“Judgments about what is good and what is bad, what is worthwhile and what is a waste of talent, what is useful and what is less so, are judgments that seldom can be made in the present. They can safely be made only by posterity.” — Tulving
Hindsight bias occurs when we look backward in time and see events are more predictable than they were at the time a decision was made. This bias, also known as the “knew-it-all-along effect,” typically involves those annoying “I told you so” people who never really told you anything.
For instance, consider driving in the car with your partner and coming to a T in the road. Your partner decides to turn right and 4 miles down the road when you realize you are lost you think “I knew we should have taken that left.”
Hindsight bias can offer a number of benefits in the short run. For instance, it can be flattering to believe that our judgment is better than it actually is. And, of course, hindsight bias allows us to participate in one of our favorite pastimes — criticizing the decisions of others for their lack of foresight.
Aside from helping aid in a more objective reflection of decisions, hindsight bias also has several practical implications. For example, consider someone asked to review a paper but knows the results of the previous review from someone else? Or a physician asked for a second opinion after knowing the results of the first. The results of these actions will likely be biased by some degree. Once we know an outcome it becomes easy to find some plausible explanation.
Hindsight bias helps us become less accountable for our decisions, less critical of ourselves, and over-confident in our ability to make decisions.
One of the most interesting things I discovered when researching hindsight bias was the impact on our legal system and the perceptions of jurors.
Harvard Professor Max Bazerman offers:
The processes that give rise to anchoring and overconfidence are also at play with the hindsight bias. According to this explanation, knowledge of an event’s outcome works as an anchor by which individuals interpret their prior judgments of the event’s likelihood. Due to the selective accessibility of the confirmatory information during information retrieval, adjustments to anchors are inadequate. Consequently, hindsight knowledge biases our perceptions of what we remember knowing in foresight. Furthermore, to the extent that various pieces of data about the event vary in support of actual outcome, evidence that is consistent with the known outcome may become cognitively more salient and thus more available in memory. This tendency will lead an individual to justify a claimed foresight in view of “the facts provided.” Finally, the relevance of a particular piece of that may later be judged important to the extent to which it is representative of the final observed outcome.
In Cognitive Illusions, Rudiger Pohl offered the following explanations of hindsight bias:
Most prominent among the proposes explanations are cognitive accounts which assume that hindsight bias results from an inability to ignore the solution. Among the early approaches are the following three: (1) Fischhoff (1975) assumed an immediate and irreversible assimilation of the solution into one’s knowledge base. As a consequence, the reconstructed estimate will be biased towards the solution. (2) Tversky and Kahneman (1974) proposed a cognitive heuristic for the anchoring effected, named anchoring and insufficient adjustment. The same mechanism may apply here, if the solution is assumed to serve as an “anchor” in the reconstruction process. The reconstruction starts from this anchor and is then adjusted in the direction of one’s knowledge base. However, this adjustment process may stop too early, for example at the point where the first plausible value is reached, thus ending to a biased reconstruction. (3) Hell (1988) argued that the relative trace strengths of the regional estimate and of the solution might predict the amount of hindsight bias. The stronger the trace strength of the solution relative to that of the original estimate, the larger hindsight bias should be.
Pohl also offers an evolutionary explanation of hindsight bias:
Finally, some authors argued that hindsight bias is not necessarily a bothersome consequence of a “faulty” information process system, but that is may rather represent an unavoidable by-product of an evolutionary evolved function, namely adaptive learning. According to this view, hindsight bias is seen as the consequence of our most valuable ability to update previously held knowledge. This may be seen as a necessary process in order to prevent memory overload and thus to maintain normal cognitive functioning. Besides, updating allows us to keep our knowledge more coherent and to draw better inferences.
Ziva Junda, in social cognition, offers the following explanation of why hindsight bias occurs:
Preceding events take on new meaning and importance as they are made to cohere with the known outcome. Now that we know the our friends have filed for divorce, any ambiguous behavior we have seen is reinterpreted as indicative of tension, any disagreement gains significance, and any signs of affection seem irrelevant. It now seems obvious that their marriage was doomed from the start…Moreover, having adjusted our interpretations in light of current knowledge, it is difficult to imagine how things could have happened differently.
When making likelihood judgments, we often rely on the availability heuristic: The more difficult it is for us to imagine an outcome, the more unlikely it seems. Therefore, the difficulty we experience imagining how things might have turned out differently makes us all the more convinced that the outcomes that did occur were bound to have occurred.
Hindsight bias has large implications for criminal trials. In Jury Selection Hale Starr and Mark McCormick offer the following:
The effects of hindsight bias – which result in being held to a higher standard – are most critical for both criminal and civil defendants. The defense is more susceptible to the hindsight bias since their actions are generally the ones being evaluated fro reasonableness in foresight-foreseeability. When jurors perceive that the results of particular actions were “reasonably” more likely after the outcome is known, defendants are judged as having been capable of knowing more than they knew at the time the action was taken and therefore as capable of preventing the “bad” outcome.
In post-verdict surveys jurors unknowingly demonstrate some of the effects of hindsight bias:
“I can’t understand why the managers didn’t try to get more information or use the information they had available. They should have known there would be safety problems at the plant”.
“The defendants should have known people would remove the safety shield around the tire. There should have been warnings so people wouldn’t do that”
“Even though he was a kid, he should have known that once he showed the others who had been drinking that he had a gun, things would get out of hand. He should have known guns invited violence.
Jurors influenced by the hindsight bias look at the evidence presented and determine that the defendants knew or should have known their actions were unsafe, unwise, or created a dangerous situation. Hindsight bias often results in the judgment that the event was “an accident or tragedy waiting to happen.”
Protection Against Hindsight Bias
In Principles of Forecasting, Jon Scott Armstrong, offers the following advice on how to protect yourself:
The surest protection against (hindsight bias) is disciplining ourselves to make explicit predictions, showing what we did in fact know (sounds like a decision journal). That record can also provide us with some protection against those individuals who are wont to second guess us, producing exaggerated claims of what we should have known (and perhaps should have told them). If these observers look to this record, it may show them that we are generally less proficient as forecaster than they would like while protecting us against charges of having blown a particular assignment. Having an explicit record can also protect us against overconfidence in our own forecasting ability: If we feel that we “knew all along” what was going to happen, then it is natural enough to think that we will have similar success in the future. Unfortunately, an exaggerated perception of a surprise-free past maybe portend a surprised-full future.
Documenting the reasons we made a forecast makes it possible for us to know not only how well the forecast did, but also where it went astray. For example, subsequent experiences may show that we used wrong (or misunderstood) inputs. In that case, we can, in principle, rerun the forecasting process with better inputs and assess the accuracy of our (retrospectively) revised forecasts. Perhaps we did have the right theory and procedures, but were applying them to a mistaken picture of then-current conditions…Of course inputs are also subject to hindsight bias, hence we need to record them explicitly as well. The essence of making sense out of outcome knowledge is reinterpreting the processes and conditions that produced the reported event.
Hindsight Bias is part of the Farnam Street latticework of mental models.