Diversity Can Trump Ability

From "The Trouble With Elections: Everything We Thought We Knew About Democracy is Wrong," Chapter 11.5

The notion that diversity can trump expertise contradicts the “common sense” assumption that groups of experts should outperform panels of non-experts. The “Delphi method,” named for the ancient Greek oracle and advanced by the RAND corporation at the beginning of the Cold War, was based on this assumption. Experts were polled about a particular issue, such as the effects of technological changes on future wars. Initial results were shared with the experts, who then had the chance to modify their previous responses, in a series of iterative rounds.  

The Delphi method had mixed results, sometimes accurate and sometimes far off the mark. By keeping the experts anonymous and isolated from each other, it was hoped they could avoid the groupthink dynamic that promotes agreement, regardless of whether the point of convergence is right or wrong. But confirmation bias among experts is still a problem. Also, the method is intended to predict what is likely to happen, not what we want to happen. Policy decisions are grounded in community values, which the method isn’t intended to tackle. So, at best, it can only advise and inform decision-makers. The Delphi method also suffers from initial assumptions of who qualifies as an “expert.” Relatively homogeneous experts often have gaping holes in their knowledge of matters outside their area of expertise that nevertheless are of key importance. James Surowiecki explains this counterintuitive reality:

“Bringing new members into the organization, even if they are less experienced and less capable, actually makes the group smarter simply because what little the new members do know is not redundant with what everyone else knows.”  

Professors of business and finance, Lu Hong and Scott E. Page published a paper in 2004 offering a theorem that shows the inherent weakness of the approach to problem solving that seeks to find a group of “best” problem-solvers, who maximize individual abilities, and who inevitably also are less diverse than a group of randomly selected problem-solvers.

“Ultimately, the gain in individual abilities is more than offset by the functional [cognitive] diversity of a group of randomly selected people. It is in this sense that we might say diversity trumps ability.” 

In Page’s subsequent book, The Difference: How the Power of Diversity Creates Better Groups, Firms, Schools, and Societies, he describes how he discovered this unexpected dynamic while running a computer model about problem solving. He found that a group of random agents out-performed a group made up of the best individual performers in a problem solving task. This didn’t make intuitive sense to him and he at first assumed he had made some coding error in writing the computer program. But there was no error. The result kept repeating in scenario after scenario. With deeper examination, the benefit of diversity in solving difficult problems became clear. He bluntly named this the “Diversity Trumps Ability Theorem.” 

There are, of course, limits to the benefits of diversity. To consistently out-perform the group of “experts,” the groups in Page’s computer model couldn’t be too small, and the members of the diverse group also had to be reasonably smart.

“These conditions all made intuitive sense. A collection of third graders probably has little to add to the study of global warming.”

Although the attempt to create a mathematical proof, which relies on specific conditions and assumptions, may be erroneous and it might be an over-reach to apply this theorem to the real world – and indeed mathematicians have criticized the paper – there is real world experience that supports their initially counter-intuitive conclusions.

For this theorem to be applicable in the real world there are many conditions that must be met. Two of the most challenging are the ability of the participants to effectively communicate with each other, and to properly evaluate what they learn. This is where an understanding of group social dynamics and good process design come in. The problem being tackled by the group can be difficult, but can’t be so arcane that only specially trained experts can understand or evaluate potential solutions.  As I stated earlier, and bears repeating, “experts” are those who have experience solving old problems, and this experience may be helpful or act like blinders in dealing with new problems. Since future political challenges are inherently unpredictable, we can never know in advance which sort of problem-solving styles, or experts, might be most beneficial to have in a deliberative body. Diversity allows us to cover all the bases. 

Even when diversity does not introduce new and more varied information to a group, research has shown that merely by reducing homogeneity, better decision-making is possible. Research by social psychologists such as Katherine Phillips of Northwestern University and others indicates that it is not simply a matter of diverse members (outsiders, if you will) bringing new information and perspectives to the deliberation. They have found that

“even when out-group members do not bring minority viewpoints to the table, the mere presence of social category diversity in task groups can fundamentally change the behavior of the social majority and enhance group performance.”

A team of researchers conducted studies with experimental markets in Singapore, and replicated it in the United States, in which participants could earn money through virtual trading of stocks. Homogeneous groups were far more likely to overvalue stocks and develop market bubbles. In a paper published in November 2014 in the Proceedings of the National Academy of Sciences, they wrote

"In our experiments, ethnic diversity leads all traders, whether of majority or minority ethnicity, to price more accurately and thwart bubbles."

Rather than minority members bringing new information to the fore, in these cases diversity promoted better individual decision making in general. One of the researchers, MIT professor Evan Apfelbaum, noted that

“it is clear that people work harder in diverse settings, socially and cognitively. So when it comes to creating the conditions that promote cold hard accuracy in decision-making, diversity may be an invaluable asset."

The paper notes that

“the evidence may inform public discussion on ethnic diversity: it may be beneficial not only for providing variety in perspectives and skills, but also because diversity facilitates friction that enhances deliberation and upends conformity.”

Cognitive biases are not merely an attribute of individual humans. There are also much less studied collective cognitive biases. In the paper on diversity vs. homogeneity and stock trading, the authors wrote:

“Economists and psychologists have cataloged numerous individual cognitive biases. However, we suggest that biases may stem not only from the limits of individual cognition, but also from the social context in which decisions are embedded. Homogeneity (or diversity) is not a feature of individuals, but of a collective: a team, a community, or a market.”  

The researchers in that study also note that

“the presence of more than one ethnicity fosters greater scrutiny and more deliberate thinking, which can lead to better outcomes. Such friction, however, can cause conflict and complicate collective decisions. The challenge, then, is in establishing rules and institutions to address ethnic diversity and its effects. Without them, conflict can be destructive; with them, diversity can benefit the collective.”

Whether these social cognitive biases and behaviors are innate or culturally created, they are pervasive, and need to be anticipated in designing optimal deliberative procedures.

Mere diversity in the room is not enough. One design option that is being experimented with is the use of segmented deliberation spaces. It has been found that when a few members of a marginalized minority are included, they may not be inclined to fully express concerns about a policy topic. Their participation often changes when there is a critical mass of them in the room. Among many cultures, if even fifty percent of those present are women, they might contribute less than ten percent of the speaking. In a group of all women it turns out they have plenty to say, but cultural norms of deferral may trump this. Another approach involves facilitating separate caucusing by identity groups, which may bring forth perspectives and information that the whole group would not otherwise benefit from. But there are also polarization risks from this procedure, and more research is needed. It is beyond the scope of this book to examine the array of facilitation methods and group deliberation designs, but it is important to note that diversity alone is insufficient.