18.5: Confirmation Bias

Many studies (as well as a bit of careful observation) document our tendency to look for, remember, and acknowledge the value of positive evidence that supports our beliefs, while overlooking or undervaluing negative evidence that tells against them. The distinction between positive and negative evidence may be clearer if we consider a couple of examples.

• Suppose that you believe (or hypothesize) that all swans are white.

Then a white swan is positive evidence for your view; it confirms or supports (though it doesn’t conclusively establish) your belief. By contrast, swans that are not white are negative evidence against your view; they disconfirm your belief. In fact, even one non-white swan shows that your belief is false. One black swan falsifies your hypothesis right then and there.

In many cases, though, negative evidence disconfirms a hypothesis without utterly refuting it. For example, Wilbur might wonder whether Wilma has a crush on him. The fact that she goes out of her way to chat with him is confirming, though by no means conclusive, evidence that she does. And the fact that she sometimes seems to avoid him is disconfirming evidence, though it doesn’t prove that she doesn’t.

Confirmation bias is our common tendency to look for, notice, and remember confirming or positive evidence (that supports what we think) while overlooking or downplaying disconfirming or negative evidence (which suggests that what we think is wrong). For example, if Wilbur is already convinced that women are bad drivers, he may be more likely to notice or remember cases where women drove badly and to overlook or forget cases where they drove well.

The selective thinking exhibited in this bias makes for bad reasoning, because it allows us to support our views without running the risk of finding out that they are wrong. It really boils down to considering only those points that support our own views, which is about as far from being open minded as possible. Careful reasoning requires testing our views to see if they fit the facts. The confirmation bias also encourages beliefs in illusory correlations (Chapter 15), since it encourages us to look for cases where two variables do go together without looking for cases where they may not.

Since the confirmation bias often leads to bad reasoning (that’s why it’s called a bias), it is important to avoid it. It should help to convince yourself of the value of negative evidence and to make a practice of looking for it, but this bias is difficult to eliminate.

In a series of experiments, Mynatt, Doherty, and Tweney had subjects attempt to determine whether various laws or generalizations about the movement of a spot on a computer screen were true or not. They asked some people to confirm various generalizations about the dot’s movement, others to disconfirm them, and still others just to test them. The instructions to disconfirm were not effective; about 70% of the time people in each group looked for confirming evidence. And the ineffectiveness of asking people to look for disconfirming evidence persisted even when the value and importance of searching for such evidence were explained to them before they began their task.