Defining Confirmation Bias and Its Cognitive Mechanism
Confirmation bias is one of the most pervasive and powerful biases in human cognition. It refers to the tendency to search selectively for information that supports the beliefs, hypotheses, and expectations a person already holds, to attend to that information preferentially, to remember it more readily, and to interpret ambiguous material in its favor. Peter Wason's 2-4-6 task, introduced in 1960, was the first experimental demonstration of the bias: participants asked to discover a numerical rule generated test cases that fit their current guess rather than cases that could have shown the guess to be wrong. Confirmation bias operates at three distinct levels. The first is bias in information search, in which a person looks first and hardest for evidence that would confirm the hypothesis and does not go looking for evidence that would disconfirm it. The second is bias in interpretation, in which ambiguous information is read in whichever direction fits the existing belief. The third is bias in memory, in which supporting information is retained better while contradictory information is forgotten more easily. Because the three levels compound one another, a belief can grow more confident over time even when the actual evidence encountered has been neutral. In evolutionary terms, confirmation bias is thought to have been an efficient information-processing strategy, since acting quickly on a working hypothesis is usually better than suspending judgment while gathering complete data. In judgments that require weighing complex and mixed evidence, however, that same efficiency becomes a source of systematic error, and the errors are difficult to notice from the inside because the reasoning feels like ordinary observation.
Confirmation Bias in Close Relationships
Confirmation bias exerts a particularly strong influence in close relationships, where the same person is observed continuously over years and where almost every behavior admits of more than one reading. Once a first impression has formed, subsequent information tends to be interpreted in whatever direction confirms it. If a person holds the belief that a partner is trustworthy, behavior that supports trust attracts attention while behavior that contradicts it is discounted as insignificant. If instead the belief is that the partner is cold, behavior showing coldness is noticed while warm behavior is filed away as an exception that requires no revision of the overall judgment. The concepts of positive sentiment override and negative sentiment override, widely used in Gottman's work on couples, can be understood as the relationship-specific expression of confirmation bias. Stable couples interpret a partner's behavior charitably, which is positive sentiment override, while unstable couples interpret the same behavior negatively, which is negative sentiment override. The two states are self-perpetuating for the same reason any confirmation bias is: each interpretation supplies further evidence for the belief that produced it. Once a couple has fallen into negative sentiment override, recovery becomes difficult, because even the partner's positive behavior is reinterpreted as having some ulterior motive behind it. A gesture intended as repair is received as manipulation, so the ordinary means by which couples correct their course stops functioning, and the deterioration continues without either partner intending it.
Reading a Compatibility Result Without Confirmation Bias
Confirmation bias is also at work when you read the result of a compatibility assessment. A favorable result may lead you to attend only to your partner's good qualities and to overlook genuine problems, treating the result as license to stop paying attention. An unfavorable one may lead you to overweight minor friction and give up on a relationship that was workable, since every subsequent disagreement now arrives as confirmation of the verdict. The mechanism is the same in both cases: a single belief, the result itself, begins to select what you notice and what you remember. That is why a result is worth reading for its specifics, the strengths and the challenges alike, rather than filed away as a compatible-or-incompatible label, since a single categorical label is exactly the sort of belief that confirmation bias amplifies most efficiently. Simply pausing when you finish and asking whether you have begun interpreting your partner's behavior in a slanted way is enough to reduce the influence of the bias. That pause works not by eliminating the tendency, which is not achievable, but by inserting a moment of deliberate review at the point where automatic interpretation would otherwise proceed unchecked.