Abstract
Abstract In 2010, a new research stream began on collective intelligence (CI), defined as a group's general ability to perform consistently well across a wide variety of tasks. Subsequent empirical evidence presents a mixed picture. Some studies have found groups to exhibit CI while others have not. To resolve these disparate results, we compare 21 experimental studies to understand what influences whether groups exhibit CI. We find that task structure is a boundary condition for CI in that groups exhibit CI across well‐structured tasks but not across ill‐structured tasks. For ill‐structured tasks, CI has a more nuanced set of multiple factors that may be interpreted as different facets of CI. This research extends our understanding of CI by suggesting that the original definition of CI was too all‐encompassing. CI should be reconceptualized as a multi‐dimensional phenomenon, similar to research on individual intelligence. We highlight avenues for future research to continue to move CI research forward, particularly regarding ill‐structured tasks.