Top

How to Make the Best Decisions

Artwork By Anja Bartelt

Playing the devil's advocate can lead to bad decision-making. Here's what's better…

One thing you learn as a social psychologist is that we all have biased and inflated beliefs about our own rationality and independence. We all think that we assess information fairly and rationally and that our knowledge coupled with good intentions will lead to good decisions. What social psychology teaches you is that you, like everyone else, select information and consider alternatives in a limited fashion. Years of careful research have shown me how hard it is to make really good decisions.

Training and education certainly help, but neither will save us from biased thinking and poor judgments. A better route to a good decision is for someone who authentically believes differently to directly challenge our beliefs and ways of thinking: to debate with a genuine dissenter. I learned that over many years of studying the behavior of business groups, juries, and even Supreme Court justices.

For many people, perhaps most people, the value of dissent lies almost totally in the outside possibility that the minority opinion might be correct. That is why we listen to the dissenter, at least for a while. What goes unrecognized is that dissent stimulates divergent thinking in the majority even when the dissenter is wrong.

Let me repeat that: We are stimulated to think divergently regardless of the merit of the dissenting position. In my study of the power of authentic dissent, I have seen its ability to stimulate a broader search for information, a consideration of more alternatives, the use of multiple strategies, and more original thinking.

What is fascinating is that people recognize the importance of divergent thinking and still resist dissent as the vehicle for it. They want a mechanism that will clone the stimulating effects of dissent, but without the conflict or dislike between team members that dissent engenders. High on their list of techniques for achieving divergent thinking is playing devil’s advocate, and its sister technique, dialectical inquiry, in which a person doesn’t just argue all the negatives of a proposed position but also offers a counterposition. Recent books such as Tom Kelley’s Ten Faces of Innovation go so far as to suggest that even these forms of “pretend dissent” are too negative and “idea stifling.” They join the chorus advocating agreement and role-playing as the routes to creativity and good decision-making.

Instead, I suggest that if we are looking for the best decision, the real need is for challenge and debate—the curious ways of thinking that are stimulated by authentic disagreement.

Can We Fake Dissent?

Having expressed my doubts about the devil’s advocate for years, I undertook the first study in 1998 when three graduate students at Berkeley expressed their interest. Our study was quite simple: We selected groups of four individuals. In one group there was no dissent. In the second group there was authentic dissent. In the third group there was devil’s advocate “dissent.” To hold the arguments constant, and all communication as well, we had the four individuals in each group seated at a table, separated by partitions. They communicated via computer. They were aware of one another but could not communicate directly, either verbally or nonverbally. All the communication was done via the computers, which we could control.

The individuals “deliberated” a personal injury case and were asked to agree on compensation for “pain and suffering.” They could choose any one of eight options in $75,000 increments from “$1 to $75,000” up to “over $525,000.” From pretesting this case, we knew that they all would pick a low compensation, and they all did. Everyone’s position was below $150,000.

Each person saw the arguments of the others, but in fact all of these arguments were preprogrammed. Each learned on the first ballot that two others agreed with her, favoring a position of low compensation. What differed was whether a third individual, person B, agreed with her as well (the no-dissent condition) or took a different position of high compensation (being then a dissenter).

That “dissent,” however, was either “authentic”—that is, it was the true opinion of person B—or the result of person B being asked by the experimenter to play devil’s advocate. The arguments were identical. After deliberation, each participant listed her thoughts about the case (a technique that has been found to be very useful in tracking the thinking of individuals). These thoughts were coded for quantity and also for whether they were “internal” (coming from the individual) or “external” (paraphrases of others’ comments or the case material). They were also coded for the direction of the thinking. Did the thinking favor the person’s own position, or did it consider both sides of the issue?

In general, those individuals who faced authentic dissent had the most internal thoughts. They themselves were generating the thoughts, rather than paraphrasing others or the case information. They were actually thinking! More important was the direction of their thoughts. Those facing authentic dissent showed a balance in considering both sides of the issue. Those facing a devil’s advocate did not.

Anja Bartelt Parrot Without Edges
Anja Bartelt

In general, those individuals who faced authentic dissent had the most internal thoughts. … They were actually thinking!

Why the Devil Fails

Those facing the devil’s advocate had more thoughts supporting their own position than did those facing authentic dissent. Rather than stimulating thought on both sides of the issue, the devil’s advocate actually stimulated more thoughts in defense of their original position. It was not the kind of thinking that proponents of the devil’s advocate technique would have hypothesized or desired.

The whole point of the devil’s advocate is to get people to consider the downsides as well as the upsides of their preferred position. It appears to do the reverse. Those facing a devil’s advocate seem to be convincing themselves that they were right all along. By contrast, authentic dissent fostered more balanced thinking between the pros and cons of a position.

When I first presented this study at the “Knowledge and the Firm” conference at Berkeley, Dorothy Leonard, a Harvard Business School professor, confessed that she had been arguing for the devil’s advocate technique in her executive education courses for years. She took our findings to heart and promptly incorporated them into her teaching and books. Fortunately, executive education in business schools now includes a healthier skepticism about these role-playing forms of challenge.

A second study took this a step further. We wanted to compare authentic dissent to the devil’s advocate in their impact not just on thinking, but on the solutions they generated. We also wanted to test some questions related to variants on the devil’s advocate. For example, was it important for the advocate’s true position to be known? Did it matter whether the advocate’s true position coincided with what she was arguing as a devil’s advocate? We were especially interested in seeing whether the technique could have the same impact as authentic dissent when there was the closest possible match, namely when the devil’s advocate was known to believe the position she was asked to argue. Most would have predicted that this congruence would clone the effects of authentic dissent. It didn’t.

In this study, we compared authentic dissent to three versions of the devil’s advocate. In one, the devil’s advocate’s own position was unknown. In a second, it was known to be the opposite of what the devil’s advocate was asked to role-play by the experimenter. In other words, she held similar beliefs to the majority but role-played the opposite. In a third, her position was known to be the same as what she was asked to role-play. She actually believed what she was now arguing as a devil’s advocate.

The comparison of most interest was between this third condition—a devil’s advocate arguing a position she actually believed—and the authentic dissent condition, in which she argued the same position—the same way—but had not been asked to be a devil’s advocate. In both of those conditions, she was arguing something she had believed from the beginning—and everyone knew it. The arguments were exactly the same, since they were scripted. The only difference was whether or not she had been asked to be a devil’s advocate.

The four individuals in each group deliberated a management issue for eight rounds. Everyone could see the comments on their computer screen. In the “authentic dissent” condition and in each of the three devil’s advocate conditions, the dissenter was an accomplice of the experimenter and was typing in comments from a prewritten script. The comments were identical for all conditions. After the deliberation, the participants were asked to give as many solutions as possible that would solve the particular management problem. We coded the responses for the number of solutions they provided and the quality of those solutions.

The results were surprising in a couple of ways. Almost every person expected that it would matter whether the devil’s advocate position was known or unknown, and whether she believed the position she was arguing or believed the reverse. The results showed that it made no difference on any measure. None of the versions of the devil’s advocate caused significant differences in the quantity or quality of the solutions that the participants generated. 

Anja Bartelt Turtle Without Edges
Anja Bartelt

It is when you face a dissenter who believes his position, has the courage to say so, and does so persistently that you confront the possibility that you may be wrong. 

The Courage of Conviction

Authentic dissent was the one condition—and the only one—where individuals came up with more solutions. They also came up with more creative solutions, which we measured by a combination of quality and quantity. It was authentic dissent that won hands down, even when compared to the condition where the devil’s advocate argued what she believed (and where the others knew that as well). Remember: This involved believing the same position and giving the same arguments in support of that position. Though seemingly identical to the authentic dissent condition, asking the person to play the devil’s advocate eliminated the benefits for creative problem solving.

Upon reflection, we have identified some of the problems inherent in the devil’s advocate technique. Role-playing does not show the courage and conviction of authentic dissent. When a person is role-playing, you don’t really know the relationship between what he is saying and what he believes. Even when his words are consistent with his beliefs, you know that he is acting. And since you are aware that he is playing a role, you are likely to think and interact with him differently. After all, you can’t persuade someone who is playing a part to change his position.

Regardless of the reasons, the devil’s advocate, unlike authentic dissent, does not stimulate the kinds of thinking that lead to good decision-making or creative problem solving.

A Better Devil?

Some researchers are convinced by the conclusion above, but others have sought to improve the devil’s advocate technique by amplifying the debate aspect of it. Among the solutions offered by Cass Sunstein and Reid Hastie, for example, is a pumped-up version of devil’s advocate called “red teaming,” in which teams compete and try to defeat one another. Many other variants on role-playing also try to clone the properties of dissent, but what they often lack, almost by definition, is authenticity.

It is when you face a dissenter who believes his position, has the courage to say so, and does so persistently that you confront the possibility that you may be wrong.

You at least start to investigate the complexity of the information and the issue. You start to seek information and to consider alternatives, much as you do when first forming an opinion. You look at all sides. You consider the cons as well as the pros. You consider multiple possibilities. You aren’t chitchatting, making friends, or engaging in an intellectual exercise—you are thinking.

The Evolution of the Devil's Advocate

The Catholic Church doesn’t want to learn after the fact that a saint behaved in ways that were not saintly. Thus, in 1587, it instituted a practice of exploring everything negative about each candidate. God’s advocate might argue for canonization, but the devil’s advocate argued against it. He might argue, for example, that all the miracles attributed to the candidate were fraudulent. Intuitively, the technique would seem to counterbalance a bias in favor of the candidate held by most of the decision-makers, including the Pope. It is worth remembering, however, that the devil’s advocate was but one mechanism used in the process of canonization, which could take decades, if not centuries, to complete.

When the route to sainthood was streamlined, in 1983, the Church eliminated the traditional position of the devil’s advocate. Some worried that doing so would leave little room for dissenting opinions. However, in an interesting variation, the Church invited authentic dissent from the fiercest critic of the next person being considered for sainthood—Mother Teresa. An international icon, she had been widely applauded for helping the poorest of the poor in India and those dying of diseases such as AIDS and leprosy. That fierce critic was none other than Christopher Hitchens—essayist, author, and religious, literary, and social critic. Known for his confrontational style, Hitchens had criticized Mother Teresa and had even called her a fraud. He was asked to argue against her, and he did so with great flourish. One of his criticisms was that she wanted to convert people to her Roman Catholic beliefs, which Hitchens considered less than saintly. Much to its credit, the Church listened to him. In the end, it decided, his argument was irrelevant. However, I suspect that his critique stimulated a better scrutiny of the evidence and better decision-making.