Why can group decisions outperform individual decisions?

Groups have an advantage when members hold partly different information or approaches and the task lets them demonstrate why an answer is better. The group must actually surface those contributions. Independent first judgments preserve starting differences; explanation makes a proposal inspectable; and a clear decision rule helps the team revise for reasons rather than simply follow confidence or seniority.

• Members form an initial view before seeing the emerging consensus.
• The group has relevant diversity of information, experience, or problem-solving approach.
• People can explain and test the reasoning behind a choice.
• Unique information is actively requested instead of waiting for someone to volunteer it.
• The group distinguishes expertise and evidence from confidence, title, or airtime.
• The final method combines contributions rather than defaulting to the first suggestion.

In a 2006 experiment on highly demonstrable letters-to-numbers problems, groups of three, four, and five solved the tasks in fewer trials than the best of equivalent numbers of individuals. That is strong evidence for those structured, solvable problems—not proof that groups beat their best member on ambiguous workplace judgments.Journal of Personality and Social Psychology via PubMed

No. Classic crowd wisdom normally aggregates judgments that retain some independence. Face-to-face discussion creates social influence: people exchange reasons, but they also react to one another's confidence and preferences. Discussion can improve the inputs, reduce the independence between them, or do both. The design of the process matters as much as the number of people.

A 2018 live-crowd study asked 5,180 people general-knowledge questions, placed them in groups of five for deliberation, and then aggregated several group consensus estimates. In that design, aggregating a small number of group consensuses was more accurate than aggregating the original crowd estimates. The result depends on a two-level method—small-group debate followed by aggregation—and should not be reduced to “any meeting improves accuracy.”Nature Human Behaviour

Discussion can create correlated errors. In an experiment with 144 participants performing simple estimation tasks, even mild exposure to other people's estimates narrowed the range of answers without improving collective error. The crowd became more consistent in the everyday sense of giving similar answers, but not more accurate.Proceedings of the National Academy of Sciences via PubMed

Groups also favor shared information. A meta-analysis covering 65 hidden-profile studies, 101 independent effects, and 3,189 groups found that groups mentioned substantially more common than unique information. Hidden-profile groups were eight times less likely to find the solution than groups whose members all had the full information; measures of pooling unique information were positively related to decision quality. These studies use a particular experimental paradigm, but they identify a practical facilitation risk: a fact known by one person may never enter the team answer.Personality and Social Psychology Review via PubMed

• The first proposal anchors the conversation.
• A senior or confident member is treated as correct without testing the reason.
• People repeat common facts because those facts receive quick agreement.
• Unique evidence arrives after the group has publicly committed to a choice.
• Conflict avoidance turns consensus into the objective instead of decision quality.
• Everyone relies on the same source, so errors are correlated rather than canceled.

The claim is incomplete until it names a benchmark. A team can beat the average individual while still finishing behind its best member. It can produce a higher average result with greater variability, or become more consistent around a worse answer. Each comparison answers a different question.

Lower variance is not automatically desirable. It means observations are more tightly clustered. Whether that is useful depends on where they are clustered, whether lower or higher scores are better, what target matters, and how costly extreme errors would be. Likewise, “highest potential” is too vague for a result statement; name the best observed score, the theoretical limit, or the benchmark you actually measured.

1. 01

   Collect independent first views

   Ask each person to record a choice and reason before open discussion. Do not display the emerging distribution while people are still forming that view.

2. 02

   Invite unique information

   Ask what one person knows that the rest of the group may not. Round-robin evidence before debating preferred answers.

3. 03

   Separate reasons from people

   Write the evidence and tradeoffs where the group can inspect them without tying every point to status or personality.

4. 04

   Test the leading answer

   Ask what would make the current choice wrong, which assumption is carrying the most weight, and what evidence has not been reconciled.

5. 05

   Record the comparison honestly

   Use the same task, reference, and scoring direction for individual and group answers. Report missing work as unavailable rather than zero.

6. 06

   Debrief the process

   Ask which information changed the decision, who influenced the revision, what was overlooked, and whether the team would use the same process again.