How Many Heads Are Better Than One?

Since the beginning of human history, people have made decisions in groups. In modern societies, they do so in companies, law firms, school boards, labor unions, religious organizations, governments, and international institutions. In all these cases and countless others, people assemble in groups, small or large, to decide what to do. As the saying goes, two heads are better than one, and if this is so, then three heads should be better than two, and four are better still, and with a hundred or a thousand, well, things are bound to go well. (We’ll discuss some mathematical results that describe and constrain the power of group thought in chapter 8.) If group members really talk to one another, they can learn and make better choices, and the group can correct the mistakes made by some or many of its members.

This optimistic view has a long legacy. A prominent version can be traced directly to Aristotle, an early advocate of collective intelligence, who insisted that when people “all come together . . . they may surpass—collectively and as a body, although not individually—the quality of the few best . . . When there are many who contribute to the process of deliberation, each can bring his share of goodness and moral prudence; . . . some appreciate one part, some another, and all together appreciate all.”²

Aristotle was alert to the possibility that group members, deliberating together, may add up what they know and disregard mistakes and thus improve on “the quality of the few best.” The key point involves information aggregation, as different people appreciate different parts of a body of information and lead everyone to appreciate everything. We will spend a lot of time trying to figure out what Aristotle’s claim might mean and how to make it become true.

In the twentieth century, the philosopher John Rawls spoke in similar terms: “The benefits from discussion lie in the fact that even representative legislators are limited in knowledge and the ability to reason. No one of them knows everything the others know, or can make all the same inferences that they can draw in concert. Discussion is a way of combining information and enlarging the range of arguments.”³ While Rawls’s topic was democracy, his point extends to all groups, whether private or public, small or large.

But do groups actually succeed in surpassing the quality of the few best? Do they, in fact, combine information and enlarge the range of arguments? Do firms accomplish this feat? Do government officials? Unfortunately, the history of the human species suggests that all too often, groups fail to live up to their potential. On the contrary, many groups turn out to be foolish. They bet on products that are doomed to failure. They miss out on spectacular opportunities. They develop unsuccessful marketing strategies. Their investments and strategies go awry, hurting millions of people in the process.

Leaders of different stripes are sharply divided on the question of whether groups are likely to make good decisions. In business, many prominent leaders insist that you can’t make good decisions unless you involve a lot of people, who provide safeguards against blunders. In government, some famously successful leaders, like President Franklin Delano Roosevelt, have tended to believe the same thing. Other, less successful presidents have isolated themselves and listened to a small group of like-minded advisers. In many circles, the conventional wisdom is that if you want a good decision, you have to talk to a ton of people.

On the other hand, we are all familiar with leaders who have great confidence in their own judgments and who believe that group judgments and most inputs from advisers tend to be a distraction and a waste of time. (Vladimir Putin is an example.) Many tyrants and some geniuses tend to think this way. Sometimes they succeed; some leaders, including tyrants, really are extraordinary and can think things through on their own. But far more often, they fail, partly because they do not and cannot know enough.

Their self-confidence is both delusional and extreme. But we should pause before entirely dismissing them. Making good group decisions can be hard, and the difficulty sometimes makes us wonder whether it is worth the effort. For some people, a good group discussion is a rare experience. Economists suggest that we should assess the value of decisions in terms of two considerations: the costs of decisions and the costs of errors. Group judgments increase the costs of decisions, because many people are involved. And such judgments can increase, rather than decrease, the number and the size of errors (and hence the error costs), if the group ends up making decisions worse.

This point brings us to groupthink. The idea, first elaborated by Irving Janis, suggests that groups may tend toward uniformity and censorship.⁴ Janis was right: much of the time, both private and public groups blunder not in spite of group deliberation, but because of it. After deliberation, companies, labor unions, and religious organizations often make disastrous decisions. The same point holds for governments.

The term groupthink is memorable, and it has deservedly entered the popular lexicon. But many researchers have tried, with little success, to find evidence to support Janis’s claims.⁵ Janis believed that groups are especially likely to suffer from groupthink if they are cohesive, have highly directive leadership, and are insulated from experts. To support his account, he offered some arresting case studies.

But it is hazardous to draw conclusions from case studies that aren’t exactly randomized controlled trials, and in any case, other case studies (involving the Nixon White House and the launch of the Challenger space shuttle) do not support Janis’s claims. Experimental research fails consistently to link particular group characteristics, including those that Janis emphasized, to groupthink.⁶ Janis’s evocative account can be seen as akin to a work of literature rather than as a precise account of how groups go wrong or as helpful guidance for group success.

Janis also produced his theory before the rise of modern behavioral science, which has led to an independent literature—represented not only in Kahneman’s Thinking, Fast and Slow, but also in Ariely’s Predictably Irrational, Mullainathan and Shafir’s Scarcity, and Thaler and Sunstein’s Nudge.⁷ To date, however, no one has focused, in any sustained way, on how the recent behavioral findings bear on the performance of firms and other groups. We aim to fill that gap. We want to bring behavioral research, including behavioral economics, in direct contact with the question of group performance.

In this way, we seek to go beyond the idea of groupthink to obtain a more precise understanding of the problem of group failures (including business and government failures involving teams and organizations) and to identify potential solutions. We elaborate on several decades of research, some of it our own, that helps to specify the mechanisms that account for the failures of various deliberating groups, and that points the way toward successful practices. Many of the most constructive ideas show the extraordinary value of small steps that can be implemented immediately and that have a big impact. Countless businesses can do better for their employees and their customers. The same is true for governments, and for many other groups, including charities, religious organizations, and educational institutions.

Sharing What We Know

The two of us have studied group decision making for many years, and this book draws heavily on our own empirical work as well as our experience.

From 2009 to 2012, one of us (Sunstein) served as administrator of the White House Office of Information and Regulatory Affairs (OIRA). OIRA helps to oversee federal rule-making in the areas of environmental protection, workplace safety, health care, immigration, agriculture, homeland security, and much more. In the federal government, as in so many other places, group decision making is the norm, and the fate of large projects and policies may depend on group dynamics.

The point of the OIRA process is to increase the likelihood that important rules and requirements will be consistent with law, will actually benefit the American people, and will not have unjustified costs. When the process works well, it is in part because different agencies of the federal government, all with their own information, are able to share what they know and thus to produce a better outcome. For example, the Council of Economic Advisers has specialists on economic effects. The Environmental Protection Agency (EPA) has experts on the environment. The US Trade Representative knows a great deal about trade implications. The Office of Science and Technology Policy has plenty of scientific expertise. When people from different offices and departments feel free to say what they know, group decisions get a lot better and big mistakes are avoided.

If, for example, the EPA is considering a rule that would reduce air pollution from power plants, it needs information from other people within the federal government. It needs to know about economics and science, and it might need to know about trade. True, the agency has its own specialists, but if it hears from other experts in those fields, it is likely to make much better decisions. As we will emphasize, the assignment of particular roles, ensuring that important perspectives are not ignored, is a hallmark of well-functioning groups. Careful cost-benefit analysis, a paradigmatic example of System 2 thinking, is also a crucial way of overcoming behavioral biases, including group biases.

Something similar can be said in the area of foreign policy. The idea of good process may sound boring, but when nations avoid foreign policy mistakes or make canny decisions, it is often because they have “good processes” in place to ensure that leaders learn what they need to know, and that valuable information is not missing or lost. Within the White House, the national security adviser has the job of running that process. If the person who serves in this role is dictatorial or proceeds without giving careful consideration to what a lot of people think, the process is unlikely to go well.

During the Bush administration, the judgment that Iraq had weapons of mass destruction was, in part, a failure of an entire process in which information was not properly sifted and aggregated. During the Obama administration, the decision to go after Osama Bin Laden was a gamble, but it was a good one, because National Security Adviser Thomas Donilon ran a masterful process and kept the president superbly informed.

Complacency, Anxiety, and Happy Talk

Good managers are acutely aware of the benefits of information aggregation, and they know that role assignment can reduce the risk of “happy talk.” Happy talk occurs when group members say that all is going well and likely to go even better—that there is nothing to worry about. We think that happy talk is a pervasive source of group failures, and we will have a fair bit to say about why it is so harmful. In fact, happy talk is a principal target of this book.

In government, as elsewhere, role assignment may not be enough. We can divide the world of leaders into two kinds of people: the complacent and the anxious. Complacent leaders are relaxed, upbeat, and contented. They think that things are entirely under control. By contrast, anxious leaders are focused on possible disasters. They fear that things are about to go wrong and perhaps spiral out of control.

Complacent people are full of energy, excitement, enthusiasm, and optimism. They believe in their projects. Their enthusiasm is contagious. In the Obama administration, for example, most officials were not complacent, but some of them were; they thought that the Affordable Care Act could be implemented with only a few glitches, that Democrats and Republicans would work together, and that policies would work just as the administration hoped.

Complacent people tend to be immensely likable. They seem like perfect team players. They can tell a terrific story. Some of them are visionaries. They don’t rock any boats. They are buoyant and full of smiles. They have plenty of ideas, many of which are excellent. They tend to be happy, and they like happy talk. It’s tempting to hire and promote them. We won’t name any names, but every presidential administration, like many other workplaces, has plenty of complacent, easygoing people.

By contrast, the anxious people may be optimistic, nice, even enthusiastic and full of smiles, but they are also troubled by concern, skepticism, and doubt. They believe in their projects and want them to work, but anxious people see obstacles, downsides, and challenges everywhere. Working with the Obama administration’s complacent people, the anxious people would sometimes look doubtful, focused, negative, and occasionally severe. They would ask probing questions along the lines of, “What could go wrong? Did you think of this? Why haven’t you planned for that?”

We will name two names. Nancy-Ann DeParle, who served as deputy chief of staff for policy from 2010 to 2012, is an exemplary anxious person. DeParle has had a lot of government experience. She is keenly well aware that competent, smart, and enthusiastic people can make big mistakes, especially if these people are not looking around the corner and thinking about worst-case scenarios. While she is immensely fair and one of the nicest people in the world, she is not always smiling. She has no problem looking people in the eye and doubting their plans, or telling them a firm no, in part because she is anything but complacent.

Here’s a little story. When Sunstein was running OIRA, DeParle sent him a mildly anxious email late one night, asking, “How’s the regulation coming?” Sunstein had no idea which regulation she meant. He responded, “I don’t know which regulation you have in mind, but most of them are coming pretty well, so chances are that this one is coming well too.” DeParle replied briefly: “hug.”

Sunstein was a bit amazed, and also touched, because nice as DeParle is, she isn’t prone to sending hugs. Sunstein answered, “That’s the nicest email of the year.” DeParle promptly wrote back to explain that she had written “ghg,” for “greenhouse gas,” and her cell phone had autocorrected to “hug.” (Oh well.)

DeParle isn’t a hugger. She doesn’t like happy talk. She makes everyone around her, and everyone below or above her, much wiser, although sometimes also a bit uncomfortable. You can’t rerun history, but there’s an argument that if she had continued to serve in the Obama administration, the troubled rollout of HealthCare.gov in 2013 would have been executed better and the website would have worked. Whether or not that is so, anxiety concentrates the mind, and an anxious leader concentrates a lot of minds. Anxiety, like most emotions, is contagious—which is an important lesson for managers to keep in mind.

In President Obama’s first term, Jeff Zients served as deputy director for management and later as acting director of the Office of Management and Budget. He is upbeat, funny, and unfailingly genial, and his brow is seldom furrowed. Everyone loves Zients, but he’s fully capable of productive anxiety, especially when the stakes are high. In the private sector, he was known as a superstar manager. During the 2013 fiasco that was the early days of HealthCare.gov, President Obama made a really smart move in asking Zients to fix the problem—as Zients did. The president also made a smart move in asking Zients to return to the White House as director of the National Economic Council.

Good managers have to be anxious, even as they must inspire loyalty and affection in their colleagues and subordinates, and Zients has an extraordinary ability to combine anxiety, likability, and competence. Some of his colleagues have referred to him as “the quarterback,” not only because he has a terrific throwing arm, but also because his leadership ability is manifested in insights and performance, not happy talk.

Anxious people like DeParle and Zients are indispensable in business and government, because they cut through, and overcome, the risk of groupthink. They work a bit like devil’s advocates (discussed in chapter 6), but they’re a lot better, because everyone knows that they mean business and aren’t playing any kind of game. They say what they think. Even better, they say what they fear. With their very different styles, they keep asking, “What can go wrong?”

When groups ultimately do well, it is often because they have anxious leaders and hence are better able to obtain and aggregate dispersed information—and to ensure that the groups know everything that their members know. Groups are also capable of learning, because anxiety is a great motivator. Groups need a little anxiety, maybe even a lot of it. They need a culture that enables them to find out what they need to know.

Realism, Optimism, and the Path to Wiser Groups

The good news is that if discussion is properly structured and if groups adopt the right norms and practices, they can create that culture. The bad news is that in the real world, discussion often leads people in the wrong directions. Many groups fail to correct the mistakes of their members. On the contrary, groups often amplify those mistakes. If group members are unrealistically optimistic, groups may be more unrealistic still. If people within a firm are paying too little attention to the long term, the firm will probably suffer from a horrible case of myopia. There is no evidence that the judgment mistakes uncovered by behavioral scientists are corrected as a result of group discussion.

Many of the most serious blunders made by companies, public officials, and religious organizations are a direct result of the problems that we will trace here. When companies launch products that no one wants to buy, it is often because the doubters have tried to be team players and silenced themselves. When officials propose environmental rules that are too weak, too draconian, or simply misdirected, it is often because group discussion has failed.

As we will see, wise groups and their managers are fully alert to the risks of suboptimal group processes and take concrete steps to counteract them. With the rise of the internet, there are unprecedented opportunities to take such steps. Groups can pool the information that their members have. They can also reach outside themselves to draw on the dispersed knowledge of hundreds, thousands, or even millions of other people.

Drawing on empirical research, we identify a number of concrete solutions for making groups wiser. Some of those solutions are straightforward; for managers, the steps require leadership, determination, understanding, and a little creativity. Other solutions call for the use of modern technologies and may require some investment in both time and resources. Some of the solutions are informal and can be instituted overnight; others require at least a degree of formality and call for advance planning. Different approaches will work for different groups.

One of this book’s recurring themes is that one size does not fit all. But if groups and their leaders have a fuller understanding of how things go wrong, they will be in a much better position to make things go right.

The Plan

This book is divided into two parts. Part 1 explores the sources of group failure. Chapter 1 establishes the context, describing why we deliberate and discuss problems in groups, what our hopes are when we do so, and the challenges that groups face. The principal focus is on how groups may fail to obtain important information—and on how their leaders and members tell dissenters, or people who have a different perspective, to shut up.

Chapters 2 through 5 explore four causes of group failure. First, groups do not merely fail to correct the errors of their members; they can amplify these errors. Second, groups fall into herds, as group members follow the statements and actions of those who speak or act first, even if those statements and actions lead the group in unfortunate, terrible, or tragic directions. Third, groups tend to get more extreme—as, for example, when a group of people inclined to suffer from excessive optimism becomes still more optimistic as a result of internal discussions. Fourth, members emphasize shared information at the expense of unshared information and thus do not give the group the benefit of critical and perhaps disturbing information that one or a few people have.

Part 2 turns to sources of group success. Chapter 6 begins with an account of seven methods, most of them quite simple, that can be used to reduce the risk of failure. Each method is designed to counteract the four failures that we identified in part 1. Our focus is on group deliberation and how to make it better. We emphasize the importance of self-silencing leaders and the value of role assignment, implicit or explicit.

Chapter 7 details the immense importance of distinguishing between two very different processes: (1) the identification of a list of potential solutions and (2) the selection of the preferred solution. In identifying a list of potential solutions, groups must be flexible, imaginative, and even a bit sloppy to open things up. In selecting an actual solution, it is important for groups to be tight and analytic—to close things down. Many groups do poorly because they mix up the two kinds of inquiries and, in particular, because they close things down at the very stage where they should be opening up to produce a list of potential solutions. We draw out this distinction in detail in chapter 7, but it is helpful to keep it in mind throughout the book.

In chapter 8, we explore the purported wisdom of crowds. The idea amounts to a celebration of the occasionally amazing performance of statistical groups, that is, groups whose members do not deliberate with one another at all but who answer a question or solve a problem independently. We show that crowd wisdom comes from some simple arithmetic—and that the same arithmetic explains why crowds can also be unwise. In some circumstances, statistical groups do poorly, and we spell out what those circumstances are. For groups, and especially firms, it is a mistake to assume that crowds are always wise. It is much better to understand why and when crowds turn out to be wise, and to build on that understanding to improve group deliberation. A central point here involves the importance of diversity and dissent.

In chapter 9, we discuss the role of experts. We show that it is often a big mistake to chase the expert. If you want to get the right answer, you’ll do better to consult a bunch of experts, rather than looking for the one who seems best.

In chapters 10, 11, and 12, we explore new and more creative methods, often enlisting the internet to obtain ideas. For example, both the private and the public sectors are making innovative use of tournaments, by which large populations can be asked to submit their ideas. Even if the prize is relatively small, this approach can produce pathbreaking innovations. chapter 10 shows how.

In chapter 11, we turn to prediction markets, which invite people to put their money on the line, saying, for example, whether a product or a movie will do well, whether the weather will get lousy, or whether a leader will lose his or her job. As we will see, prediction markets have proved stunningly accurate. chapter 12 describes how groups can learn a lot from seeking public comment.

In chapter 13, we encounter the mysterious Factor C, a collective-intelligence factor that suggests that certain personality characteristics—above all, social sensitivity—can make groups perform better. Wise groups rely on a combination of incentives and architecture, but they also benefit from certain personalities, which they can also cultivate and help to produce. Some personal characteristics are innate, but others can be learned. Basketball’s all-time best player, Michael Jordan (sorry, LeBron), was a great learner. With respect to Factor C, he got a lot better. Groups are unlikely to have anyone like Michael Jordan, but they do far better when they promote learning of multiple kinds.