Hindsight Bias

Hindsight bias is the tendency to see a past event as having been more predictable than it actually was, once its outcome is known. After the fact, people revise their memory of what they had expected toward what in fact happened, so a result that genuinely surprised them comes to feel as though it had been obvious in advance. It is sometimes called the "knew-it-all-along" effect, and it is one of the most reliably reproduced findings in the psychology of judgment.

Hindsight bias has more than one component. Part of it is a genuine memory distortion: learning an outcome changes how the earlier evidence is recalled and reconstructed, so the recollected prediction drifts toward the known result. Part of it is a sense of inevitability, the feeling that events could only have unfolded one way. And part of it is a motivated element, because remembering oneself as having been right is more comfortable than remembering a miss.

The practical damage is to learning. Learning from an outcome requires an honest comparison between what was expected and what occurred. Hindsight bias erases the gap by editing the expectation after the fact, so a forecaster who was wrong remembers having been roughly right, draws no lesson, and repeats the error. In investing this shows up after every large market move. A crash that almost no one positioned for is described a month later as the obvious consequence of warning signs that were "clearly" there, and the people who did not act on those signs nonetheless remember having seen them.

Because hindsight bias makes a person's past judgment look more accurate than it was, it is a direct feeder of overconfidence bias: an inflated memory of a track record inflates trust in the next call. It also interacts with confirmation bias, since the evidence that gets recalled as having pointed to the outcome is disproportionately the evidence that now fits it.

The countermeasure is to remove the memory from the loop. A prediction written down before the outcome is known, with the reasons and the probability attached, cannot be edited by later knowledge. Reviewing that record against what happened is the only way to see the real gap between expectation and result, which is what a decision journal is for.

The effect can be shown by comparing a prediction made before an outcome with the memory of that prediction made afterward.

A hypothetical illustration. Before an earnings announcement, an investor writes that a company has maybe a 40 percent chance of beating expectations. The company beats, and the stock jumps. Asked weeks later what probability they had assigned, the same investor recalls something like 70 percent, and feels they had basically seen it coming. The 40 they actually wrote is gone from memory, replaced by a number that fits the result. If they had never written the 40 down, there would be no way to catch the revision, and they would carry away a false lesson: that their read on the company had been sharp, when it had in fact been close to a coin flip. The written record is the only thing that preserves the original 40 and exposes the edit.

The one upside

• A sense that the world is coherent and comprehensible is not worthless, and hindsight bias partly reflects the useful work of fitting events into a story.

The costs

• It blocks learning from experience by erasing the gap between what was expected and what occurred.
• It inflates confidence in one's own forecasting ability, feeding overtrading and oversized bets.
• It fuels unfair blame, because a decision that was reasonable given what was known at the time looks negligent once the outcome is known.
• It is nearly impossible to detect from the inside, since the edited memory feels exactly like an accurate one.