliberate underestimation, or genuinely unusual events, so a late project should not automatically be diagnosed as a planning fallacy. Reviewed lens: Anchoring Effect. Anchoring is a well-supported effect in which an initial numerical value can pull a later estimate toward it. A 2026 meta-analysis covering 2,601 effect sizes found a large overall effect, but also substantial variation across studies. The effect should not be treated as a rule that every number changes every judgment: incidental anchors, anchors from a different dimension, clearly random values, incentives, and some debiasing conditions were associated with smaller or null effects. Reviewed lens: Availability Heuristic. Availability is a judgment heuristic: people can use how easily examples or scenarios come to mind when estimating frequency or probability. That shortcut is not automatically an error because memorable or accessible examples can correlate with real frequency. Bias appears when accessibility is driven by factors that are not diagnostic of the quantity being judged. Reviewed lens: Hindsight Bias. Knowing an outcome can make the outcome look more predictable in retrospect. The effect has been studied for decades and across many settings, but it does not mean that every confident explanation after an event is biased.","reviewState":"reviewed-links","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"f0811d12eced34dbfb1163312d573f28b501641f4aefaf6b7be447802352d7a1"}
{"chunkId":"cb-skill-metacognition-guide","canonicalId":"metacognition","canonicalUrl":"https://cognitive-biases.github.io/skills/metacognition/","resourceType":"skill","section":"guide","title":"Metacognition","text":"Inspect how a judgment was formed, what influenced it and what would make you change it. Learning outcome: You can review your reasoning process without confusing confidence, outcome and hindsight with decision quality. Use when: A good result is being treated as proof that the original decision process was good. Use when: A bad result makes the warning signs seem obvious in retrospect. Use when: You are mainly testing arguments that support a preferred answer. Use when: You need to learn from a past decision without rewriting what was knowable at the time. Action: Freeze what was actually known before the outcome. Action: Judge decision-process quality separately from the eventual result. Action: Write what would change your mind before collecting more evidence. Action: Compare the review with the original forecast, notes or assumptions. Reviewed lens: Outcome Bias. Outcome bias occurs when knowledge of a result changes how people evaluate the quality of a decision even when the information available at the time of the decision is held constant. Outcomes can still be relevant for learning, so the error is not 'never look at results'; it is using luck or hindsight as if it had been available to the original decision-maker. Reviewed lens: Hindsight Bias. Knowing an outcome can make the outcome look more predictable in retrospect. The effect has been studied for decades and across many settings, but it does not mean that every confident explanation after an event is biased. Reviewed lens: Confirmation Bias. Confirmation bias is an umbrella label for several ways existing beliefs or hypotheses can influence information search and interpretation. It should not be reduced to one behaviour such as reading only agreeable news, and a preference for confirming tests is not irrational in every task or environment. Reviewed lens: Moral Luck. Resultant moral luck describes cases where judgments of blame, punishment, or moral evaluation differ because otherwise similar actions lead to different outcomes partly outside the agent's control. Outcome information does affect moral judgment in experiments, but the effect should not be reduced to 'people ignore intent.' Mental states, causal responsibility, belief justification, negligence, and the kind of moral judgment being asked about all matter. Some studies find that false or unjustified beliefs explain more of classic moral-luck asymmetries than the bad outcome itself, while still detecting an independent outcome effect.","reviewState":"reviewed-links","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"a4de01c6e20a8012587e7464fcb992473282752abf6b1962e9ab946cbca0ec8b"}
{"chunkId":"cb-skill-information-verification-guide","canonicalId":"information-verification","canonicalUrl":"https://cognitive-biases.github.io/skills/information-verification/","resourceType":"skill","section":"guide","title":"Information verification","text":"Check where information came from, whether sources are independent and whether a correction has really replaced the old explanation. Learning outcome: You can verify important claims without treating repetition as independent confirmation. Use when: A claim feels familiar because it has appeared repeatedly. Use when: Several summaries may all depend on the same original source. Use when: A correction was issued but the old story still shapes later reasoning. Use when: Generated answers repeat a claim without showing where it came from. Action: Trace the claim to the earliest reliable or primary source you can inspect. Action: Count independent sources, not repetitions. Action: After a correction, restate the supported replacement explanation explicitly. Action: Mark claims as verified, unverified or disputed instead of flattening them into one confidence level. Reviewed lens: Illusory Truth Effect. Repeated information is, on average, judged as more truthful than comparable new information. The size of the effect varies with the material and procedure, and repetition does not make every claim believable or erase all prior knowledge. Reviewed lens: Continued Influence Effect. The continued influence effect describes a specific pattern: false or outdated information can keep affecting later reasoning even after a clear correction. It does not mean the correction had no effect, that the person rejected the correction, or that belief in the false claim became stronger. Corrections often reduce misinformation reliance without removing it completely. That distinction matters because continued influence is much broader than the rarer backfire outcome. Reviewed lens: Confirmation Bias. Confirmation bias is an umbrella label for several ways existing beliefs or hypotheses can influence information search and interpretation. It should not be reduced to one behaviour such as reading only agreeable news, and a preference for confirming tests is not irrational in every task or environment. Reviewed lens: Availability Heuristic. Availability is a judgment heuristic: people can use how easily examples or scenarios come to mind when estimating frequency or probability. That shortcut is not automatically an error because memorable or accessible examples can correlate with real frequency. Bias appears when accessibility is driven by factors that are not diagnostic of the quantity being judged.","reviewState":"reviewed-links","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"80255e8083204d37c615e47f30c7371ca1dfa5209649e1675c85e8591a06a334"}
{"chunkId":"cb-skill-ai-assisted-reasoning-guide","canonicalId":"ai-assisted-reasoning","canonicalUrl":"https://cognitive-biases.github.io/skills/ai-assisted-reasoning/","resourceType":"skill","section":"guide","title":"AI-assisted reasoning","text":"Use AI as a decision aid without confusing fluent output, humanlike presentation or repeated generated claims with verified capability and evidence. Learning outcome: You can use an AI system to expand reasoning while keeping important judgments independently checkable. Use when: A chatbot recommendation is influencing an important decision. Use when: The model gives the first number, score or estimate in the discussion. Use when: A polished conversational style makes the system feel more capable than you have tested. Use when: You are using AI mainly to strengthen a conclusion you already prefer. Action: Write a provisional judgment or success criterion before asking AI for advice when the decision matters. Action: Separate interface quality from the capability needed for the task. Action: Verify consequential claims against an independent source rather than another generated answer. Action: Ask for counterevidence and alternative explanations before finalizing the decision. Reviewed lens: Automation Bias. Automation bias is a documented pattern of inappropriate reliance on automated cues or recommendations. It can produce commission errors when a user follows incorrect advice and omission errors when a user fails to act because automation did not signal a problem. This does not mean automation is generally harmful: decision support can improve overall performance, and the relevant question is whether reliance remains calibrated when the system is wrong, incomplete, or difficult to verify. Reviewed lens: Anchoring Effect. Anchoring is a well-supported effect in which an initial numerical value can pull a later estimate toward it. A 2026 meta-analysis covering 2,601 effect sizes found a large overall effect, but also substantial variation across studies. The effect should not be treated as a rule that every number changes every judgment: incidental anchors, anchors from a different dimension, clearly random values, incentives, and some debiasing conditions were associated with smaller or null effects. Reviewed lens: Anthropomorphism. Anthropomorphism is the attribution of humanlike properties, intentions, emotions, or mental states to nonhuman agents. It is a well-established psychological phenomenon, but it is not automatically a cognitive error: humanlike models can sometimes be useful. The risk appears when humanlike cues are treated as evidence for capabilities, understanding, accuracy, consciousness, or motives that have not actually been demonstrated. Reviewed lens: Appearance–Capability Expectation. Research supports the broader pattern that a robot's appearance, morphology, framing, and human-likeness shape expectations about its competence, social qualities, and likely behavior. However, 'Form-Function Attribution Bias' is not an established standardized name in the literature. On this site it should be treated as a project label for appearance-driven capability expectations, not as a universally recognized cognitive-bias construct. Reviewed lens: Confirmation Bias. Confirmation bias is an umbrella label for several ways existing beliefs or hypotheses can influence information search and interpretation. It should not be reduced to one behaviour such as reading only agreeable news, and a preference for confirming tests is not irrational in every task or environment. Reviewed lens: Illusory Truth Effect. Repeated information is, on average, judged as more truthful than comparable new information. The size of the effect varies with the material and procedure, and repetition does not make every claim believable or erase all prior knowledge.","reviewState":"reviewed-links","sourceIds":[],"releaseVersion":"2026.08.18","schemaVersion":"1.0.0","contentHash":"04d73674b30f83c9422d95a1fcf244ace6ba77a0ee638e57a9fe1467fb9c87aa"}