The Psychology of Crowds: Why Groups Make Worse Decisions Than Individuals

On April 17, 1961, fourteen of the most experienced foreign policy and intelligence professionals in the United States government gathered in the Cabinet Room of the White House to finalize the plan for what would become the Bay of Pigs invasion. The men in that room included a former general who had commanded the Allied invasion of North Africa, a future Secretary of Defense, a former president of the Ford Foundation, and the director of the Central Intelligence Agency. Their collective résumé was staggering. Their collective judgment was catastrophic. Within seventy-two hours of the invasion’s launch, the operation had collapsed entirely, leaving fifteen hundred CIA-trained Cuban exiles captured or dead, and the United States government humiliated before the entire world.

The Bay of Pigs disaster has been analyzed more thoroughly than almost any other foreign policy failure in American history. What the analysis consistently reveals is not that the individual members of that group were incompetent. It is that the group was incompetent in ways that none of its individual members were. The process of collective deliberation had not pooled their wisdom. It had suppressed it. The social dynamics of the group had made every person in that room measurably worse at their job than they would have been thinking alone.

Irving Janis, the psychologist who studied this failure most carefully, coined the term “groupthink” to describe what he observed. But the phenomenon he named was not new. It is as old as organized human decision-making, and it is far more destructive, and far more structurally embedded, than the word “groupthink” — which sounds like a quirk, an occasional malfunction — suggests.

The Social Brain Is Not Designed for Truth

Human beings are social animals in a specific and consequential sense: our brains are wired to track and respond to social signals with a priority that systematically overrides our tracking of factual reality. This is not a design flaw. For most of human evolutionary history, being wrong about a fact was far less dangerous than being wrong about your social standing. A person who misjudged the intentions of their tribal allies was likely dead. A person who held a false belief about, say, the behavior of a particular plant was merely less efficient at foraging. The brain optimized for the more dangerous error.

The result is a cognitive architecture in which social information — what other people believe, what the dominant group thinks, what the leader signals — is processed with more urgency and more weight than factual information. When these two sources of information conflict, social information usually wins. This is not a weakness of character or intelligence. It is the normal operation of a brain built for a social environment where consensus and hierarchy were survival mechanisms.

The implications for group decision-making are severe. In a group setting, the brain is simultaneously trying to process factual information and social information, and the social processing tends to dominate. A person in a meeting is not simply thinking about the problem on the table. They are also monitoring the emotional states of other people in the room, tracking who defers to whom, calculating the social cost of disagreement, and adjusting their expressed views to match the perceived consensus. All of this happens largely below the level of conscious awareness. The person genuinely believes they are reasoning about the problem. They are actually reasoning about the social situation.

This is why so many studies of group decision-making find that groups converge on the view of the highest-status member regardless of that member’s actual expertise. The convergence is not the product of persuasion, exactly. It is the product of everyone else in the room recalibrating their expressed views to match the social signal. The highest-status person speaks first, or speaks most confidently, and the group’s collective judgment gravitates toward that anchor not because the arguments were evaluated and found compelling, but because the social cost of sustained disagreement is too high to maintain.

Why Committees Are Systematically Wrong

The committee as a decision-making structure deserves more systematic criticism than it receives. Committees are bad at making decisions not because of occasional failures or because they attract the wrong people, but because their structural properties reliably produce worse outcomes than individual decision-makers facing the same problems.

The first structural problem is that committees reward the expression of views over the expression of uncertainty. An individual thinking alone can sit with uncertainty, can hold multiple contradictory hypotheses, can say “I don’t know” as a genuine resting state. In a committee, every person who speaks is under social pressure to have a position. Uncertainty does not play well in group settings. It signals weakness, indecision, lack of preparation. The result is that individuals who are actually uncertain express false confidence, and the committee’s deliberation is conducted on the basis of inputs that have all been laundered through a confidence-inflation filter. The group ends up with a picture of the situation that looks more certain than it actually is.

The second structural problem is that committees share information badly. Research on group deliberation consistently finds that groups spend most of their discussion time on information that all members already share, and very little time on unique information that only one or a few members possess. This is the opposite of what you would want from a decision-making body. The whole point of assembling a group of diverse experts is to surface knowledge that no single member has. But the social dynamics of group deliberation punish the introduction of novel information (it disrupts the emerging consensus, it signals that you have been holding back, it requires everyone else to update) and reward the reiteration of known information (it creates agreement, it signals team membership, it is emotionally comfortable). Groups are uniquely bad at the thing that is uniquely valuable about them.

The third structural problem is diffusion of responsibility. When a decision is made by a group, no individual feels fully responsible for the outcome. This sounds like it should reduce anxiety and improve reasoning. In practice, it enables lower-quality thinking by each individual, because the internal check — the sense that this is my decision and I will have to live with it — is absent. Studies of individual versus group risk-taking consistently find that groups take more extreme risks than individuals, precisely because the accountability is distributed across enough people that no single person feels the weight of it.

The Experiments That Made Intellectuals Uncomfortable

The psychology of conformity has been studied experimentally since the 1950s, and the results have never stopped being disturbing to people who believe that education, intelligence, and expertise protect against social pressure. They do not.

Solomon Asch’s conformity experiments are the foundational demonstration. Asch showed subjects a card with a single line on it, then asked them to identify which of three comparison lines matched it in length. The correct answer was obvious — the lines differed by several inches. When subjects answered alone, error rates were under one percent. When subjects answered in a group where confederates had been instructed to give the wrong answer, error rates jumped to thirty-seven percent. More than a third of responses were knowingly wrong, simply because the social environment had endorsed the wrong answer. And the subjects who conformed were not doing so consciously. Most reported, when interviewed afterward, that they had genuinely begun to doubt their own perception.

The implications are not limited to simple perceptual tasks. Subsequent research extended the finding to factual judgments, ethical evaluations, risk assessments, and policy preferences. The degree of conformity varies — it is larger when the group is cohesive, when the individual is uncertain, when the stakes appear high — but it is always present. Even people with strong prior views will shift them substantially when placed in a social environment that endorses a different view, even briefly. The shift persists after the social pressure is removed. This is not stubbornness being overcome by persuasion. It is the brain rewriting its model of reality to match the social signal.

Stanley Milgram’s obedience experiments demonstrated a related but distinct phenomenon: the capacity of social authority to override individual ethical judgment. Sixty-five percent of subjects in Milgram’s baseline study administered what they believed were dangerous electric shocks to a stranger, at the direction of a researcher in a white coat, simply because an authority figure told them to continue. The subjects were not sadists or psychopaths. They were ordinary people from the New Haven area responding to an ordinary social dynamic — the presence of an authority who appeared to know what they were doing.

What both sets of experiments reveal is the same underlying truth: individual judgment is not the stable, autonomous, reality-tracking process that we assume it to be. It is continuously calibrated against social information, and under sufficient social pressure, it will track social signals rather than reality. Groups do not aggregate individual judgments. They transform them.

Exceptions and the Conditions That Produce Them

The picture so far is unrelievedly bleak, which should prompt suspicion. Groups sometimes do make good decisions. Markets — which are a kind of distributed group decision-making mechanism — sometimes aggregate information with remarkable accuracy. Juries sometimes deliver correct verdicts. Scientific communities, over long periods, converge on true models of the world. What conditions produce these exceptions?

The conditions are not mysterious, but they are demanding. Accurate group decision-making requires that individual judgments be formed independently before aggregation, that the aggregation mechanism capture the distribution of views rather than collapsing it into a consensus, and that there be a feedback mechanism that penalizes systematic error. Markets work when they work because prices aggregate information from buyers and sellers who have formed their views independently, and because the feedback mechanism — profit and loss — is real and immediate. They fail when participants are heavily influenced by each other’s behavior (herding, bubbles), when the feedback mechanism is delayed or decoupled (leveraged speculation), or when social dynamics infect the price signal (panics).

The implications for organizational design are significant. The structure of most institutional decision-making — committees, boards, leadership teams — is optimized for exactly the conditions that produce groupthink. Meetings are held in which the leader speaks first. Dissent is managed rather than encouraged. Consensus is treated as a signal of quality rather than a warning sign. The people in the room are selected for their compatibility with the existing institutional culture. Almost none of the conditions for accurate group judgment are present, and almost none of the common-sense reforms — better facilitation, devil’s advocate roles, anonymous voting — reliably compensate for the structural defects.

The Stubborn Implication

The research on group decision-making has been accumulating for seventy years. It is consistent, robust, and largely uncontested in academic psychology. Its practical implications have been almost entirely ignored by the organizations that most need to apply them.

This is itself a social phenomenon worth noting. The finding that groups make bad decisions has been studied, published, discussed, and demonstrated repeatedly — and the response of most organizations has been to add more meetings. The social utility of meetings — the sense of participation, the feeling of shared ownership, the performance of deliberation — is so valuable to organizations as social entities that the epistemic costs are simply ignored. Meetings are not primarily decision-making mechanisms. They are social rituals. When you understand them as rituals, their persistence despite their poor decision-making record makes complete sense.

The lesson that should have been learned from the Bay of Pigs is the same lesson that could have been learned from any number of corporate collapses, military disasters, and policy catastrophes before and since: the quality of a decision is not improved by adding people to the room. It is improved by ensuring that the people making the decision have access to accurate information, genuine accountability, and the social freedom to disagree. Most organizations provide none of these things, and then wonder why their collective judgment is worse than their individual talent would predict.

The uncomfortable conclusion is this: for most important decisions, a well-informed individual thinking carefully alone will outperform a committee of equally well-informed individuals deliberating together. This is not an argument for authoritarianism. It is an argument for designing decision processes that treat the social brain’s vulnerabilities as engineering constraints rather than character flaws to be overcome by the right culture or the right people. The brain will always respond to social signals. The question is whether you have designed your decision-making process to work with that fact or against it.