Confirmation Bias
A reader's summary of the tendency to search for, favor, and recall information that confirms what you already believe — Peter Wason's 1960 rule-discovery experiments, Francis Bacon's 1620 description of the same pattern three centuries earlier, and why showing someone the other side of an argument so often backfires.
The effect at a glance
Confirmation bias is the tendency to search for, interpret, and recall information in ways that support a belief you already hold, while giving disproportionately less weight to information that would count against it. It doesn't require dishonesty or even awareness — the selection happens upstream of any deliberate decision about what to believe, which is what makes it a standing finding in cognitive psychology rather than just a rhetorical accusation to throw at an opponent.
Origin
The clearest early lab demonstration is Peter Wason's 1960 rule-discovery task. Participants saw the number triple 2-4-6, were told it satisfied a rule, and had to discover the rule by proposing their own triples and receiving yes-or-no feedback before stating their answer. The actual rule was simply “any three ascending numbers,” but most participants formed a narrower hypothesis early on, such as “numbers increasing by two,” and then kept testing triples consistent with that guess instead of triples designed to falsify it. A large share announced a confident, wrong rule without ever having tried to break their own hypothesis. Wason described the underlying tendency; the term “confirmation bias” itself came into standard use through the psychology literature that followed.
History and context
The pattern was described long before it was measured. Francis Bacon's 1620 Novum Organum put it almost exactly in modern terms: “the human understanding, when it has once adopted an opinion... draws all things else to support and agree with it.” Wason's 1960 task and his related 1968 four-card selection task turned that observation into something measurable under controlled conditions in the twentieth century. The next major step came from social psychology rather than logic puzzles: Charles Lord, Lee Ross, and Mark Lepper's 1979 study on attitudes toward capital punishment showed the bias operating on interpretation, not just search — both sides of a real, emotionally loaded disagreement read the identical mixed evidence as vindicating their own side. By 1998, Raymond Nickerson's review article had gathered decades of findings from medicine, law, science, and politics under the single “confirmation bias” heading that's now the common reference point.
Main ideas
It bundles three separate mechanisms
"Confirmation bias" is really shorthand for at least three distinguishable tendencies: biased search (seeking out evidence that supports what you already think), biased interpretation (reading ambiguous evidence as more supportive than it is), and biased memory (recalling confirming instances more easily than disconfirming ones) — treated as one effect in casual use but studied somewhat separately in the literature.
Wason's task caught biased search directly
In his 1960 experiment, participants were shown the triple 2-4-6, told it fit a rule, and asked to discover the rule by proposing further triples and getting yes/no feedback. Most fixated on an early hypothesis like "ascending by two" and kept proposing triples that fit it rather than triples designed to break it, so many announced a wrong rule with real confidence — a case of failing to seek out disconfirmation, not of misreading it.
Mixed evidence can polarize rather than resolve
Charles Lord, Lee Ross, and Mark Lepper's 1979 study gave people who already held opposing views on capital punishment the same two studies, one supporting deterrence and one undermining it. Both sides rated the study agreeing with their prior view as more methodologically sound and came away more confident in their original position — mixed evidence moved each side further apart rather than toward agreement.
It runs harder on identity-linked beliefs
The effect shows up on trivia and abstract rule-guessing tasks, but it's consistently stronger for beliefs tangled up with self-image, group identity, or prior investment — which is why it overlaps heavily with, but isn't identical to, the broader idea of motivated reasoning.
Nickerson's 1998 review made it the standard frame
Raymond Nickerson's paper "Confirmation Bias: A Ubiquitous Phenomenon in Many Guises" synthesized decades of scattered findings across psychology, medicine, law, and science into one umbrella account, and is the reference point most later work, including popular coverage, cites back to.
A feed optimized for engagement rewards exactly this
Content you already agree with is faster to process, more comfortable to sit with, and more likely to get a reaction — all signals a ranking algorithm built to maximize time-on-platform and interaction will pick up on and reinforce, whether or not it was designed with confirmation bias in mind.
Critique
- It's become a debate-ending accusation. Telling an opponent they're “just showing confirmation bias” is now a common way to dismiss a claim without engaging its content, even though the bias, if real, would apply to the accuser's own reasoning just as much as to the target's.
- The three mechanisms don't replicate equally well. Biased search is the most robustly demonstrated component; biased memory for confirming versus disconfirming information has produced a more mixed and contested body of results, which the umbrella term tends to flatten.
- It overlaps with, but isn't, motivated reasoning. Confirmation bias is often described as a largely automatic search-and-weighting problem, while motivated reasoning adds a directional goal — arriving at a preferred conclusion. Popular usage frequently treats the two as interchangeable when the underlying literature keeps them conceptually distinct.
- Blaming it entirely on algorithms overstates the case. How much of measured political polarization comes from platform ranking specifically, versus from confirmation bias operating on choices people would make anyway, is a genuinely disputed empirical question — see Echo Chambers and Filter Bubbles for what the data on that split actually shows.
Impact
The term is now a standard citation in science communication, misinformation research, and journalism explaining why corrections so often fail to change minds — see The Backfire Effect for the narrower and more contested claim that a correction can actively deepen the original belief. It also supplies part of the mechanism behind the Illusory Truth Effect, since a claim that already matches what you believe gets the easiest, most fluent processing of all. Outside psychology, it underpins standard advice in debate training, journalism, and forecasting to actively seek out the strongest version of the opposing case, on the theory that the ordinary search process won't surface it on its own.