Definition
Collective intelligence is the ability of a group, organisation, or network to solve problems and perform tasks that exceed what any individual member could accomplish alone — arising from the interactions among members rather than from the pooled average of individual abilities.
The term entered popular science largely through Lévy's 1994 book L'Intelligence collective and was given empirical grounding by Anita Williams Woolley and colleagues at Carnegie Mellon, who in 2010 identified a measurable group-level factor (c) analogous to individual IQ. In Focus, Goleman invokes collective intelligence to argue that the social attunement of individual members — particularly social sensitivity — is the key upstream driver of how smart a group actually becomes.
Why it matters
How it works
The c-factor
Woolley's team gave 192 groups a battery of cognitively diverse tasks — brainstorming, puzzle-solving, moral reasoning, negotiation — and found that performance across tasks correlated strongly: groups that did well on one tended to do well on others. This group-level general factor was distinct from the mean IQ of members.
The structural conditions that produced high c were attentional and social, not cognitive in the narrow sense. Groups where one person talked most of the time performed worse; groups where conversational turns were distributed broadly performed better. This mirrors the argument in Focus that attention hoarding — one person consuming the group's collective attention — suppresses emergence.
Social sensitivity as the upstream variable
Social sensitivity — the capacity to accurately read others' emotional states — turned out to be the strongest single predictor. A group of moderately smart, highly perceptive people consistently outperformed a group of brilliant but socially blunt ones. Goleman frames this as an argument for emotional intelligence as a group-level resource: when each member is accurately reading the others' states, the group's information-processing fidelity is higher.
Digital collective intelligence
Research on Wikipedia editing, open-source software, and prediction markets has extended the concept to decentralised networks. James Surowiecki's Wisdom of Crowds (2004) documented that large, independent, diverse crowds reliably outperform individual experts at estimation tasks — provided crowd members are not herding or imitating one another. Herding collapses collective intelligence toward individual errors.