Confirmation Bias: Definition, Examples and Why IQ Doesn't Protect You

Confirmation bias at a glance

• Biased search: you look mainly for information that could support your view and rarely for information that could sink it.
• Biased interpretation: supporting evidence is accepted at face value; opposing evidence is checked for flaws.
• Biased memory: the cases that fit are the ones that come back to you later.
• Where it sits: it is one entry in the wider family of cognitive biases, and probably the one with the broadest reach.

Where the idea comes from: Wason’s 2-4-6 task

In 1960 the British psychologist Peter Wason gave people the number sequence 2, 4, 6 and told them it followed a rule he had in mind. They could test other sequences of three numbers, and he would say whether each one fitted, until they felt sure enough to announce the rule. The rule was simply “any three numbers in increasing order”. Most people formed a narrower idea, such as “add two each time”, and then tested only sequences that fitted it, like 8, 10, 12. Every test came back “yes”, so they grew confident in the wrong rule.

Of Wason’s 29 participants, intelligent young adults by his own description, only six found the correct rule without first announcing a wrong one; 13 announced one wrong rule and nine announced two or more. The way out was to try a sequence the guess said should fail, such as 1, 2, 3 or 10, 5, 1. Those who never did so had no way to discover their mistake. Wason’s later selection task (1968) showed the same pattern in logic: asked which cards to turn over to test a rule of the form “if P then Q”, most people miss the card that could prove the rule false.

The three forms, with real examples

Biased search: asking questions that can only say yes

In a 1978 series of studies by Mark Snyder and William Swann, participants were told to find out whether the person they were about to interview was an extravert. They mostly chose questions about sociable, outgoing situations, the kind an extravert answers well, rather than questions that could reveal an introvert. The interviewees, answering the questions put to them, then really did come across as more outgoing. The hypothesis produced its own evidence.

Biased interpretation: the same evidence, opposite conclusions

Charles Lord, Lee Ross and Mark Lepper gave supporters and opponents of capital punishment two purported studies, one appearing to show that the death penalty deters murder and one appearing to show that it does not. Both sides judged the study that agreed with them as better designed and more convincing, picked holes in the other, and came away holding their original position more firmly. Mixed evidence did not narrow the disagreement; it widened it. The authors called this biased assimilation and attitude polarization.

Choosing what to read

A 2009 meta-analysis led by William Hart and Dolores Albarracín pooled experiments in which people could choose between information that supported their views and information that challenged them. On average they showed a moderate preference for the supporting side (d = 0.36). The preference was weaker when the view mattered less to them or the information on offer was low in quality, and it reversed when the challenging information was useful for a goal they had at that moment. In other words, the bias responds to motives, and changing the motive changes the behaviour.

Groups, courtrooms and laboratories

Groups do not cancel the bias out. In 2000, Stefan Schulz-Hardt and colleagues at Ludwig-Maximilians-Universität München found that groups prefer supporting information just as individuals do, and that the preference grows the more members already agree before the discussion starts. The effect was replicated with working managers. A team of like-minded people can therefore be more one-sided than any of its members.

In forensic work the cost is concrete. A 2013 review by Saul Kassin, Itiel Dror and Jeff Kukucka, prompted in part by a high-profile mistaken fingerprint identification in the Madrid train bombing case, describes how a confession or other case information can shape what examiners and witnesses later see and remember. Their proposed safeguards include limiting what examiners know about the rest of the case while they do the comparison.

Is confirmation bias always irrational?

Not always, and an honest page should say so. In 1987 Joshua Klayman and Young-Won Ha argued that much of what is labelled confirmation bias is better described as a positive test strategy: checking the cases you expect to fit. Under many realistic conditions that is an efficient way to test an idea. It fails in situations like Wason’s, where the true rule is broader than your guess, so that only a test you expect to fail can reveal the error. The useful lesson is not “never look for confirmation” but “make sure at least one of your tests could prove you wrong”.

Does a high IQ protect you from confirmation bias?

This is where intelligence research has something definite to say, and it is not flattering. Keith Stanovich, Richard West and Maggie Toplak reviewed the evidence in 2013 and concluded that the size of myside bias shows very little relation to intelligence. Myside bias is the form of confirmation bias that favours your own opinions. Avoiding it is, in their words, a rational thinking skill that intelligence tests do not assess.

A 2017 experiment by Dan Kahan, Ellen Peters, Erica Cantrell Dawson and Paul Slovic went further. Participants interpreted the same table of results, presented either as a trial of a skin-rash cream or as a study of a gun-control ban. With the skin cream, the most numerate people did clearly better. With guns, answers split along political lines, and the split was larger among the most numerate participants, not smaller. Skill with numbers was used to reach the answer people preferred.

So a high score on an IQ test is no certificate of open-mindedness. Reasoning ability decides how well you can analyse evidence; whether you apply it evenly to evidence you dislike is a separate habit. Our pages on critical thinking, metacognition and the Dunning-Kruger effect look at that habit from other angles.

How to reduce confirmation bias

• Consider the opposite. In 1984 experiments by Lord, Lepper and Elizabeth Preston, asking people how they would have judged the evidence if the results had come out the other way reduced biased judgement more than telling them to be “as fair and unbiased as possible”. Good intentions alone did little.
• Run a test that could fail. Before acting on a belief, write down one observation that would show it is wrong, then look for it.
• Decide in advance what would change your mind. It is much harder to move the goalposts once you have named them.
• Build disagreement into groups. Because like-minded groups search more one-sidedly, include someone whose first preference differs, or assign a member to argue the other side.
• Separate the information from the judgement. Where possible, assess evidence before you learn which conclusion it supports, the principle behind the forensic safeguards above.

Trust and scope notes

This page summarises published research for general information, with figures taken from the studies listed below. The IQ Revealed test is an online reasoning test of 40 original items for adults aged 18 and over; it does not measure confirmation bias or any other thinking bias, and, as the research above shows, a reasoning score says little about how evenly someone weighs evidence. Our guide to how IQ tests work explains what an online test can and cannot tell you.

IQ Revealed is not affiliated with, endorsed by or connected to the researchers, universities, journals or publishers named on this page. Names are used only to identify the research being described.