Confirmation Bias

Confirmation bias is not a single failure but a family of related tendencies documented across decades of research (Nickerson, 1998). It appears in three main forms: biased search (preferring sources and questions likely to confirm what we already believe), biased interpretation (reading ambiguous evidence as supportive of prior beliefs), and biased memory (recalling supporting instances more readily than disconfirming ones). Wason's (1960) rule-discovery task offered an early classic demonstration: participants asked to discover a numerical rule almost always tested cases that fit their current guess rather than cases that could refute it. The tendency to seek confirmation rather than disconfirmation is deeply built into ordinary reasoning.

Confirmation bias persists because it is efficient. Testing every belief against every possible counter-example would be paralyzing. Prior beliefs act as priors in a Bayesian sense; the mind updates them incrementally rather than overturning them at every new datum. But the mind is not a good Bayesian without help. In socially charged domains — politics, identity, faith, in-group loyalty — confirmation bias is amplified by motivated reasoning: the desire for a conclusion actively shapes which evidence is admitted and how it is weighed (Kunda, 1990). This is why arguments about facts often fail to change minds. The problem is not lack of information; it is that the incoming information is being processed by a system already committed to a conclusion.

A handful of practices reliably reduce (though never eliminate) confirmation bias. Pre-registration: state in advance what evidence would change your mind. The question what would change my mind is the single most useful cognitive intervention documented in the debiasing literature. Consider-the-opposite prompts: deliberately generate the strongest reasons a reasonable person might disagree with you (Lord, Lepper, and Preston, 1984). Steel-manning: state the opposing view in its most defensible form before responding. Structured disagreement: seek out one thoughtful person who disagrees and try to learn from them, not defeat them. Superforecasters, studied by Tetlock (2015), share these habits. They update visibly and often. They treat changing their minds as a mark of skill, not weakness.

For the next week, notice when you feel certain and ask two questions: what evidence would change my mind, and when did I last update this belief. If you cannot answer either, the belief may be held out of habit rather than knowledge. This is uncomfortable. Confirmation bias feels like conviction; questioning it feels like disloyalty to oneself. The KAfrequency framing reframes it: a mind willing to update is a mind alive. A mind that only confirms is not thinking; it is defending.