Here’s a question you’d think would have an obvious answer. When facing a complex decision, are you more likely to reach a good outcome by thinking it through alone, or by discussing it with a group?
Most people’s instinct is a vague probably a group’s better, reasoning that more minds catch more mistakes, more perspectives surface more options, more discussion produces better outcomes. This is broadly supported by research. But there’s an important qualification hidden in the broad finding. Some groups make substantially better decisions than individuals. Others make substantially worse decisions. What distinguishes the two has been studied carefully, and the answer is specific enough to be practical.
The collective-intelligence research
The most systematic research on this comes from a team led by the American organisational scientist Anita Woolley at Carnegie Mellon. Her 2010 paper, Evidence for a Collective Intelligence Factor in the Performance of Human Groups, reported on studies in which groups of strangers worked together on diverse tasks — brainstorming, moral reasoning, negotiation, planning — and their performance was measured across all tasks.
Woolley’s team found that groups varied in performance in ways that couldn’t be explained by the individual intelligence of their members. Some groups consistently outperformed others on many kinds of tasks, suggesting there was a general group-level capacity — which she called a c-factor, analogous to the individual g-factor of general intelligence — that some groups had more of than others.
What predicted this c-factor wasn’t what most people would guess. It wasn’t the average IQ of group members; that had almost no relationship to collective performance. It wasn’t the intelligence of the smartest member; that had only a weak relationship. What predicted collective intelligence was three specific group characteristics:
Social sensitivity among members. Groups where members were good at reading each other’s emotional states — measured by a standard test of emotional intelligence — performed better on collective tasks. The mechanism seems to be that socially sensitive members notice when others are hesitating, disagreeing silently, or holding back, and create space for those contributions to surface.
Equal turn-taking. Groups where conversation was roughly balanced across members performed better than groups dominated by one or two speakers. Domination by a single voice reduces collective intelligence substantially, even when the dominant voice belongs to the most individually capable member.
More women. Groups with more women tended to score higher on collective intelligence, apparently because women scored higher on average on the social-sensitivity measure that mattered.
These findings have been replicated and extended. The broader finding — that collective intelligence is genuinely different from the sum of individual intelligences, and that it’s produced by specific group dynamics — has held up well.
Google’s Aristotle
A related and widely-cited study was conducted by Google itself, starting in 2012, under the name Project Aristotle. Google’s people-analytics team had access to data on thousands of their own teams and wanted to identify what distinguished the effective teams from the ineffective ones. The project examined dozens of variables — team composition, demographic diversity, social ties outside work, seniority distributions, and many others — looking for patterns.
The striking result was that most of these variables had surprisingly little predictive power. Teams with similar composition could perform very differently. What ended up mattering most was a specific characteristic the researchers called psychological safety — the shared belief that the team was a safe place to take interpersonal risks. In teams with high psychological safety, members could admit mistakes, ask questions that might seem naive, challenge ideas, and share half-formed thoughts without fear of being punished or dismissed. In teams with low psychological safety, these behaviours were rare, and the team’s capacity to learn and adapt was correspondingly limited.
The concept of psychological safety had been developed earlier by the Harvard organisational psychologist Amy Edmondson, who had studied it across healthcare teams, manufacturing teams, and other settings. Edmondson’s extensive research programme has consistently found that psychological safety predicts team performance, particularly in contexts where learning, adaptation, and complex problem-solving matter.
This aligns with the Woolley findings. Social sensitivity and equal turn-taking are, in effect, mechanisms by which psychological safety gets created. When members feel safe, they contribute. When they don’t, they withdraw, and the group’s collective capacity shrinks to whatever the dominant voices can produce alone.
When groups do worse than individuals
Before concluding that groups are always better, it’s important to note the situations where the research is clear that groups perform worse than individuals working alone.
Creative idea generation. Classic research on brainstorming, from the 1950s onward, has consistently found that individuals generating ideas alone produce more and better ideas than groups of the same size brainstorming together. The dynamics of group conversation — one person talking at a time, social pressure to stay on topic, anchoring on early suggestions — substantially reduce the variety and quantity of ideas produced. The classic solution is nominal group technique — have members generate ideas independently first, then come together to share and build on them. This captures the benefits of multiple perspectives without the losses from group dynamics.
Time-pressured decisions. Groups deliberate more slowly than individuals decide. In true emergencies — where a decision is needed in seconds or minutes, where delay produces harm — individual authority usually outperforms group process. This is why command structures exist in militaries, emergency responses, and surgical teams. The discussion that would improve decisions under normal conditions becomes counterproductive when time is the binding constraint.
Highly technical judgements. Where a problem has a clearly correct answer accessible only through specialised knowledge, the group’s judgement is no better than the best member’s judgement, and often worse if the best member is outvoted by less expert members. Diagnosing a rare medical condition, for example, is better done by a specialist than by a committee.
Situations where groupthink takes hold. The American psychologist Irving Janis documented in the 1970s that cohesive groups under pressure can develop what he called groupthink — a collective mode in which the desire for group harmony overrides realistic appraisal of alternatives. Dissenting views are self-censored, consensus is prematurely reached, warnings are dismissed, and the group can collectively decide on courses of action that any of its individual members would have rejected if they’d been asked separately. Janis’s original cases were major American foreign-policy disasters — the Bay of Pigs, Vietnam escalation — but the pattern appears in corporate decisions, community decisions, and family decisions too.
The diversity-of-perspectives finding
A related finding comes from the American social scientist Scott Page, whose research on diversity and problem-solving produced what’s sometimes called the diversity bonus. Page’s mathematical models, supported by empirical research, suggest that groups of moderately-skilled people with diverse perspectives often outperform groups of highly-skilled people with similar perspectives — provided the problem being solved is genuinely complex. On simple problems, expertise dominates. On complex problems, diverse approaches matter more.
This helps explain why groups that look, on paper, like they should be highly effective — selected for individual brilliance and common background — sometimes underperform groups that look less impressive but contain more varied ways of thinking. The varied ways of thinking become advantageous exactly when the problem is hard enough that no single perspective is adequate.
The Page finding has real policy implications. It suggests that diverse hiring isn’t only a matter of fairness (though it is that); in certain problem domains, it’s also a matter of better outcomes. Organisations facing complex problems benefit from teams with genuinely different backgrounds, training, and mental models. Homogeneous teams have systematic blindspots that even very talented individuals can’t compensate for.
What this suggests practically
For a student or early-career professional learning to work with groups, several working principles emerge.
Think about the type of decision before choosing how to make it. Creative generation? Do it individually first. Technical judgement with a clear answer? Use the expert. Complex problem requiring diverse perspectives? A real group discussion is probably worth the time. Emergency? Single authority.
In group decisions, watch the conversational dynamics, not just the content. Who’s speaking? Who isn’t? Is turn-taking roughly equal? Are quiet members being drawn out, or are dominant members filling the space? The dynamics are often more predictive of the eventual decision than the arguments being made.
Create psychological safety deliberately where you have influence. The specific behaviours that produce it are small and learnable. Acknowledging your own mistakes. Asking genuine questions rather than rhetorical ones. Responding to dissent with interest rather than defence. Not punishing people for the content of their contributions. These behaviours, practised over time, produce a substantially different team dynamic.
When you’re in a position to lead group decisions, don’t assume consensus means agreement. Sometimes it means everyone has self-censored. One of the most useful habits is deliberately asking for disagreement, especially from people who haven’t spoken. What are we missing? Who sees this differently? What would make this decision wrong? The questions signal that disagreement is welcome and often surface real concerns that were otherwise going to stay hidden.
Recognise when the group is slipping into groupthink. The warning signs Janis identified are specific: pressure toward unanimity, self-censorship of doubts, stereotyping of outsiders who disagree, illusions of invulnerability, rationalisation of warnings. A group that’s showing these patterns is producing a collective decision that no individual member would endorse if asked alone. Breaking the pattern — often by introducing outside perspectives, or by having members privately rate their confidence before speaking — can substantially improve the decision that eventually emerges.
The question that remains
The deepest thing the collective-intelligence research teaches is that groups are not automatically smarter than individuals. Their intelligence depends on the specific conditions under which they operate. Groups with psychological safety, social sensitivity, equal participation, and diverse perspectives genuinely outperform what any individual could do alone. Groups without these conditions are often slower and worse than individual decision-makers would have been.
What this implies is that leading or participating in groups well is a specific skill, learnable but not automatic. The person who knows how to create the conditions under which groups make good decisions is doing something valuable that many people in positions of authority don’t know how to do. Over a career, the capacity to produce good collective decisions probably matters more than the capacity to be individually brilliant.
The question worth carrying, the next time you’re in a meeting that’s about to make a consequential decision:
Are the conditions in this room producing collective intelligence, or producing collective mediocrity — and what’s actually going to determine which one this meeting delivers?
Key research referenced: Anita Woolley and colleagues’ 2010 Science paper on collective intelligence; Google’s Project Aristotle (publicly reported 2015-2016); Amy Edmondson’s research on psychological safety (The Fearless Organization, 2018); Scott Page, The Difference (2007); Irving Janis’s research on groupthink (Victims of Groupthink, 1972).