Why Do You Only Find Evidence for What You Already Believe?

Show someone evidence that contradicts something they strongly believe.

Watch what happens.

They don’t update their belief. They question the evidence. They find flaws in the methodology. They wonder about the motivations of the people who produced it. And by the end, they often believe their original position more strongly than before — because now they’ve had to defend it.

This is not unique to uninformed people or bad-faith argumentation. It happens in laboratories, in published research, and in your own head, several times a day, on topics you consider yourself to be reasonable about.

The Wason Selection Task

In 1960, psychologist Peter Wason published what has become one of the most cited problems in the history of cognitive psychology.

Subjects were shown four cards. Each card had a letter on one side and a number on the other. The cards showed: E, K, 4, 7. They were told the rule: If a card has a vowel on one side, it has an even number on the other.

The question: which cards do you need to flip to test whether the rule is true?

Most people answer E and 4. The correct answer is E and 7.

To test the rule, you need to check whether any vowel lacks an even number (flip E) and whether any odd number has a vowel (flip 7). The 4 is irrelevant — finding a vowel behind it confirms nothing; the rule doesn’t say vowels are the only things with even numbers. But flipping 4 feels like testing the rule because it looks for the pattern.

Only about 10% of people get this right on the first try.

Wason’s insight: human reasoning is systematically skewed toward confirmation. We test hypotheses by looking for examples that match them, not by looking for evidence that would falsify them. The concept of falsification — what would disprove this? — is counterintuitive. The brain defaults to: what would prove this?

Attitude Polarization

In 1979, Charles Lord, Lee Ross, and Mark Lepper at Stanford published a study that showed something more troubling: not only do people seek confirming evidence, they evaluate evidence differently depending on whether it confirms or contradicts their prior beliefs.

They showed subjects — half pro-capital punishment, half anti — the same two studies. One study supported the deterrent effect of capital punishment. The other challenged it. Both studies were presented with equal methodology details.

Subjects rated the study that supported their prior view as more methodologically rigorous and convincing than the one that contradicted it. The same methodology, evaluated in opposite directions based on which conclusion it supported.

More disturbing: after reading both studies — the same two studies — both groups became more extreme in their positions, not less. The evidence that might have created common ground instead drove the groups further apart.

The researchers called this attitude polarization. Mixed evidence doesn’t split the difference; it sharpens division. Each side found the confirming study compelling and the disconfirming study flawed.

Why Smarter Isn’t Better

You might expect that higher intelligence protects against confirmation bias. The research suggests it does not.

Political scientist Philip Tetlock, in decades of research on expert political prediction, found that the most confident and articulate experts — the ones with the most sophisticated reasoning — were often no better than chance at predicting political outcomes. More to the point: they generated more elaborate justifications for their prior positions.

Dan Kahan at Yale Law School has conducted research on what he calls “motivated reasoning” — the use of high cognitive ability in service of confirming existing beliefs rather than updating them. He finds that people with high math and science literacy perform better at analyzing data when it confirms their priors and worse when it challenges them — the opposite of what you’d expect from pure reasoning.

The intelligent brain is not a neutral reasoning engine. It’s a motivated advocate for its current beliefs. Greater cognitive ability produces more effective motivated reasoning, not less.

The Architecture of Confirmation Bias

Several mechanisms contribute to the pattern.

Selective exposure. People seek out information sources that tend to confirm their existing views and avoid those that challenge them. This is partly deliberate and partly a consequence of how recommendation algorithms work — they optimize for engagement, and confirming content tends to produce more engagement than disconfirming content.

Selective recall. Memory is reconstructive. When you try to remember evidence on a topic, you retrieve the examples that support your current position more readily than those that challenge it. The confirming cases are better encoded (because they were processed as coherent with your existing beliefs) and more accessible.

Selective interpretation. When confronted with ambiguous evidence, the brain interprets it in whichever direction fits the current belief. This is not deliberate dishonesty — it is the default output of a pattern-completion system that is trying to be efficient.

Differential scrutiny. Confirming evidence passes quickly. Disconfirming evidence triggers critical analysis. The result: confirming evidence accumulates; disconfirming evidence gets picked apart.

Why Debates Don’t Work

If confirmation bias operates through these mechanisms, the structure of most debates becomes predictable.

When you argue with someone about a strongly held belief, you are providing disconfirming evidence in an adversarial context. Their brain applies maximum scrutiny to your arguments while producing minimum scrutiny for its own. You may construct perfect counterarguments. They will find flaws in each one. You will both leave more confident than you entered.

Research on actual belief change suggests it happens differently than debates imply: over time, through quiet exposure to new experiences, through relationships with people who hold different views (not through confrontation), and crucially through sleep — REM sleep appears to play a role in integrating new, potentially conflicting information into memory without the defensive arousal present in waking debates.

The One Practice That Helps

The most evidence-backed approach to counteracting confirmation bias involves a specific, counterintuitive exercise: actively generate the strongest version of the opposing argument yourself.

This is called steel-manning — the opposite of straw-manning. The research shows that trying to articulate why you might be wrong, or why someone who disagrees with you might have a legitimate point, activates different neural systems than the default defensive reasoning mode. It does not eliminate bias, but it measurably reduces it.

The reason it works is that it forces you to do what the brain normally refuses to do: take disconfirming information seriously, on its own terms, without the adversarial context that triggers defensive processing.

You believe most things for good reasons. A few things you believe for bad reasons, but convincingly bad ones. The trouble is that the feeling of having a good reason is identical in both cases.

The question is not whether you have confirmation bias. You do. Everyone does.

The question is whether you have ever tried to find a case where you were wrong — and if so, what you did when you found one.