What is confirmation bias?

Confirmation bias is the tendency to seek, interpret, and remember information in a way that confirms what you already believe. It has three parts: search (you look for supporting evidence and not for disconfirming evidence), interpretation (the same ambiguous fact reads as supporting your side), and memory (you recall the hits and forget the misses). There is a 'cold' version that is just a byproduct of how we test ideas, and a 'hot' version driven by wanting something to be true.

What was Wason's 2-4-6 task?

In Wason (1960), 29 people were told that 2, 4, 6 fits a hidden rule and asked to discover it by proposing their own triples. Most guessed a narrow rule like 'even numbers going up by two' and only tested triples that fit it, which can confirm a guess but never disprove it. The real rule was simply 'any three numbers in increasing order.' Only 6 of the 29 discovered it without first announcing a wrong rule, because almost nobody tested a triple they expected to fail.

Is confirmation bias always a flaw?

Not necessarily. Testing an idea by looking for cases where it holds is a 'positive test strategy' that works well in most real environments; Wason's puzzle is rigged so that it fails. One influential theory argues reasoning evolved to win and evaluate arguments in a group, so myside bias is the system working as designed, more lawyer than scientist. Studies also find it barely correlates with intelligence. It becomes dangerous mainly when you search using a claim's own loaded terms, which can lead you into a 'data void' that confirms a falsehood.

Heuristic ▶ Subscribe

Episode 5 · Cognitive biases · Research notes

A rule fits 2, 4, 6. Why almost nobody can solve it.

You are told a hidden rule fits the numbers 2, 4, 6, and you can test any triple you like. Almost everyone tests numbers that fit their first guess, hears "yes," and confidently declares the wrong rule. The move that would actually reveal the answer, testing something you expect to fail, is the one almost nobody makes. That is confirmation bias.

The short answer: a triple that fits your guess can only make you feel surer; it can never prove you wrong. The real rule was just "any three numbers in increasing order," and only 6 of 29 people found it without a wrong guess first. The honest twist: this may not be a defect at all. It looks like a shortcut that usually works, your brain may be built to win arguments more than to find truth, and "doing your own research" can make a false belief stronger.

Episode 5 premieres

Wednesday, July 29 · 8:00 AM PT

▶ Subscribe to catch it

The two experiments

Peter Wason ran the 2-4-6 task in 1960. People tested only triples that fit their guess (8-10-12, 20-40-60), which felt confirming but proved nothing, and 4 in 5 announced a wrong rule at least once. In his 1968 selection task, people were shown four cards (E, K, 4, 7) and a rule, "a vowel means an even number on the back." Most turned the E (correct) and the 4 (useless, it cannot break the rule), while hardly anyone turned the 7, the one card that could actually falsify it. Together: we neither generate the disconfirming test nor choose it when it is right in front of us.

Same evidence, opposite conclusions

In a classic 1979 study, people with strong opposite views on the death penalty read the same two studies. Each side judged the study that agreed with them as careful and picked apart the one that disagreed. The identical page was brilliant research to one person and garbage to another. This biased grading of evidence is one of the most reliable findings in the field (the stronger claim that people polarize further apart has been harder to reproduce, so hold that one loosely).

The twist: maybe it is not a bug

Testing where an idea holds is a "positive test strategy" that works fine in most of the real world; Wason's puzzle is a rigged exception. One influential theory holds that reasoning evolved to win and evaluate arguments in a group, so your brain behaves less like a broken scientist and more like a working lawyer, and being smart does not save you, myside bias barely correlates with intelligence. The real danger shows up in 2024 research: people who searched online to check a false story often believed it more, because they searched using the story's own loaded words and landed in a "data void" of low-quality pages echoing the claim. The lesson is not that research is bad; it is that how you search decides what you find.

What actually helps

The single move with the best evidence is to consider the opposite: before you settle, ask whether you would judge this evidence the same way if it pointed the other direction. That one question measurably reduced biased grading in studies, where simply being told to "be fair" did almost nothing. So when you catch yourself collecting reasons you are right, go hunt the best reason you are wrong; search the neutral version of a question, not the loaded one; and try to state the other side so well they would agree with your version.

Sources

Nickerson, R. S. (1998). Confirmation bias. Review of General Psychology, 2(2), 175-220.
Wason, P. C. (1960). On the failure to eliminate hypotheses in a conceptual task. QJEP, 12(3), 129-140.
Wason, P. C. (1968). Reasoning about a rule. QJEP, 20(3), 273-281.
Lord, C. G., Ross, L. and Lepper, M. R. (1979). Biased assimilation and attitude polarization. JPSP, 37(11), 2098-2109.
Klayman, J. and Ha, Y. (1987). Confirmation, disconfirmation, and information in hypothesis testing. Psychological Review, 94(2), 211-228.
Mercier, H. and Sperber, D. (2011). Why do humans reason? Behavioral and Brain Sciences, 34(2), 57-74.
Stanovich, K. E., West, R. F. and Toplak, M. E. (2013). Myside bias, rational thinking, and intelligence. Current Directions, 22(4), 259-264.
Aslett, K. et al. (2024). Online searches to evaluate misinformation can increase its perceived veracity. Nature, 625(7995), 548-556.