Confirmation Bias and How to Reduce It

Confirmation bias is the habit of looking for, interpreting and remembering evidence in ways that favor what you already believe. The APA Dictionary of Psychology describes it as gathering evidence that confirms your expectations, usually by focusing on support and dismissing contradictory evidence or never looking for it (APA Dictionary of Psychology). In a 1998 review, psychologist Raymond Nickerson defined it as seeking or interpreting evidence in ways that are partial to existing beliefs, expectations or a hypothesis in hand (Nickerson, 1998). He stressed that people usually do this without meaning to and without noticing (same source).

Where the idea comes from

People noticed this long before psychologists tested it. Nickerson’s review quotes Francis Bacon, who wrote in 1620 that once the mind adopts an opinion, it draws everything else in to support it and neglects or sets aside the cases on the other side (Nickerson, 1998).

In a 1960 experiment by P. C. Wason, people tried to work out the rule behind a set of three numbers, such as 2-4-6, by proposing new sets and hearing whether each one fit (Nickerson, 1998). People typically tested only sets that matched their own guess, like more even numbers going up by two. If the real rule was broader, such as any three increasing numbers, every test came back “yes,” and they never found out their guess was wrong (same source). Wason built the task so that relying on confirming evidence alone would almost certainly lead to a wrong answer. Only 6 of his 29 participants reached the correct rule without first announcing a wrong one (Wason, 1960).

Nickerson’s review pulled together laboratory findings, examples from medicine, law and science, possible explanations, and the question of whether the bias ever serves a purpose. He separated a motivated form, where people defend a belief they want to keep, from an unmotivated form, where they treat evidence unevenly even with no stake in the answer (Nickerson, 1998).

The forms it takes

It helps to think of confirmation bias at three stages: what you look for, how you read it and what you remember.

Form	What it looks like	Example from research
Biased search	Seeking information that supports your view, and testing ideas only in ways that could confirm them	A meta-analysis found a moderate preference for information that supports people’s existing attitudes, beliefs or behaviors over information that challenges them (d = 0.36) (Hart et al., 2009)
Biased interpretation	Accepting supporting evidence at face value while picking apart evidence against you	Supporters and opponents of the death penalty each rated the study that agreed with them as more convincing (Lord, Ross & Lepper, 1979)
Biased memory	Remembering what fits your view better than what doesn’t	People recognized information that supported their position on gun control better than opposing information (Frost et al., 2015)

A few studies show how this plays out:

• In personality studies, people asked to check whether someone was an extrovert tended to ask questions that an extrovert would answer with “yes” (Nickerson, 1998).
• In the death penalty study, 48 undergraduates read two made-up studies on whether capital punishment deters crime, one for and one against. Both sides found the study on their side more convincing, and their self-reported beliefs moved further apart after reading the same mixed evidence (Lord, Ross & Lepper, 1979).
• Two groups watched the same video of a child taking a test. The group told she came from a high socioeconomic background rated her performance above grade level, and the group told her background was low rated the same performance below grade level (Nickerson, 1998).
• Mock jurors were more likely to remember statements as trial evidence when the statements fit the verdict they had chosen, including some statements that had never been presented (same source).

Everyday examples

Nickerson’s review gives many real-world cases (Nickerson, 1998):

• People who want to believe in horoscopes or psychics can find a few predictions that came true, ignore the misses, and forget that anyone could have made guesses with a similar hit rate.
• Gamblers often explain away their losses, which lets them believe their odds are better than they are.
• The body sends a steady stream of small signals, so someone afraid of being ill can find plenty that seem to confirm the fear.
• Doctors tend to form one or a few hypotheses early, which can make them less likely to consider another explanation and can color how they read later test results.
• In the heat of a dispute, each side builds the strongest case for its own position and ignores evidence against it, which is why both sides so often claim victory.

The same thing shows up in what people choose to read. The selective-exposure meta-analysis found that people generally lean toward information that agrees with them (Hart et al., 2009). Our guide on how to evaluate sources has practical checks for this.

Why it happens

Researchers point to several causes, and Nickerson thought motivational and cognitive factors are probably both at work, each able to feed the other (Nickerson, 1998).

Wanting a particular answer is the first. Kunda’s review concluded that people are more likely to reach conclusions they want to reach, but only as far as they can build reasonable-looking justifications for them (Kunda, 1990). In the selective-exposure meta-analysis, the pull toward agreeable information was weaker when a belief wasn’t tied to someone’s values or wasn’t held with conviction. When challenging information was useful for a goal people had at the moment, they actually preferred it (Hart et al., 2009).

Limited attention plays a part too. People tend to think about one hypothesis at a time and to consider only the possibility that it’s true, so a wrong idea that’s close to right can keep collecting apparent support (Nickerson, 1998).

Then there is the habit of testing the cases you expect to fit. Klayman and Ha argued that many results labeled confirmation bias are better explained by a “positive test strategy.” In realistic conditions it is often a good way to check an idea, but it can lead to systematic errors (Klayman & Ha, 1987). Wason’s task is a case where it fails, because the real rule is broader than the guess being tested (Nickerson, 1998).

How to reduce it

One well-tested technique is called considering the opposite. In two experiments, people who were prompted to think about possibilities opposite to their view, either through direct instructions or through materials that made those possibilities stand out, judged evidence less one-sidedly. The prompt worked better than simply telling them to be as fair and unbiased as possible (Lord, Lepper & Preston, 1984). Nickerson also cited studies suggesting that thinking of reasons against a judgment, and of alternative explanations early on, can help (Nickerson, 1998).

Training has some support as well:

• In two experiments, one session of training (a computer game or an instructional video) reduced several biases. In the first experiment the targets were confirmation bias, the bias blind spot and the fundamental attribution error. The effects were medium to large right away and still present at least two months later, and games with personal feedback worked better than videos (Morewedge et al., 2015).
• In a field test with 290 graduate students in three professional programs, the timing of training happened to put some students through it before an unannounced business case modeled on the Space Shuttle Challenger launch decision. Trained students were 19% less likely to pick the inferior solution that confirmed their hypothesis (Sellier et al., 2019). A 2020 correction fixed a miscalculated percentage in the original paper, and 19% is the corrected figure (Psychological Science, 2020).

These results are encouraging but limited. They come from a small number of studies, and the 2015 effects are reported across several biases at once. In 1998, Nickerson wrote that the question of how far training can change the bias deserved more research (Nickerson, 1998).

Put into everyday habits, those findings look something like this:

1. Before you settle on a view, write down reasons it could be wrong.
2. Ask what you would expect to see if your idea were false, then go and look for that, the way a 2-4-6 player should test a set that breaks their own rule.
3. Read the strongest version of the other side’s case. Our guide to the straw man fallacy covers the opposite habit.
4. Check most carefully the claims you’d like to be true.

For more ways to build these habits, see how to improve your critical thinking.