What is confirmation bias?
The research behind confirmation bias started with a football game, and the best-tested remedy comes from aviation.
Answer to the question
What is confirmation bias?
Confirmation bias is the tendency to search for, read and remember evidence in ways that favour what we already believe. It is usually unwitting rather than deliberate. Classic studies show opposing groups rating the same material differently, and a 1998 review treats the term as an umbrella over several habits, some motivated and some not.
Key takeaways
- Hastorf and Cantril (1954) compared how students at Princeton and Dartmouth saw a 1951 football game, and each side counted more infractions by the opposing team on film.
- Lord and colleagues (1979) had supporters and opponents of the death penalty read two studies; each group found the study fitting its view more convincing, and the two ended further apart.
- Nickerson (1998) treats confirmation bias as an umbrella over several habits and finds evidence for both a motivated form and an unmotivated one.
- Guess and Coppock (2020) ran three survey experiments and found no sign of backlash from counter-attitudinal information, even in conditions set up to make it most likely.
- Syed (2015) presents aviation, with independent investigation and shared findings, as a system that keeps the reader of the evidence separate from the person who made the error.
Students at Princeton and Dartmouth watched film of one football game and counted the rule infractions. Each side counted more against the other team, an early picture of confirmation bias.
A football film and a flight recorder
Confirmation bias is the tendency to give evidence an easier ride when it agrees with what we already think. In their 1954 paper on the match played in 1951, Albert Hastorf and Hadley Cantril began by comparing how students at the two universities described it afterwards, then ran the film test. The authors concluded that the two groups had in effect watched different games, and that each version felt real to the person holding it.
The study concerned loyalty and perception and appeared well before the phrase became common. It supplies the picture that the label confirmation bias later described: two people given the same material who come away with different records of it. The students were asked only to count, and the counts still followed the colours they wore, which is why the study is a useful starting point. Confirmation bias, in this picture, operates inside the act of looking, before any argument begins.
Matthew Syed's book Black Box Thinking approaches the same problem from the other direction. He describes aviation as a system in which every accident is investigated independently and the findings are shared, so one crew's mistake becomes a lesson for the whole industry. He calls this an open loop, and he sets it against closed loops, where failure is explained away and the information never travels.
His account of why closed loops form leans on Leon Festinger's work on cognitive dissonance. People tend to reframe evidence that threatens a belief about themselves rather than revise the belief, and the more senior and publicly committed the person, the harder the reframing. That is confirmation bias at the scale of an institution, with investigators and managers in the place of students.
What the studies show, and where they stop
A 1979 experiment moved the football game's lesson onto the page. Charles Lord and his colleagues recruited people with strong opinions on the death penalty, some in favour and some opposed, and had them read descriptions of two studies on whether it deters crime. One study appeared to support their side and the other to cut against it.
Both groups rated the study that fitted their view as more convincing and better conducted, and both moved further in its direction. The authors called the outcome attitude polarisation: after reading the same balanced material, the two sides ended further apart than they began. The authors' explanation is that people accept supportive findings with little scrutiny and pick apart unwelcome ones. Everyone read identical material, so the difference lay in the reading. For anyone asking what confirmation bias looks like in a controlled setting, this is the standard answer: a fair, balanced body of evidence can feed two opposite conclusions when the readers start from opposite places.
A long review by Ray Nickerson, published in 1998, treats confirmation bias as an umbrella over several related habits rather than one mistake. It covers seeking evidence that favours a hypothesis, giving extra weight to supportive evidence, interpreting ambiguous material in the favoured direction and remembering the past as more consistent with a view than it was. It follows the idea into medical diagnosis, jury verdicts and scientific resistance to new theories.
Nickerson separates a motivated form, in which people defend beliefs they value, from an unmotivated form, in which they are partial even on questions they care little about. He found evidence for both forms of confirmation bias. He also notes that the term usually describes unwitting selectivity, not the deliberate case-building of a lawyer or a debater.
The widening-gap picture of confirmation bias has a limit. Andrew Guess and Alexander Coppock defined backlash as people who meet evidence against their view holding it more strongly afterwards, and tested for it in three survey experiments, two on online worker samples and one on a nationally representative sample. They found no sign of backlash, even in conditions set up to make it most likely. They concluded that it is much less common than the information-processing literature often implies.
The results measure different things. The 1979 study tracked how convincing people found each report and where their attitudes ended up, while the survey experiments asked whether counter-attitudinal information pushed views the wrong way. Uneven reading of evidence is well documented, and the stronger claim that opposing evidence hardens opinion has found less support.
Why outside readers matter
Aviation's practice deals with confirmation bias without asking investigators to be unbiased. Someone other than the people involved reads the evidence, and the conclusions go to everyone. The football students and the death-penalty readers were not short of evidence. They lacked a reader with no side.
- Someone without a stake reads the evidence, not the person who made the call.
- The conclusions go to everyone, so one case becomes a lesson for the rest.
- The result that would count against the view is fixed before the evidence arrives.
The same logic applies to ordinary decisions where a view has settled. A purchase defended after the fact, as in Why You Defend What You Just Bought, and a decision judged only by how it turned out, as in What Outcome Bias Does to a Good Call, both leave the judging to the person with the stake, and We Define First, Then We See follows the same habit into advertising, where a claim is read as a hypothesis to be confirmed.
Passing the material to someone with nothing riding on the answer, and settling beforehand which outcome would count against the view, changes who makes the call, which is the arrangement aviation reached when it gave the reading of its failures to people outside the crews involved.
Common questions
What is confirmation bias?
Confirmation bias is the tendency to search for, read and remember evidence in ways that favour what we already believe. It is usually unwitting, and the research covers seeking, interpreting and remembering.
Is confirmation bias the same as motivated reasoning?
Not exactly. Nickerson's 1998 review separates a motivated form, in which people defend beliefs they value, from an unmotivated form, in which they are partial even when the answer matters little to them. The review found evidence for both.
Is confirmation bias deliberate?
Usually not. Nickerson describes the term as covering unwitting selectivity, which is different from a lawyer or debater building a case on purpose.
Does confirmation bias always make opinions more extreme?
No. The 1979 death-penalty experiment found opposing groups ending further apart after reading the same studies, but Guess and Coppock's three survey experiments found no sign of backlash when people met evidence against their views.
Sources
Syed, M. (2015). Black Box Thinking: Marginal Gains and the Secrets of High Performance. London: John Murray. Seed book; vault hub Books/Black Box Thinking.
Festinger, L. (1957). A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press.
Hastorf, A. H., & Cantril, H. (1954). “They saw a game: A case study.” The Journal of Abnormal and Social Psychology, 49(1), 129-134. doi:10.1037/h0057880 https://doi.org/10.1037/h0057880
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:10.1037/0022-3514.37.11.2098 https://doi.org/10.1037/0022-3514.37.11.2098
Nickerson, R. S. (1998). “Confirmation bias: A ubiquitous phenomenon in many guises.” Review of General Psychology, 2(2), 175-220. doi:10.1037/1089-2680.2.2.175 https://doi.org/10.1037/1089-2680.2.2.175
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. doi:10.1017/S0007123418000327 https://doi.org/10.1017/S0007123418000327