by Matthew Syed
Of course, if the individuals in the group don’t know much then combining their judgements won’t achieve much. If you ask a group of laypeople to estimate how much ocean levels will rise over the next decade you won’t get very far. To achieve group wisdom, you need wise individuals. But you also need diverse individuals, otherwise they will share the same blind spots.
Now, with this in mind, let us perform a thought experiment. Suppose you identified the fastest runner in the world. Let us call him Usain Bolt. Suppose, too, you could clone this runner. If you were putting together a relay team of, say, six runners, your team of Usain Bolts would smash the opposition (assuming they pass the baton effectively). Every single one would be faster than anybody in any other team.
This tells us something that was alluded to near the beginning of this book. When it comes to simple tasks, diversity is a distraction. You just want to hire people who are smart, fast, knowledgeable, whatever. Things are not just different, but the polar opposite, when it comes to complex problems, however. Let us return to economic forecasting. Suppose you could identify and clone the most accurate forecaster in the world. If you were putting together a team of six forecasters, would it make sense to put six of these clones together? On the surface, the team sounds unbeatable. Each member is more accurate than any forecaster in any other team. Isn’t this the perfect team?
We can now see that the answer is an emphatic ‘no’! They all think in the same way. They use the same model, and make the same mistakes. Their frames of reference overlap. Indeed, the Soll experiment implies that a diverse group of six forecasters, while individually less impressive, would be 15 per cent more accurate.
It is worth pausing to reflect upon how potentially world-changing this result is. It reveals – in precise, mathematical terms – the sheer power of cognitive diversity. A team of world-class forecasters who think in the same way are less dramatically intelligent than a group of forecasters who think differently.
Of course, most of us do not sit around the table at work, or in life, making numerical forecasts of the kind familiar to economists. But we do try to solve problems, come up with creative ideas, determine strategies, spot opportunities and much else besides. This is the essence of the group-based work that is coming to dominate our world. And yet we can expect diversity to have even stronger effects on these tasks.
Let us take creativity and innovation. Ask yourself this question: suppose that you put together a team of ten people to come up with ideas to solve, say, the obesity crisis. Suppose, too, that each of these ten people comes up with ten useful ideas. How many useful ideas do you have in total?
NOW CLOSE THE BOOK TO FIGURE OUT YOUR ANSWER.
In fact, this is a trick question. You can’t infer the number of ideas in a group from the number of ideas of its members. If these people are clone-like and come up with the same ten ideas, you have only ten ideas overall. But if the ten people are diverse, and come up with different ideas, you could have one hundred useful ideas. That is not 50 per cent more ideas, or 100 per cent more ideas, but almost 1,000 per cent more ideas. This is another huge effect solely attributable to diversity.
In problem-solving teams, we see the same pattern. We noted that in prediction tasks, taking the average of independent forecasts is an effective way of pooling information. With problem solving, however, averaging is often a terrible idea. Taking the average of two proposed solutions can often lead to incoherence. This is where the phrase ‘a camel is a horse designed by committee’ comes from. With most problems, a team has to reject some ‘solutions’ in favour of others.
But this again reveals why diversity matters. With homogenous groups people tend to get stuck in the same place. Diverse teams, on the other hand, come up with fresh insights, helping them to become unstuck. Rebel ideas are effectively firing the collective imagination. As the leading psychologist Charlan Nemeth puts it: ‘Minority viewpoints are important, not because they tend to prevail but because they stimulate divergent attention and thought. As a result, even when they are wrong they contribute to the detection of novel solutions that, on balance, are qualitatively better.’6
But the power of diversity is more subtle than even these examples might suggest. The deepest problem of homogeneity is not the data that clone-like teams fail to understand, the answers they get wrong, the opportunities they don’t fully exploit. No, it is the questions they are not even asking, the data they haven’t thought to look for, the opportunities they haven’t realised are out there.
The more challenging the domain, the less that any single person – or perspective – can hope to grasp. With prediction teams, homogenous minds make the same errors. With problem-solving teams, they get stuck in the same place. With strategy teams, they miss the same opportunities.
When Justice Scalia argued that there was a trade-off between performance and diversity, he was making a seductive conceptual error. It is the same error that leads most people to express surprise when told that the average of six forecasters is more accurate than the top forecaster, and that deludes people into thinking that a group of wise individuals must constitute a wise group. Scalia was, in effect, looking at the problem from the individualistic perspective, not the holistic perspective. He didn’t take account of the fact that collective intelligence emerges not just from the knowledge of individuals, but also from the differences between them. Let us call this the ‘clone fallacy’.
The tragedy is that this fallacy is pervasive. Indeed, perhaps the most striking conversation I had while researching this book was with a renowned economic forecaster. I asked if he preferred to work with people who think in the same way, or who think differently. He replied: ‘If I truly think my model is the best one out there, then I should work with people who think like me.’ This logic is highly compelling. It is also spectacularly wrong.
V
Most organisations have an avowed policy of meritocratic hiring. The idea is to recruit on the basis of skill and potential, rather than on arbitrary factors like social connections, race or gender. This is both morally commendable and self-interested. Institutions are hiring talent regardless of what it looks like. But it also contains latent dangers. Let us take a hypothetical example to flesh out the logic. Suppose that some universities have a strong reputation for, say, software development. These universities are likely to attract the smartest software students. These students, in turn, will graduate with the most impressive credentials. Now, suppose you are running a top software company. Wouldn’t you want these students? Wouldn’t you want to pack your organisation with the brightest and the best?
The enlightened answer is ‘no’. These graduates will have studied under the same professors and absorbed similar insights, ideas, heuristics and models, and perhaps world views, too. This is sometimes called ‘knowledge clustering’. By selecting graduates in a meritocratic way, organisations can find themselves gravitating towards clone-like teams. This is not to dismiss meritocracy. It is merely to point out that collective intelligence requires both ability and diversity.
Indeed, no test that ranks individuals can – on its own – construct intelligent groups, another point made by Scott Page: ‘Suppose you are building a team to come up with creative ideas. First, any test applied to an individual can only measure that individual’s ideas. Second, a clone of the person who scores highest on whatever test we apply necessarily adds less to the group than a second person with a single different idea. Therefore, no test can exist.’7
Now let us return to the distinction between cognitive diversity (differences in thoughts, insights, perspectives) and demographic diversity (differences in race, gender, class and so on). We noted in Chapter 1 that demographic diversity often overlaps with cognitive diversity. This is intuitive since our identities influence our experiences, perspectives and more. Advertising firms, for example, rely on demographic diversity to create campaigns that appeal to the breadth of their client base.
This helps to explain
the study by Professor Chad Sparber (along with dozens of others), which found that an increase in racial diversity of just one standard deviation increased productivity by more than 25 per cent in legal services, health services and finance. In any domain that requires an understanding of broad groups of people, demographic diversity is likely to prove vital.
But there are other contexts where the overlap is less significant, or even non-existent. In the very same piece of research, Sparber found that increases in racial diversity offered no efficiency gains for firms producing aircraft parts, machinery and the like. Why? Because the experience of being, say, black provides few, if any, novel insights into the design of, say, engine parts.
We can make this point in a different way with economic forecasting. Take two economists: one white, gay, male and middle-aged, the other black, young, female, heterosexual. These economists are different in demographic terms – and might tick all the boxes on a conventional diversity matrix. But suppose they went to the same university, studied under the same professor and left with similar models. In these circumstances, they would be clone-like in relation to the problem.
Now take two white, middle-aged, bespectacled economists, who have the same number of children and like the same TV programmes. They may seem homogenous and, from a demographic perspective, they are. But suppose that one is a monetarist and the other a Keynesian. These are two different ways of making sense of the economy; two very different models. Their collective prediction will, over time, be significantly better than either alone. The two economists may look the same, but they are diverse in the way they think about the problem.
This is worth keeping in mind because hiring someone who is different in terms of colour or gender does not guarantee an increase in cognitive diversity. Building collective intelligence cannot be reduced to a box-ticking exercise. Consider, too, that people who start out diverse can gravitate towards the dominant assumptions of the group. This can lead to a situation where leadership teams look diverse, but who are – in cognitive terms – anything but. They have all been at the organisation so long that they have come to share identikit views, insights and patterns of thinking.
Successful teams are diverse, but not arbitrarily diverse. A group of scientists designing a hadron collider is unlikely to benefit from hiring, say, a skateboarder, of whatever colour or gender. Or consider what would have happened if the FA board of Brailsford, Campbell, Giles, Lancaster, Badale had been invited to advise not on English football, but on, say, DNA sequencing. The team would have had diverse information, but it would scarcely have impinged on the problem space.
Diversity contributes to collective intelligence, then, but only when it is relevant. The key is to find people with perspectives that are both germane and synergistic.
For economic forecasters, collective intelligence emerges from accurate predictors, who deploy different models. For an intelligence agency, it emerges from outstanding analysts, who possess a rich diversity of experience, the better to understand the multiplicity of threats they face. For policymakers, it emerges from exceptional individual politicians with (among other things) backgrounds that span the demographic spectrum of the electorate they serve. For teams working in other contexts – well, we will see further examples as the book progresses.
Perhaps the most important point is the generalised significance of diversity. Diversity isn’t some optional add-on. It isn’t the icing on the cake. Rather, it is the basic ingredient of collective intelligence. You can see the power of diversity from a broader perspective, too. Diversity explains why price systems work so effectively, and why open-source innovation platforms and wikis have become pervasive. These all share the same underlying signature: they aggregate the disparate information contained in different minds.FN4
Diversity has even moved to the heart of artificial intelligence. A couple of decades ago, machine learning was based on single algorithms. Today, it is largely characterised by ensembles of diverse predictors. Scott Page hit upon the same pattern when creating problem-solving computer models. ‘I stumbled on a counterintuitive finding,’ he said. ‘Diverse groups of problem-solvers . . . consistently outperformed groups of the best and brightest.’8
VI
One of the much-vaunted solutions to a lack of diversity in politics and beyond is the use of focus groups. These are often hailed as a means of offering the benefits of diversity without having to dilute the clubbable homogeneity of power structures within political elites. The basic idea is that you put a representative cohort of people in a room, ask questions, find out what they like and what they don’t, note any objections and practical problems, and then refine the policy accordingly. Advertisers sometimes make the same point with ‘market research’, testing the ideas on a diverse audience to gain insights into what works and what doesn’t.
But it should be clear that such approaches, while sensible in their own terms, miss the deeper point. Why? Because diversity is not just about getting answers from focus groups or market research. It is about the questions that are asked in the first place, the data that is used as the basis for deliberation and the assumptions that permeate the problematisation of any issue.
This is not just true of politics, but science – supposedly the most objective of disciplines. One survey of sports science journals found that 27 per cent of studies focused exclusively on men, but only 4 per cent on women.9 It is no coincidence that the vast majority of sports scientists are men. This is one tiny example of how biases can be baked into deliberations before scientists start to answer questions, and where data is skewed before the lessons are probed. This shows that while demographic diversity and cognitive diversity are conceptually distinct, they typically overlap.
You can see the same point in a different way by looking at primatology. Before Jane Goodall came on the scene, the field was dominated by men. They adopted Charles Darwin’s view of evolution, focusing on competition among males for access to females. In this framework, female primates are passive, and the alpha male has access to all the females, or females simply choose the most powerful male. But this frame of reference contained a blind spot. Only when a critical mass of women arrived on the scene did primatology come to realise that female primates are far more active, and might even have sex with many males, insights that created a richer, more explanatory theory of primate behaviour.
Why did women scientists see something that men had missed? In her fascinating book The Woman That Never Evolved, the anthropologist Sarah Blaffer Hrdy writes: ‘When, say, a female lemur or bonobo dominated a male, or a female langur left her group to solicit strange males, a woman fieldworker might be more likely to follow, watch, and wonder than to dismiss such behaviour as a fluke.’
We saw in the opening chapter that Japanese people tend to focus more on context and less on individuals when compared with Americans. It is noteworthy that primatology has benefited from this very effect. As the academics Douglas Medin, Carol D. Lee and Megan Bang put it in a lead article for Scientific American:
In the 1930s and 1940s U.S. primatologists . . . tended to focus on male dominance and the associated mating access. Rarely were individuals or groups tracked for many years. Japanese researchers, in contrast, gave much more attention to status and social relationships, values that hold a higher relative importance in Japanese society. This difference in orientation led to striking differences in insight. Japanese primatologists discovered that male rank was only one factor determining social relationships and group composition. They found that females had a rank order, too, and that the core of the group was made up of lineages of related females, not males.
This takes us back to something mentioned earlier. Remember the warning of John Cleese? ‘Everybody has theories,’ he said. ‘The dangerous people are those who are not aware of their own theories. That is, the theories on which they operate are largely unconscious.’ We can now see that this applies as much to science itself as to anything else. In his subtle and bea
utiful book Conjectures and Refutations, Karl Popper, perhaps the greatest of all philosophers of science, makes this point, along with so many others. His words are among my favourite ever written, and a useful jolt not just to scientists, but all of us:
Twenty-five years ago I tried to bring home the point to a group of physics students in Vienna by beginning a lecture with the following instructions: ‘Take pencil and paper; carefully observe, and write down what you have observed!’ They asked, of course, what I wanted them to observe. Clearly the instruction, ‘Observe!’ is absurd . . . Observation is always selective. It needs a chosen object, a definite task, an interest, a point of view, a problem . . . For a scientist [a point of view] is provided by his theoretical interests, the special problem under investigation, his conjectures and anticipations, and the theories which he accepts as a kind of background: his frame of reference, his ‘horizon of expectation’ [my italics].
VII
Take a look at the crossword below. There are thirty-five clues in total, eighteen across and seventeen down. Some of the clues are general knowledge, some are riddles and others are anagrams. If you want to have a go at it, the answers are at the back of the book. This particular crossword was published in the Daily Telegraph on 13 January 1942. At the time, readers of the newspaper had been complaining that the daily crossword puzzle was becoming too easy. Indeed, some claimed they could complete it in a matter of minutes. This was met with disbelief in some quarters, prompting a man called W. A. J. Garvin, chairman of the Eccentric Club, to offer a £100 prize to be paid to charity if anyone could complete the puzzle in less than twelve minutes.