Hindsight Bias

Dendrology and hindsight bias (or, Give the chainsaw a rest)

One of the goals of mental time travel is keeping events in perspective. To understand an overriding risk to that perspective, think about time as a tree. The tree has a trunk, branches at the top, and the place where the trunk meets the branches. The trunk is the past. A tree has only one, growing trunk, just as we have only one, accumulating past. The branches are the potential futures. Thicker branches are the equivalent of more probable futures, thinner branches are less probable ones. The place where the top of the trunk meets the branches is the present. There are many futures, many branches of the tree, but only one past, one trunk.

As the future becomes the past, what happens to all those branches? The ever-advancing present acts like a chainsaw. When one of those many branches happens to be the way things turn out, when that branch transitions into the past, present-us cuts off all those other branches that didn’t materialize and obliterates them. When we look into the past and see only the thing that happened, it seems to have been inevitable. Why wouldn’t it seem inevitable from that vantage point? Even the smallest of twigs, the most improbable of futures—like the 2%–3% chance Russell Wilson would throw that interception—expands when it becomes part of the mighty trunk. That 2%–3%, in hindsight, becomes 100%, and all the other branches, no matter how thick they were, disappear from view.

That’s hindsight bias, an enemy of probabilistic thinking. Psychologists also call it the “I knew it all along” effect or creeping determinism. It is the tendency to believe, once an event has occurred, that it was predictable or even inevitable—distorting our memory of what we actually knew beforehand. The bias applies whether the outcome is positive or negative. In everyday life, it manifests in small but telling ways: Mike, having just had his car painted, notices a scratch on the door and exclaims, “I just got the car painted. I knew this was going to happen!” Only after seeing the scratch does he convince himself he had foreseen it.

Memory does not replay history like a VCR. Later information can change which earlier details we recall and how we interpret them. Evidence consistent with the outcome becomes prominent, while conflicting clues and earlier doubts fade. This produces an impression of historical determinism: the past looks more predictable than it felt at the time. We may imagine that people living through the 1929 stock market crash immediately understood its historical significance. Life, however, does not provide a dramatic movie score to announce which events will matter.

The consequences of chainsawing the branches

In the court’s opinion in Jentz v. ConAgra Foods, Judge Frank Easterbrook warned about assessing probabilities after one of the potential futures has occurred. Jentz v. ConAgra Foods involved a “hot” grain bin owned by ConAgra and the company it hired (named West Side) to investigate and deal with the cause of the burning smell, smoke, and raised temperatures from the bin. After failing to solve the problem, West Side’s foreman told ConAgra to call the fire department. He then told some of his employees to remove tools from a tunnel to the bin that could impair firefighters’ access.

While the workers were in the tunnel, the bin exploded, severely injuring the West Side employees. Those injured sued ConAgra and West Side. The jury awarded $180 million in compensatory and punitive damages. In addressing the punitive damages against West Side, Judge Easterbrook, writing for the court, pointed out that Illinois law required a “gross deviation” from the standard of care to award punitive damages. Finding no evidence in the record of any foreseeable likelihood of explosion at the time the foreman ordered the workers to remove tools from the tunnel, he concluded, “The verdict appears to be a consequence of hindsight bias—the human tendency to believe that whatever happened was bound to happen, and that everyone must have known it. If [the foreman] believed that an explosion was imminent, then he is a monster; but of that there is no evidence. Hindsight bias is not enough to support a verdict.”

Once we know there was an explosion, it is difficult to assess the situation as it appeared beforehand. The jurors heard about the men entering the tunnel with the outcome already known. The opinion identifies the danger of treating that outcome as inevitable; it does not let us reconstruct each juror’s actual reasoning.

A similar pattern appears in business. A civil servant assigned to review a group of traders’ transactions analyzed the records and found that only about 1% of the trades had generated significant profits; the rest were either small winners or outright losers. Shocked, he concluded that the traders should simply have done more of the winning trades and fewer of the losers. It was, to him, an obvious improvement—so obvious that he could not understand why highly paid professionals had not seen it themselves. But the civil servant was looking at the data after the fact, when the outcomes were already known. The trades that ended up losing he labeled “gross mistakes,” much as journalists describe a candidate’s decision that preceded an election loss as a “mistake.” Yet a mistake cannot be determined solely by its outcome; it must be judged against the information available at the time the decision was made.

Steve Bartman could have used a stadium full of Cubs fans with Judge Easterbrook’s perspective. They would have recognized that the Cubs losing the game was only one of many ways it could have turned out. After Bartman and others reached for the foul ball, the Cubs could still have won. The loss also involved later plays: the pitching ace gave up hits, and the shortstop committed an error on a potential inning-ending double play. Blaming Bartman alone erases those remaining possibilities. The final result does not by itself establish how much his action contributed to the loss.

Pete Carroll and the world of Monday Morning Quarterbacks could have used the judge’s reminder that we tend to assume that, once something happens, it was bound to happen. If we don’t try to hold all the potential futures in mind before one of them happens, it becomes almost impossible to realistically evaluate decisions or probabilities after.

In another case, a CEO regretted firing his company’s president. Although he had initially characterized the decision as one of his worst, when we reconstructed the tree, essentially picking the branches up off the ground and reattaching them, it was clear that he and his company had made a series of careful, deliberative decisions. Because they led to a negative result, however, the CEO had been consumed with regret. When he looked back at his decisions, he couldn’t see all those branches and their likelihoods. He could only see the trunk. All he saw was a bad result.

The inevitability illusion

The bias has been demonstrated in controlled settings. In one experiment, participants were asked whether a city should be held responsible for paying for a bridge. Those given only the limited information available at the time of the decision were far less likely to think the city would pay than those who were also told the eventual negative outcome: sediment blocked the river and caused flood damage. Knowing the result inflated the perceived predictability of that result. In another classic study from the 1970s, people estimated the likelihood of various diplomatic outcomes before President Nixon’s trips to China and the Soviet Union. After the trips, when asked to recall their original predictions, participants consistently overestimated the probabilities they had assigned to what actually happened and underestimated those for what did not—convincing themselves they had foreseen it all.

Modern business history offers the same illusion. The rise of Apple and Google may now seem inevitable, but that inevitability was far from obvious to competitors like Nokia in 2003. Once the success became established, many people retrospectively believed it had always been predictable, ignoring the genuine uncertainty that existed beforehand.

Reaction to the 2016 election provides another strong demonstration of what happens when we lop branches off the tree. Hillary Clinton had been favored going into the election, and her probability of winning, based on an accumulation of the polls, was somewhere between 60% and 70%, according to FiveThirtyEight.com. When Donald Trump won, pollsters got the Pete Carroll treatment, maybe no one more than Nate Silver, founder of FiveThirtyEight.com and a thoughtful analyzer of polling data. (“Nate Silver was wrong.” “The pollsters missed it.” “Just like Brexit, the bookies blew it.” Etc.) The press spun this as a certain win for Clinton, despite the Trump branch of the tree being no mere twig at 30%–40%. By the day after the election, the Clinton branch had been severed, only the Trump branch remained, and how could pollsters and the polling process have been so blind?

The CEO removed all those other branches with a chainsaw and ran them through a wood-chipper. They disappeared and he acted as if they never existed. That’s what hindsight bias is, and we’re all running amok through the forest with a chainsaw once we get an outcome. Once something occurs, we no longer think of it as probabilistic—or as ever having been probabilistic. This is how we get into the frame of mind where we say, “I should have known” or “I told you so.” This is where unproductive regret comes from. A deeper consequence is that it fosters overconfidence. Those who are adept at predicting the past often convince themselves they are equally adept at predicting the future. After the shock of a catastrophic event—from the September 11 attacks to financial crises—many will insist the warning signs were clear all along and that next time they will see it coming. The bias stops us from learning honestly from new experiences, making us overconfident in our supposed predictive abilities and leading us to take larger risks than we otherwise would.

Hindsight bias can encourage unfair blame and overconfidence. A journalist’s account of a business leader’s “timing” or an economist’s confident explanation of an earlier surprise can sound persuasive without showing that either could have predicted the outcome. Compare such accounts with what was recorded beforehand.

Keeping the branches in view

By keeping an accurate representation of what could have happened (and not a version edited by hindsight), memorializing the scenario plans and decision trees we create through good planning process, we can be better calibrators going forward. We can also be happier by recognizing and getting comfortable with the uncertainty of the world. Instead of living at extremes, we can find contentment with doing our best under uncertain circumstances, and being committed to improving from our experience.

One practical technique is to reconstruct plausible alternative outcomes and ask what evidence supported each at the time. Compare this reconstruction with any forecasts or decision notes made beforehand. That helps distinguish what was foreseeable from what merely seems obvious now.

One of the things poker teaches is that we have to take satisfaction in assessing the probabilities of different outcomes given the decisions under consideration and in executing the bet we think is best. With the constant stream of decisions and outcomes under uncertain conditions, you get used to losing a lot. To some degree, we’re all outcome junkies, but the more we wean ourselves from that addiction, the happier we’ll be. None of us is guaranteed a favorable outcome, and we’re all going to experience plenty of unfavorable ones. We can always, however, make a good bet. And even when we make a bad bet, we usually get a second chance because we can learn from the experience and make a better bet the next time.

Life, like poker, is one long game, and there are going to be a lot of losses, even after making the best possible bets. We are going to do better, and be happier, if we start by recognizing that we’ll never be sure of the future. That changes our task from trying to be right every time, an impossible job, to navigating our way through the uncertainty by calibrating our beliefs to move toward, little by little, a more accurate and objective representation of the world. With strategic foresight and perspective, that’s manageable work. If we keep learning and calibrating, we might even get good at it.