Look for evidence that could change your mind

Confirmation bias involves seeking or interpreting information in a way that favors a belief already in hand. It can operate even when the belief concerns fairness. Once a team believes a person is resistant, questions may focus only on objections, while helpful suggestions disappear from memory. Once a team believes its process is inclusive, only positive comments may be circulated. Nickerson's review describes many forms of this tendency. The practical implication is to design a search that could reveal a mistake, not merely gather more material on the same side.

A disconfirming question must be capable of changing the conclusion. 'Can you give another example of their resistance?' is confirming. 'When did this person support a comparable change, and what conditions differed?' opens another explanation. Specify the evidence standard in advance. If every contrary example is dismissed as an exception, the belief has become difficult to test. At the same time, one favorable instance need not erase a repeated pattern. Compare frequency, context, seriousness, and the quality of the evidence rather than counting anecdotes as if they were equal.

Darley and Gross demonstrated that background labels could influence interpretation of ambiguous performance in an experimental setting. The study does not tell you the motive behind a particular workplace judgment. It does give a reason to separate background impressions from the actual evidence being reviewed. One practice is to describe the work sample before discussing the person. Another is to ask two reviewers to write an initial assessment against agreed criteria before comparing impressions. These practices do not remove all bias. They make the reasoning visible enough to question and revise, which is more useful than asking reviewers simply to promise objectivity.

A fictional team believes residents never use a particular information format. Its evidence is three complaints and a staff member's impression. The team has not reviewed usage, whether the format was offered, or whether the alternative was accessible. A colleague proposes asking only people who complained why the format failed. Another suggests examining both successful and unsuccessful uses before removing it. The decision affects people who have not been present in the discussion.

Choose a low-stakes belief about a process. What evidence would you accept against it?

Write a proposed conclusion, its strongest available evidence, one alternative explanation, and the information that would change your recommendation.