Confirmation bias

Confirmation bias is the tendency to seek, interpret, and remember information in ways that confirm existing beliefs and hypotheses while ignoring or undervaluing contradicting information. The bias operates across three distinguishable components that often co-occur: biased information search (selectively seeking evidence that supports prior belief rather than evidence that would test it), biased interpretation (treating ambiguous evidence as supporting the held position), and biased memory (better encoding and recall of belief-consistent information). Confirmation bias is one of the most-replicated findings in cognitive psychology, operating independently of expertise, intelligence, and material stakes.

The contemporary empirical literature begins with Peter Wason’s (1960) 2-4-6 task in the Quarterly Journal of Experimental Psychology, in which participants tasked with discovering a rule generating number sequences systematically tested confirming examples rather than potentially falsifying ones — finding the simple “ascending by 2” hypothesis but failing to discover the actual rule (“any ascending sequence”). Nickerson (1998) reviewed several decades of follow-up work in Review of General Psychology; Klayman and Ha (1987) clarified that the underlying tendency is better described as a positive-test strategy than as confirmation per se. The bias overlaps with related constructs including myside bias (Stanovich, West & Toplak 2013) in political-reasoning research and motivated reasoning (Kunda 1990) in social-cognition research.

Three points are routinely missed in popular treatments. First, confirmation bias is not a flaw of stupidity or dishonesty; it appears in all reasoners and is structurally amplified in information environments that allow active selection (search engines, social media feeds). Second, intelligence and expertise do not protect against it — in some paradigms, more knowledgeable participants generate more sophisticated confirming arguments rather than less. Third, the reliable interventions are structural — adversarial review, pre-registered hypotheses, devil’s-advocate roles — rather than introspective resolve.

Confirmation bias is the bias most directly opposed to the scientific method. Science depends on actively seeking evidence that could falsify a hypothesis; confirmation bias inclines people to seek evidence that would confirm it. The bias scales from individual reasoning to institutional decision-making to political discourse to information consumption. In medicine, it contributes to diagnostic errors. In law, it contributes to wrongful convictions. In science, it contributes to publication bias and the replication crisis.

The bias is amplified by environments where evidence is selected by the seeker rather than presented adversarially — a description that fits much of the modern internet. A 2025 review in the Annals of the New York Academy of Sciences documented how generative AI tools further extend this dynamic: conversational AI systems can be steered toward confirming responses through subtle phrasing, and the private nature of these interactions limits the corrective effect of public disagreement (Lopez-Lopez et al., 2025). Social platforms add to this, although studies disagree on how much feed algorithms themselves narrow what people see.

For individuals, recognizing the patterns is the foundation of better epistemic practice — the ability to update beliefs in response to evidence rather than to one's prior commitments.

Peter Wason's four-card selection task (Wason 1966, 1968) is the best-known experimental demonstration. Participants were given four cards showing letters and numbers (such as E, K, 4, 7) and a hypothesis to test ("If a card has a vowel on one side, it has an even number on the other"). They were asked which cards needed to be turned over to test the hypothesis. Most people chose the E, which is correct because an odd number on its back would break the rule, and often the 4, which cannot break the rule, but rarely the 7, which would break it if it had a vowel on the other side.

Wason's earlier 2-4-6 task (1960) was the first. Participants given a numerical sequence and asked to discover the rule generating it consistently tested only sequences they expected to fit, missing the broader rule. Klayman and Ha (1987) reframed the phenomenon as positive test strategy — a more general tendency to seek information that would arise if the hypothesis were true. Earlier philosophical roots trace to Francis Bacon's Novum Organum (1620), which warned of the human mind's tendency to support its first conclusions.

Mercier and Sperber's (2011) argumentative theory proposes that confirmation bias evolved not as a flaw in individual reasoning but as a feature of group reasoning — useful when conclusions are tested adversarially in dialogue, problematic when reasoning is solitary. This framing helps explain why peer review, devil's advocacy, and structured disagreement are unusually effective debiasing strategies: they restore the adversarial context in which the bias was originally adaptive.

Reviews such as Nickerson's (1998) describe confirmation bias in how people search for evidence, how they interpret it and what they remember. Each requires different mitigation strategies.

• Selective search. The mind preferentially seeks information that supports its current view. People consult sources they expect to agree with, ask questions whose answers they expect to like, and pursue tests that would confirm rather than falsify a hypothesis. Mitigation: deliberate consultation of opposing views, structured pre-mortems, formal devil's advocacy.
• Selective interpretation. Even when contradictory evidence is encountered, it is processed differently from confirming evidence. Identical studies are judged more rigorous when their conclusions match the reader's prior view. Critics are seen as biased; allies are seen as objective. Mitigation: blinded evaluation, criteria pre-specification, third-party review.
• Selective memory. Information that confirms beliefs is recalled more readily and more vividly than disconfirming information, which fades. Over time, this produces the illusion of having always known one was right. Mitigation: contemporaneous decision logs, pre-registered predictions, calibration tracking.

The bias is amplified when the hypothesis is identity-relevant: people are particularly prone to seek confirming evidence for beliefs that are central to their political identity, professional self-concept, or personal worldview. People are more likely to believe headlines that fit their politics, but Pennycook and Rand (2019) found that people who think more carefully are better at telling true headlines from false ones whether or not the headlines fit their politics.

What it can do. Recognizing confirmation bias provides a practical framework for designing better personal and institutional epistemic practices. Pre-registering predictions, seeking out the strongest version of an opposing argument before forming a view, and structurally separating the role of evidence-gathering from the role of evidence-evaluation all reduce the bias's impact. A 2024 study in Frontiers in Public Health found that explicit awareness of confirmation bias mechanisms measurably increased resilience to misinformation in test populations (Piksa et al., 2024).

What it can't do. Awareness alone is insufficient. The bias is robust to expertise, intelligence, and stakes. Researchers, judges, and clinicians who care deeply about getting the right answer still exhibit it. The structural practices that work — pre-registration, blind review, devil's advocacy, deliberate consideration of disconfirming evidence — work because they remove the bias from the loop, not because they make individuals immune to it.

"Confirmation bias is the same as motivated reasoning." The two overlap but are distinct. Confirmation bias operates even when the seeker has no emotional stake in the answer; it reflects a general cognitive default toward positive testing. Motivated reasoning specifically describes biased reasoning driven by what the person wants to be true. Both contribute to the same observed patterns, but the mechanisms are partly different.

"Smart or well-educated people are immune to confirmation bias." They are not. The bias is robust to expertise. Higher cognitive ability is associated with stronger arguments for one's existing position — sometimes called the "smart-people-have-better-rationalizations" effect — but is not associated with reduced susceptibility to the underlying bias.

"You can fix confirmation bias by being open-minded." Knowing about the bias and wanting to be fair do not by themselves remove it. Safeguards built into the process, such as writing predictions down in advance, blind review and structured devil's advocacy, are often recommended because they do not depend on noticing the bias in the moment.

"Confirmation bias only matters in politics." The bias operates in every domain where evidence is selected and weighed: medical diagnosis, hiring decisions, scientific research, legal proceedings, financial analysis, personal relationships. Political discourse is where the bias is most visible, but not where its most consequential effects occur.

Consider a clinician evaluating a patient with abdominal pain. After hearing the initial history, the clinician forms a preliminary hypothesis — perhaps appendicitis. The selective-search mechanism now activates: the clinician asks questions whose answers will help confirm appendicitis (location of tenderness, fever, nausea), and may not ask questions that would point elsewhere (gynecological history, recent travel, bowel habit changes). The selective-interpretation mechanism activates next: ambiguous findings are interpreted as consistent with the working hypothesis. If the patient's pain has shifted location — a finding that could weaken the appendicitis hypothesis or strengthen alternatives — the clinician may unconsciously dismiss the shift.

The structural protection is not "be a more open-minded clinician." It is the diagnostic checklist, the differential diagnosis discipline of explicitly listing alternative explanations and seeking evidence for and against each, and the second-clinician review when the initial diagnosis is uncertain. These practices restore the adversarial context that solitary reasoning lacks. Awareness of confirmation bias makes clinicians more likely to use these structures; it does not, by itself, make them less biased.

• Cognitive bias — the broader category of systematic deviations from rational judgment. Confirmation bias is the most-studied example.
• Heuristic — the mental shortcuts from which biases including confirmation bias emerge. Positive testing is the heuristic; confirmation bias is the predictable error.
• Bias blind spot — the meta-bias of seeing confirmation bias clearly in others while missing it in oneself.
• Decision hygiene — the structural practices (pre-mortems, blind review, devil's advocacy) that mitigate confirmation bias and other systematic errors.
• Motivated reasoning — biased reasoning driven by what the person wants to be true. Overlaps with confirmation bias but is mechanistically distinct.
• Echo chambers and filter bubbles — information environments that may reinforce confirmation bias by showing people mostly content that matches their existing views; how strong these effects are online is debated.
• Falsificationism — Karl Popper's principle that scientific hypotheses must be testable in ways that could prove them wrong. The methodological response to confirmation bias.
• Negativity bias — the two biases both shape how people take in social and political information, though how they interact is not well established.
• Survivorship bias — both biases produce overconfidence in conclusions drawn from incomplete information; confirmation bias by ignoring disconfirming evidence, survivorship bias by ignoring failed cases.

The LifeByLogic Cognitive Bias Test does not measure confirmation bias. It scores seven other biases from short decision tasks (anchoring, hindsight bias, overconfidence, belief bias, base-rate neglect, sunk cost and outcome bias) and compares those scores with twelve self-rating statements to give a Bias Blind Spot. The closest of its tasks to this entry is belief bias: judging an argument by whether its conclusion sounds believable rather than whether it follows. The questions are LifeByLogic's own, and the test has not been normed or validated, so its scores are not percentiles. The full methodology is documented on the tool methodology page.

Take the Cognitive Bias Test →

This entry is intended as a citable scholarly reference. Choose the format that matches your context. The retrieval date should reflect when you accessed the page, which may differ from the entry's last-reviewed date shown above.

LifeByLogic. (2026). Confirmation Bias: Definition, Examples and How to Reduce It. https://lifebylogic.com/glossary/confirmation-bias/

LifeByLogic. "Confirmation Bias: Definition, Examples and How to Reduce It." LifeByLogic, 25 Oct. 2026, https://lifebylogic.com/glossary/confirmation-bias/.

LifeByLogic. 2026. "Confirmation Bias: Definition, Examples and How to Reduce It." October 25. https://lifebylogic.com/glossary/confirmation-bias/.

@misc{lblconfirmationbias2026,
  author = {{LifeByLogic}},
  title = {Confirmation Bias: Definition, Examples and How to Reduce It},
  year = {2026},
  month = {oct},
  publisher = {LifeByLogic},
  url = {https://lifebylogic.com/glossary/confirmation-bias/},
  note = {Accessed: 2026-10-25}
}