Skip to main content
    Path to Science

    Thinking Dictionary

    Confirmation bias

    What it is

    Confirmation bias is our habit of looking for, trusting, and often remembering information that agrees with what we already believe. At the same time, we tend to overlook or explain away information that doesn't agree.¹ It isn't lying, and it isn't stupidity. It shows up in ordinary people and in experts alike.¹

    There seem to be two reasons for it. One is a mental shortcut. We check new information against what we already think, because most of the time that works fine.² The other is wanting. We hold evidence we like to a lower standard than evidence we don't like.³˒⁴

    The result is that we tend to test our ideas only where they predict a "yes." Some mistakes never get a chance to show up. We end up more sure of ourselves than the evidence deserves.⁵

    Why it matters

    Good thinking means letting reality correct us. Confirmation bias gets in the way. It lets a belief collect support without ever facing a real test.

    Science at its best tries hard to prove ideas wrong. An idea that survives an honest attempt to break it is worth far more than one that was only shown friendly evidence. But scientists are human too. In one experiment, people playing scientist in a computer simulation mostly chose tests that would confirm their ideas. The good news from the same experiment: when they did get clear evidence against an idea, they used it and dropped the idea.⁶ So the problem is mostly in which tests we choose, not in refusing to accept a clear "no."

    The bias also changes how we judge evidence. In a classic study, supporters and opponents of the death penalty read the same two studies. One seemed to show the death penalty works, and one seemed to show it doesn't. Each side rated the study that agreed with them as more convincing and better done.⁷ (That study also reported that people's views became more extreme. That part has not held up well in later research.⁸˒⁹)

    In everyday life

    • At work. You decide early that a new colleague is careless. After that, every typo in their emails feels like proof, and their good work slips by unnoticed. A month later your impression feels "confirmed," but you were only counting one side. Studies find the same pattern. People watched the very same video of a schoolgirl, and judged her as above or below her grade depending on a label about her they had been given beforehand.¹⁰ And people checking a guess about someone tend to ask questions that can only confirm it.¹¹
    • Health. You start a new supplement. On good days, you credit the pills. On bad days, you blame poor sleep or stress. Unless you count both kinds of day, you can't tell whether the supplement did anything.¹ Arthritis patients who were sure their pain followed the weather found no real link once their pain was recorded for more than a year.¹²
    • Shopping. After buying an expensive phone, you read reviews. The glowing ones feel accurate. The critical ones feel nitpicky. You aren't really learning about the phone. You're defending your choice. Back in the 1950s, people who had just bought a new car read more adverts for the car they had bought.¹³
    • News and social media. Your feed shows you more of what you click, and you mostly click what you agree with. On Facebook, people's own clicks and choices cut out more opposing news than the feed's ranking program did.¹⁴ Opposing views don't vanish. Social media even shows people some of the other side.¹⁵ But those views get scarce and easy to scroll past, and it can start to feel like "everyone" agrees with you. Still, don't blame the feed for everything. One large study cut the posts from like-minded sources in people's feeds by about a third and found no change in the attitudes it measured.¹⁶
    • Relationships. If you suspect a friend doesn't care, a late reply reads as indifference. If you trust them, the same late reply reads as a busy day. The evidence is the same. Your belief decides what it means. Married couples show this too: their view of the relationship shapes how they explain the very same behaviour from their partner.¹⁷

    The common trap

    The trap is thinking you've tested an idea when you've only looked for agreement. Searching "does X work" tends to bring up pages that say it does, and that feels like research.¹ A meta-analysis is a study that combines the results of many earlier studies. One meta-analysis pooled 67 reports with nearly 8,000 people. It found a moderate preference for information that agrees with what people already think. The preference got weaker when getting the right answer really mattered.¹⁸

    Some ways out:

    • Consider the opposite. Ask yourself: "What would I expect to see if I were wrong?" Then go and look for exactly that. For example, also search "X doesn't work" or "X evidence review." This is the best-tested tip. In experiments, it reduced bias more than simply telling people to "be fair."¹⁹ Listing reasons why your answer might be wrong also makes people less overconfident.²⁰
    • Be wary of pleasant news. We ask for less proof when we like the conclusion.⁴ So treat evidence that feels good as a cue to check twice.
    • Steelman the other side. Describe the opposing view so well that someone who holds it would nod along. This is sensible advice, but it hasn't been tested directly. And in experiments, simply telling people to "look for evidence against your idea" hasn't always helped.²¹

    Try it: the 2, 4, 6 game

    You need a friend and about five minutes. Give your friend the numbers 2, 4, 6. Use this secret rule: any three numbers that go up. Their job is to find the rule. They do it by suggesting other sets of three numbers, and you answer only "fits" or "doesn't fit." Tell them to announce the rule only when they feel very confident.

    Most people try sets like 8, 10, 12. These fit their guess, "add 2 each time." They hear "fits" again and again, and confidently announce the wrong rule. In the psychologist Peter Wason's original 1960 study, only 6 of 29 students got it right on their first try.⁵

    The people who crack it try sets that their own guess says should not fit, like 1, 2, 50 or 2, 4, 7. Hearing "fits" for those shows that the guess is too narrow. A set like 6, 4, 2 doesn't help. It breaks both the guess and the real rule, so it gets "doesn't fit" either way and can't tell them apart.⁵ Swap roles afterwards and see if you do better.

    A fair warning about the game

    This puzzle is rigged to punish confirming tests. The guess "add 2" fits completely inside the real rule, so every set that fits the guess also fits the rule. In everyday life, checking where you expect a "yes" is often a sensible first step, and it can still turn up surprises.² So the lesson is not "never confirm." The habit worth building is to also check where your idea predicts a "no."

    Related concepts

    Falsifiability · Cherry-picking · Motivated reasoning · Correlation vs causation · Placebo effect · Credible sources · Lateral reading · Probabilistic reasoning · Iteration · Fast and slow thinking · Cognitive shortcuts · Actively open-minded thinking

    Sources

    1. Nickerson, R. S. (1998). Confirmation bias: A ubiquitous phenomenon in many guises. Review of General Psychology, 2(2), 175–220. https://doi.org/10.1037/1089-2680.2.2.175

    2. Klayman, J., & Ha, Y.-W. (1987). Confirmation, disconfirmation, and information in hypothesis testing. Psychological Review, 94(2), 211–228. https://doi.org/10.1037/0033-295X.94.2.211

    3. Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108(3), 480–498. https://doi.org/10.1037/0033-2909.108.3.480

    4. Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differential decision criteria for preferred and nonpreferred conclusions. Journal of Personality and Social Psychology, 63(4), 568–584. https://doi.org/10.1037/0022-3514.63.4.568

    5. Wason, P. C. (1960). On the failure to eliminate hypotheses in a conceptual task. Quarterly Journal of Experimental Psychology, 12(3), 129–140. https://doi.org/10.1080/17470216008416717

    6. Mynatt, C. R., Doherty, M. E., & Tweney, R. D. (1977). Confirmation bias in a simulated research environment: An experimental study of scientific inference. Quarterly Journal of Experimental Psychology, 29(1), 85–95. https://doi.org/10.1080/00335557743000053

    7. 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. https://doi.org/10.1037/0022-3514.37.11.2098

    8. Kuhn, D., & Lao, J. (1996). Effects of evidence on attitudes: Is polarization the norm? Psychological Science, 7(2), 115–120. https://doi.org/10.1111/j.1467-9280.1996.tb00340.x

    9. Guess, A., & Coppock, A. (2020). Does counter-attitudinal information cause backlash? Results from three large survey experiments. British Journal of Political Science, 50(4), 1497–1515. https://doi.org/10.1017/S0007123418000327

    10. Darley, J. M., & Gross, P. H. (1983). A hypothesis-confirming bias in labeling effects. Journal of Personality and Social Psychology, 44(1), 20–33. https://doi.org/10.1037/0022-3514.44.1.20

    11. Snyder, M., & Swann, W. B., Jr. (1978). Hypothesis-testing processes in social interaction. Journal of Personality and Social Psychology, 36(11), 1202–1212. https://doi.org/10.1037/0022-3514.36.11.1202

    12. Redelmeier, D. A., & Tversky, A. (1996). On the belief that arthritis pain is related to the weather. Proceedings of the National Academy of Sciences, 93(7), 2895–2896. https://doi.org/10.1073/pnas.93.7.2895

    13. Ehrlich, D., Guttman, I., Schönbach, P., & Mills, J. (1957). Postdecision exposure to relevant information. Journal of Abnormal and Social Psychology, 54(1), 98–102. https://doi.org/10.1037/h0042740

    14. Bakshy, E., Messing, S., & Adamic, L. A. (2015). Exposure to ideologically diverse news and opinion on Facebook. Science, 348(6239), 1130–1132. https://doi.org/10.1126/science.aaa1160

    15. Flaxman, S., Goel, S., & Rao, J. M. (2016). Filter bubbles, echo chambers, and online news consumption. Public Opinion Quarterly, 80(S1), 298–320. https://doi.org/10.1093/poq/nfw006

    16. Nyhan, B., Settle, J., Thorson, E., Wojcieszak, M., et al. (2023). Like-minded sources on Facebook are prevalent but not polarizing. Nature, 620(7972), 137–144. https://doi.org/10.1038/s41586-023-06297-w

    17. Bradbury, T. N., & Fincham, F. D. (1990). Attributions in marriage: Review and critique. Psychological Bulletin, 107(1), 3–33. https://doi.org/10.1037/0033-2909.107.1.3

    18. Hart, W., Albarracín, D., Eagly, A. H., Brechan, I., Lindberg, M. J., & Merrill, L. (2009). Feeling validated versus being correct: A meta-analysis of selective exposure to information. Psychological Bulletin, 135(4), 555–588. https://doi.org/10.1037/a0015701

    19. 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. https://doi.org/10.1037/0022-3514.47.6.1231

    20. Koriat, A., Lichtenstein, S., & Fischhoff, B. (1980). Reasons for confidence. Journal of Experimental Psychology: Human Learning and Memory, 6(2), 107–118. https://doi.org/10.1037/0278-7393.6.2.107

    21. Tweney, R. D., Doherty, M. E., Worner, W. J., Pliske, D. B., Mynatt, C. R., Gross, K. A., & Arkkelin, D. L. (1980). Strategies of rule discovery in an inference task. Quarterly Journal of Experimental Psychology, 32(1), 109–123. https://doi.org/10.1080/00335558008248237