"One of the most systematic errors in human perception is what psychologists call hindsight bias -- the feeling, after an event happens, that we knew all along it was going to happen. Across a wide spectrum of issues, from politics to the vagaries of the stock market, experiments show that once people know something, they readily believe they knew it all along. "This is not to say that no one predicted the war in Iraq would go badly, or that the insurgency would last so long. Many did. But where people might once have called such scenarios possible, or even likely, many will now be certain that they had known for sure that this was the only possible outcome." --Shankar Vedantam, Washington Post
"This is not to say that no one predicted the war in Iraq would go badly, or that the insurgency would last so long. Many did. But where people might once have called such scenarios possible, or even likely, many will now be certain that they had known for sure that this was the only possible outcome." --Shankar Vedantam, Washington Post
Are health insurance companies guilty of denying care or, as Shannon Brownlee's new book argues, are they failing to prevent Overtreatment?
When it comes to surveillance and terrorism, are we not doing enough to prevent the next terrorist attack, or are we collecting too much unnecessary information?
In the mortgage market, are we making it too difficult or too easy for first-time homebuyers to purchase homes?
Would one have expected that the cost of restoring order after an invasion of Afghanistan to be relatively light compared to the cost of restoring order after an invasion of Iraq?
Most people approach these issues with hindsight bias. In situations of uncertainty, hindsight bias causes a number of problems.
First, it sets up a false set of expectations about what government can do. For example, people believe that government can regulate financial markets so that there are never any bubbles or miscalculations, no decent borrower is turned down, and no one ever defaults on a loan.
People fail to recognize trade-offs. For example, Brownlee writes (p. 10),
"in politics, overtreatment is routinely left out of any discussion of health care reform. That's partly because getting rid of it smacks of rationing. But rationing is when you deny patients health care that could potentially help them...Rationing is when you limit the number of MRI machines in order to discourage doctors from ordering an MRI test for a patient. But getting rid of overtreatment, care that's useless and potentially harmful? That isn't rationing; that's improving the quality of medicine."
It's also hindsight bias. We do not know ahead of time whether that "unnecessary" MRI is going to discover a malignant tumor and save a patient's life. Such was the case in an example that I used as an anecdote in my book, Crisis of Abundance, in order to illustrate the dilemmas and trade-offs in modern medicine.
As I write this, my mother-in-law just underwent knee replacement surgery. This could turn out to be "useless and potentially harmful," or it could enable her to use the toilet herself and thereby postpone the day when she needs to be in a nursing home.
There is a great deal in Brownlee's book that is insightful, and I plan to return to it in future essays. But its value is diminished because of pervasive hindsight bias. She writes as if the errors of overtreatment can be avoided at no cost. In fact, there are a number of potential costs. The cost might involve undertreatment for some patients. Or it could involve putting significant resources into medical decision-making.
It is fair to say that the biggest challenge for health care policy is to make the best trade-off among these costs -- the cost of overtreatment, the cost of undertreatment, and the cost of resources used in the decision-making process. Hindsight bias instead makes it appear as if such a trade-off does not exist, and that only ignorance and greed are preventing better outcomes.
Recently, the lead story in the Washington Post told about surveillance of travelers. The story appears to be designed to foment protest and alarm over unnecessary and intrusive spying by the government.
My guess is that Congress, using state-of-the-art hindsight, will investigate this program and have it canceled or curtailed, as it has done to other surveillance and data mining efforts. Then, when our intelligence agencies fail to prevent the next terrorist attack, Congress will launch another hindsight-driven commission to probe the reason for failure.
With hindsight bias, our terrorism policy is going to be that of an amateur shower-taker -- alternately scalding ourselves with the failure to prevent attacks or freezing ourselves with abusive collection of data. A better approach would be to focus on the most effective way to collect data and to prevent its abuse, while recognizing that there can be no perfect solution. As outlined in an earlier essay, The Constitution of Surveillance recognizes trade-offs and offers an alternative to hindsight bias for addressing this challenge.
In hindsight, most people made the right decision about Afghanistan and Iraq. That is, in hindsight, most people are in favor of invading the former and against invading the latter.
However, go back to 2001. Which country is going to pose more long-term problems for an invader -- Afghanistan or Iraq? Afghanistan, with its mountainous territory and history of rebellion against the Soviets, probably would have been viewed by many experts as posing the greatest difficulty. At the time of the invasion, many feared that it would prove disastrous, and in fact in October of 2001 the New York Times pronounced it a quagmire.
In fact, the unexpectedly low cost of invading Afghanistan may have been one of the reasons for the unexpectedly high cost of invading Iraq. The Bush Administration probably based its expectations of the latter on the outcome of the former.
For now, it is not clear what is the best strategy in Iraq. Some argue that the larger the role that Americans take in the war, the less incentive for the Iraqi government to address difficult issues. Others argue that without a major American presence, security will deteriorate and the country will sink into sectarian violence. Years from now, we may know the answer to these and other questions. And with hindsight bias, we will wonder how those who were on the wrong side of the issue could have been so blind. Meanwhile, real decisions have to be made with imperfect information.
Humans have a cognitive blind spot when it comes to making decisions under uncertainty. Steven Pinker's wonderful new book, The Stuff of Thought, discusses the theory that humans are hard-wired to see the world in terms of substance, space, time, and cause-effect. We are not hard-wired to think in terms of statistics. Pinker writes (p. 85-86)
"When the mind locates one object with respect to another, it is apt to compress the first one into a pinpoint or blob whose shape and parts are no longer discernible, like a thing in a box...I suspect this is one of the reasons people have so much trouble understanding statistical comparisons...[For example] the distributions of talents and temperaments for men and women are not identical...Yet when people hear about this research, they tend [to] mangle it into the claim that every last man is better than every last woman (or vice-versa)...It's as if people heard the statistic that women outlive men on average and concluded that every woman outlives every man."
Concerning hindsight bias, this story describes the work of three psychologists who argue that it is a necessary byproduct of the way that we learn.
"any feedback or correct information a person receives after he has given his initial judgment automatically updates the knowledge base underlying the initial judgment. If a person cannot remember this initial judgment, he will reconstruct it from what he currently knows about the situation. And what he currently knows is the updated version of what he used to know. So while feedback does not directly affect a person's memory for the original response, it indirectly affects the memory by updating the knowledge used to reconstruct the response. Rather than thinking of hindsight bias as a flaw of human cognition, as previous research suggests, Hoffrage, et al. argue that it's a by-product of an adaptive mechanism - one that makes human memory more efficient."
Suppose that I have faced a decision between driving and taking an airplane, and the airplane turned out to be the better choice. What the authors are suggesting is that it is easier for me to remember "I knew that I should have taken the plane" than to remember all of the data and probability calculations that I was making ahead of time.
Economizing on memory may be useful more often than not. Next time I am near a hot stove, it is better for me to make a quick and automatic decision to keep my hand away rather than make calculations of the probability of getting burned based on how far my finger is from the heating element.
However, for complex issues involving uncertainty, hindsight bias is very deceptive. We think that we can stop terrorist attacks without collecting any unnecessary information. We think that we can get a "free lunch" in health care by simply willing doctors to make better decisions about when to use expensive procedures. And we think that the President makes poor decisions about foreign policy, because we always remember ourselves as agreeing with the choices that worked well and disagreeing with those that proved disappointing.
