Confirmation Bias: Definition, Examples & How to Beat It

What is confirmation bias?

Confirmation bias is your mind’s habit of treating the evidence for your current belief as stronger than the evidence against it. You look for support, read ambiguous facts as support, and remember the support better.

It shows up in three places:

1. Search: which questions you ask and where you look. You google “is coffee good for you,” not “is coffee bad for you.”
2. Interpretation: how you read the same evidence. A study that agrees with you looks well designed; one that disagrees looks flawed.
3. Memory: what sticks. You recall the times your gut was right and forget the times it wasn’t.

The bias is not the same as lying to yourself. Much of it is mechanical: testing an idea by checking cases where it should be true is simply the most natural way to test. That is what makes it so common and so hard to see from the inside. It feels like doing research.

Where it comes from

The first clear experimental demonstration came from Peter Wason, a psychologist at University College London, in 1960.

Wason’s 2-4-6 task

Wason told participants that the sequence 2, 4, 6 fit a rule he had in mind. Their job was to discover the rule by proposing other sequences of three numbers. After each one, he would say whether it fit. When they felt sure, they announced the rule.

Most people quickly formed a hypothesis such as “even numbers going up by two,” then tested it with sequences like 8, 10, 12 and 20, 22, 24. Every test got a “yes.” Confident, they announced their rule, and many were wrong.

Wason’s actual rule was much broader: any three numbers in ascending order. The sequence 1, 2, 3 fits. So does 5, 17, 300. The only way to find that out was to propose sequences that should fail under your own hypothesis, such as 1, 3, 5 or 10, 4, 1. Few people did. They looked for confirmation, and confirmation kept arriving, because every sequence that fit their narrow rule also fit the wider one.

That is the core lesson: evidence that fits your theory is not proof of your theory if it would also fit other theories. The informative test is the one that could prove you wrong.

What came after

Wason later designed the four-card selection task (1966), which showed people again failing to look for disconfirming cases. In 1979, Charles Lord, Lee Ross and Mark Lepper gave people who supported or opposed the death penalty the same mixed set of studies. Each side rated the studies that agreed with them as more convincing, and both sides ended up more confident in their original view. The same evidence pushed them further apart. Raymond Nickerson’s 1998 review in Review of General Psychology called confirmation bias “a ubiquitous phenomenon in many guises,” which remains a fair summary.

Psychologists Joshua Klayman and Young-Won Ha (1987) added a nuance: testing cases you expect to fit (a “positive test strategy”) is often a sensible default. It just fails in situations like the 2-4-6 task, where the truth is broader than your guess.

The idea is old. Francis Bacon wrote in Novum Organum (1620) that once the human understanding has adopted an opinion, it draws everything else to support and agree with it.

How to beat confirmation bias

You beat confirmation bias by deliberately looking for the evidence that would prove you wrong, before you look at anything else. Six steps:

1. Write your belief as a testable claim. “This hire will work out” is too vague. “She will ship the onboarding redesign by March” can be checked.
2. Invert it. Ask: “What would I expect to see if I were wrong?” This is inversion, and it’s the most direct fix for confirmation bias because it forces the 1, 3, 5 test instead of the 8, 10, 12 test.
3. Define your kill criteria in advance. Before you look at new data, decide what result would change your mind. If you set the bar afterwards, you’ll set it where you’re already standing.
4. Search for the other side by name. Type the opposite query. Read the best critic, not the weakest. Steelmanning the view you reject means stating it so well its supporters would agree.
5. Get an outside reader. Assign a colleague to argue against the plan, or run a pre-mortem: imagine it’s a year from now and the decision failed, then list why.
6. Write down your prediction and confidence. A dated record stops memory from editing itself. Over time, it also shows whether you have a pattern of seeing what you want.

The asymmetry to remember: one strong piece of disconfirming evidence can tell you more than ten pieces of agreeing evidence, especially if the agreeing evidence would show up either way.

3 real-world examples

The examples below are hypothetical, written to show common patterns.

Business

Say a product team believes customers want a new dashboard. They run interviews and ask, “Would a dashboard help you track your projects?” Most people say yes. Questions phrased that way invite agreement, and saying yes costs nothing. A disconfirming version would be: “Walk me through the last time you checked on a project. What did you use?” If nobody mentions wanting a dashboard, that is the 1, 3, 5 result the team needed. Confirmation bias is one of the reasons overconfidence bias runs so high in product planning: the research was designed to agree.

Personal life

Say you’ve decided a new neighbor is unfriendly because they didn’t wave once. Now every short reply and closed curtain counts as proof. The friendly moments (they held the door, they returned a package) get filed as exceptions. A quick fix: ask what else would explain the behavior. Busy? Shy? Didn’t see you? If several explanations fit equally well, the evidence isn’t telling you much.

Investing

Say you own shares in a company and you’re convinced it will win its market. You follow the founder on social media, read bullish forums and skim the critical reports. That is a curated feed, not research. A better routine: once a quarter, read the strongest bear case you can find and write down what would have to happen for it to be right. If those things start happening, you’ll recognize them instead of explaining them away. This is also how cognitive dissonance gets defused before it builds.

Common mistakes

• Thinking smart people are immune. Intelligence makes you better at building arguments, including arguments for what you already believe. Expertise doesn’t switch the bias off.
• Confusing more evidence with better evidence. Ten confirming anecdotes that would exist whether or not you’re right are worth less than one real test.
• Only checking your opponents for it. It’s easy to see confirmation bias in people you disagree with. The useful practice is spotting it in your own reasoning.
• Seeking “balance” by reading weak opponents. Reading a bad argument against your view and finding it bad only confirms the view further.
• Treating the bias as a character flaw. It’s a default way of thinking. Fix the process (questions, criteria, records), not your self-image.

Confirmation bias vs motivated reasoning

The two overlap heavily in real life. When you care about the answer, motivated reasoning supplies the reason and confirmation bias supplies the method.

The Mentarium take

Most advice about confirmation bias says “keep an open mind.” That doesn’t work, because your mind already feels open. Confirmation bias doesn’t feel like bias; it feels like noticing that you’re right.

What works is structure that operates before the bias gets a chance. Confirmation bias appears on three of the four Bias Radar lists (Money, Relationships and Business), which tells you how often it matters. In the Mentarium Decision Loop (Frame → Map → Stress-test → Decide → Review), the Stress-test step exists mainly for this bias: run inversion against your favorite option and write down the kill criteria before you look at new evidence.

The Wason task carries a lesson that goes beyond psychology. Every “yes” those participants received was true. None of it was useful. When everything you check comes back agreeing with you, don’t relax. Ask whether you’ve been running tests that could ever have said no.