Confirmation bias
We seek, notice, and remember what fits what we already believe.
What it means
Confirmation bias is the tendency to search for, interpret, weigh, and recall information in ways that confirm one's prior beliefs while giving disconfirming evidence less attention and tougher scrutiny. It operates through several channels at once: biased search (looking only where supporting evidence is likely), biased interpretation (reading ambiguous data as favorable), and biased memory (better recall for confirming cases). Crucially, it is not limited to motivated or emotional topics or to careless thinkers — it appears even among people sincerely trying to be objective, which is part of what makes it so insidious. A key debate is whether it reflects a genuine cognitive defect or an adaptive feature of reasoning: some argue that individual confirmation-seeking can still yield good collective conclusions through argument and debate. Either way, it is a central driver of belief polarization, overconfidence, persistent stereotypes, and the failure of evidence to change minds, so good practice deliberately seeks disconfirming evidence and considers the opposite.
From the 2-4-6 problem to positive testing
Wason coined the term after his 1960 '2-4-6' task: told a triple that fit a hidden rule, people proposed further triples that fit their guessed rule and rarely tried ones that would break it, then announced wrong rules with confidence. Klayman and Ha later argued this reflects less a thirst for agreement than a general positive test strategy — checking cases you expect to have the target property. That strategy is often sensible, but in many real settings it fails to expose a wrong hypothesis, because a too-narrow guess is confirmed by every positive case while its errors stay hidden. The distinction matters: the flaw is usually not a love of being right but a testing habit that interrogates only one side of a claim.
What the evidence shows
In a classic study, Lord, Ross and Lepper gave supporters and opponents of capital punishment the same mixed research; each side judged the study agreeing with them as better conducted, and both left more convinced than before — biased assimilation producing attitude polarization. That polarization result is real but fragile: it leans on ambiguous, emotionally charged evidence and often shrinks or disappears with clearer data, so it is not an iron law. Newer work points to a mechanism. Kappes and colleagues found people update their confidence normally from opinions that agree with them but discount the strength of opinions that disagree, tracked by muted activity in a prefrontal region. The bias is thus better described as selective weighting than as an inability to take in contrary facts.
Where it does damage
The bias bites hardest where careful judgment is expected. In medicine, an early diagnosis shapes which tests get ordered and how borderline results are read, a pattern sometimes called diagnostic momentum. In forensics, an examiner who knows who the prime suspect is can read an ambiguous fingerprint match more confidently, which is why some laboratories now shield analysts from case background. Auditing, intelligence analysis, hiring and peer review share the shape — one slice of what Nickerson's review documented as a ubiquitous phenomenon in many guises: an initial hunch quietly organizes the hunt for evidence. Finance is the everyday version, where a committed investor treats each headline as vindication. Because the mechanism works through what gets examined and how heavily it counts, the people most exposed are often experts operating inside their own zone of confidence.
What actually reduces it
Awareness is a weak fix: people spot the bias readily in others while missing it in themselves, and merely being warned changes little. What helps is structural. Considering the opposite — deliberately asking why your favored view might be wrong — reliably lowers biased evaluation in experiments. Formal methods build the same move in: analysis of competing hypotheses scores each piece of evidence against every rival at once; pre-registration locks predictions before data can be reinterpreted; red teams and assigned devil's advocates make someone argue the other side. None of these erase the bias, and half-hearted versions decay into box-ticking. The durable lesson is that disconfirmation must be engineered into the process, because individuals seldom produce it on their own against beliefs they already hold.
Examples
An investor convinced a stock will rise reads only bullish analysis and dismisses warning signs as noise.
A parent sure that sugar makes her child hyper remembers every wild afternoon after a birthday party and forgets the equally wild ones that followed a plain lunch.
A doctor who settles on a diagnosis in the first minute orders the tests that would confirm it, reads borderline results as supporting it, and explains away the symptom that does not fit.
A hiring manager who forms a quick impression of a candidate asks softball questions that let a favorite shine and tough ones that trip up someone she has already dismissed.
Convinced a bug lives in the database, an engineer adds logging only there, reads every slow query as the culprit, and never instruments the caching layer where the fault actually sits.
First described in Peter Wason (1960s).
Key references
- Kappes, A., Harvey, A. H., Lohrenz, T., Montague, P. R., & Sharot, T. (2020). Confirmation bias in the utilization of others' opinion strength. Nature Neuroscience, 23(1), 130-137. doi.org/10.1038/s41593-019-0549-2
- Nickerson, R. S. (1998). Confirmation bias: A ubiquitous phenomenon in many guises. Review of General Psychology, 2(2), 175-220. doi.org/10.1037/1089-2680.2.2.175
- Klayman, J., & Ha, Y.-W. (1987). Confirmation, disconfirmation, and information in hypothesis testing. Psychological Review, 94(2), 211-228. doi.org/10.1037/0033-295X.94.2.211
- Lord, C. G., Lepper, M. R., & Preston, E. (1984). Considering the opposite: A corrective strategy for social judgment. Journal of Personality and Social Psychology, 47(6), 1231-1243. doi.org/10.1037/0022-3514.47.6.1231
- Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology, 37(11), 2098-2109. doi.org/10.1037/0022-3514.37.11.2098
- Wason, P. C. (1960). On the failure to eliminate hypotheses in a conceptual task. Quarterly Journal of Experimental Psychology, 12(3), 129-140. doi.org/10.1080/17470216008416717
Where this comes up
- 7 Must-Watch Films for Behavioral Science EnthusiastsDiscover seven must-watch movies that explore behavioral science, cognitive biases, decision-making, persuasion, and hu…