Chapter 1 discusses how groups can arrive at more accurate conclusions than individual experts when they are tackling cognition problems—problems with finite solutions.

The chapter opens with a story about the game show Who Wants to Be a Millionaire? In this game show, contestants must answer a series of increasingly difficult multiple-choice questions. When they do not know the answer, they have the option to remove half the incorrect answers, call someone and ask for their opinion, or invite the audience to vote. Statistically, calling an expert yields the right answer 65% of the time, whereas the majority vote from the audience is right 91% of the time. This demonstrates that crowds can be wise and are often wiser than a single individual.

Rigorous studies done by sociologists and psychologists have also found that a group’s aggregate answer is more accurate than most individual answers for questions with definitive answers. For example, crowds are better at guessing a room’s exact temperature or estimating the number of jellybeans in a big jar.

It is important to note that aggregate answers from the group are not always the most accurate. In guessing the number of jellybeans in a jar, the answer of an individual or a small number of individuals could possibly be more accurate than the group aggregate. However, the wisdom of crowds comes from its accuracy: The individuals who guess correctly one time are not likely to maintain the same accuracy if asked to repeat the task.

In order for the aggregate answer of the crowd to be consistently right, three crucial conditions must be met: diversity of opinion, independence, and decentralization. Diversity means that members of the group must have distinct information or backgrounds that distinguish their opinions or their approaches from others. Independence means that individual members of the group have to guess without consulting others. Finally, decentralization means that the individuals must be allowed to tackle the problem and resolve it using their own methods rather than led down a specific path.

If these conditions are met, the wisdom of crowds can be likened to a mathematical truism: If the correct answer is the average, then the larger, more independent, and more diverse the group is, the more that wrong answers cancel each other out on both extremities.

To illustrate the above points, Surowiecki uses the example of decision markets, a system where participants place bets to predict a future event. Decision markets are better predictors than experts in a variety of fields, from politics (predicting election results) to sports (predicting the winning team) to the film industry (predicting the Oscars) to economics (predicting fluctuations in the stock market). This betting system captures the collective wisdom without significant money incentive necessary, as people find that the status and reputation that being correct gives them is incentive itself.

Decision markets work because they follow the fundamental rules of diversity, independence, and decentralization. One of the earliest and most famous implementations of the decision market is the Iowa Electronic Markets (IEM) project founded in 1988, which aimed to predict election outcomes at the presidential, congressional, and gubernatorial levels in the US and abroad. To participate, people could buy or sell “contracts” based on which candidate they thought would win, and the percentage of votes that the winning candidate received at the end would be paid out to participants who voted for them. Between 1988 and 2000, the IEM was off only 1.37% on presidential elections, 3.43% on other US elections, and 2.12% on foreign elections. They performed better than the largest national polls, even though regular traders at the IEM numbered a few thousand.

Despite decision markets being accurate wherever they are implemented, they are often ignored by corporate America because collective intelligence is hard to explain and believe.

This chapter explores how crowds can be wise in solving coordination problems—problems that do not have a fixed answer but require people coming together to arrive at an appropriate solution for the benefit of all.

Bees exemplify how groups solve problems of coordination: To ensure the survival of the hive, foragers go out in multiple directions to find sources of nectar. Once they find something, they come back, do a waddle dance, and convince other bees to follow them to the area. If the source is plentiful, the bees’ dance gains momentum, and more of their fellows join them. In contrast, the waddle dances that lead to smaller flower patches fail over time to convince additional bees to go, leading foragers to abandon their case altogether. In the end, this helps distribute the total bee workforce, proportionately, across various fields of flowers based on how much nectar can be foraged.

The bees arrive at a collectively intelligent solution not because they rationally consider all possibilities and eliminate those they deem bad; rather, they first uncover possible solutions and then decide which ones are best among them. This can only be possible with diversity: The more directions that forager bees can explore, the more flower fields they can potentially uncover. Without diversity, they cannot know better alternatives.

Thus, collective intelligence requires two skills: generating alternatives and distinguishing good solutions from the bad, or making sure to kill off the bad possibilities quickly to focus efforts on the positive possibilities. Diversity is required in both steps because smart people tend to reason in similar ways, and this often prevents growth or the exploration of alternatives. Diversity is especially important for small groups and in formal organizations because they allow people to explore a larger range of possibilities, without falling into the trap of evaluating only a narrow range of options. This is why the larger the crowd, the more accurate it tends to be, as there is almost certainly a level of diversity present with a greater number of people.

On the other hand, expertise can often be wrong because it is narrow. It is hard for experts to keep learning when their circle becomes too similar: They become less able to investigate alternatives. This is illustrated in the example of the chess expert: Although they can mentally recreate an entire game based on the remaining chess pieces on the board, they cannot remember the layout of said pieces if they are placed haphazardly, as it ceases to make logical sense to them. They are good at one thing, at the expense of all others.

Most importantly, there is no example that a single person can be an expert in decision-making or predicting the future consistently. Chasing the expert is not a worthwhile endeavor, as statistics are heavily stacked against them. They often not only make wrong predictions but also overestimate their accuracy entirely—they do not understand how wrong they are.

However, this does not mean that expert opinion is entirely useless. Smaller groups made of experts and non-experts often can harness collective intelligence. Nevertheless, their accuracy is statistically worse than larger, more diverse crowds. Surowiecki laments that, despite the proven accuracy of collective intelligence, people still refuse to trust crowds because they think averaging means compromising or dumbing down. To this day, most people are convinced that intelligence resides in individuals.

Surowiecki concludes that cognitive diversity is fundamental to making sound choices. Small groups that are not diverse fall prey to groupthink, a phenomenon that encourages them to all reinforce each other’s opinions, either due to social pressure of conforming or to the ridiculing of dissenting voices, which insulates outside opinions and prevents the group from considering all factors. For example, the Bay of Pigs invasion was a failure because the Kennedy administration failed to consult outside opinion, squashed opposing voices by making them sound improbable, and convinced themselves of their success. Diversity not only makes alternatives visible but also encourages individuals to say what they really think when under the pressure to conform.

This chapter seeks to answer the following question: Can people arrive at truly wise and independent conclusions despite being social creatures that evolved from imitating others in their groups?

Surowiecki uses the analogy of the circular mill to illustrate this problem. The circular mill is a real phenomenon in biology where social animals, such as ants, follow each other mindlessly in a looping circle until they run out of energy and die. The circle is formed because the ants cannot act or think independently. They only know to follow fellow ants, which prevents them from breaking the cycle once it is formed.

Under normal circumstances, ants are relatively similar to humans, as they are social creatures that function in a colony. Generally, they manage to forage food and reproduce without dying in a circular mill. However, they lack independence—the key to breaking from vicious cycles—which humans are able to harness.

Surowiecki thus argues that crowds can and should avoid interdependence if they want to remain intelligent. Independent decision-making is crucial, and it is defined as a relative freedom from outside influence. It helps on two fronts: First, it prevents mistakes from becoming correlated (meaning that if two people make the same mistake, it is not because they mutually led each other in the same wrong direction, like the ants in the circular mill). Second, it increases the chances of individuals presenting new information.

Nevertheless, independence is difficult to establish, as humans are social beings. Even though the core of Western liberalism and the foundational tenets of textbook economics stipulate that people are independent thinkers, this is not always true in practice, as people are often influenced by the actions of those around them. This is normal from an evolutionary standpoint, as humans have always grown by learning from the group. As such, people tend to influence each other’s judgments.

Surowiecki cautions, however, that the more people influence each other’s decision-making, the less likely it is that their decision will be wise. This is the case for the three phenomena discussed in this chapter: social proof, herding, and information cascades.

Social proof is the belief that if a group of people perform an action, they must have a good reason to do so. An experiment conducted by social psychologists Stanley Milgram, Leonard Bickman, and Lawrence Berkowitz found out that by planting a person on the street and making them look at the sky, they could influence passersby to also look up. The more people they planted, the more passersby would be inclined to imitate them. This is a reflexive behavior because when there is uncertainty, people tend to follow the crowd. Although this behavior makes evolutionary sense, it can impede crowd judgment.

The second phenomenon, herding, is defined as the mindless following of others to minimize the risk of failing big, even if the group still suffers smaller losses by not branching out. In prey animals such as cows, herding allows the group to come together to deter predators, even if one might still be picked off. Although the faster cows have good chances of surviving if they choose to flee, they stay with the herd even if there is a small chance that they become the unfortunate one the predators attack.

In humans, herding can most easily manifest in business, especially when it is necessary to take risks. For example, National Football League (NFL) coaches can often fall prey to herding: Although they know statistically that playing conservatively on fourth downs instead of attempting a risky first down will help them win in the long run, they refuse to do so because it is risky in the moment and breaks the convention of punting or kicking for a field goal instead. Since teams have traditionally always chosen the conservative play, NFL coaches would rather follow the herd and fail small than take the risk and potentially fail big, even if they have a statistically proven higher possibility of success by branching out. Despite being experts in the field, NFL coaches often consistently pick wrong due to herding forcing their hand into the less optimal option.

The third phenomenon, information cascades, happens when crowds source their information from each other, and it can lead to bad decision-making if the origin of that information was rooted in a false belief. Since cascades happen in a sequence where one person informs another who informs another, they gradually deter people from paying attention to their independent knowledge, as imitating others is the easier and safer option—at least in appearance. For example, an experiment conducted by economists Angela Hung and Charles Plott involved using two urns, where urn A contained twice as many light marbles as dark ones, while urn B contained twice as many dark marbles as light ones. The scientists then chose one urn at random and asked all participants to blindly draw a marble from it. Then, one by one, they were asked which urn they thought they had drawn from. They would be individually compensated if they were correct.

Every participant but the first had two sources of information. First was private information, which came from the marble they drew: If it was light, it was more likely drawn from urn A. Then, they also had public information: They could hear the guesses of the participants before them. If the participant drew a light marble but the two people before them guessed urn B, that might encourage them to change their mind to follow the group. Indeed, in 78% of the cases, information cascades started.

Then, the experiment was slightly changed. Instead of individually being compensated for being right, everyone would be paid only if the group was, on average, correct. Now, participants had more incentive to rely on their private information: Even if their individual judgments were imperfect, they were more likely to be collectively right. The experiment concluded that making potentially incorrect guesses actually allowed the group to perform better overall. In other words, by removing information cascades—which encourage conformity at the expense of independent thinking—the experimenters proved autonomy to be essential to collective intelligence. People can be autonomous or encouraged to think independently, but this is difficult to enforce in practice.