Collective Intelligence During Consensus Decision-Making in Fish

Collective intelligence is manifest when a group of animals is capable of solving a task that no member of the group alone can. One question crucial to understanding this phenomenon is whether and, if so, how individuals integrate incomplete information to reach effective group level decisions. To investigate this we trained two groups of Golden shiners (Notemigonus crysoleucas) separately to locate a food reward using one of two different stimuli dimensions: e.g., group A was rewarded on a green but not a blue substrate; group B by a striped but not blank wall. Mixed shoals comprising members of both training groups were then tested in the presence of combined stimuli (e.g., green striped, green blank, and blue striped). We find that mixed shoals of shiners can combine preferences from both stimulus dimensions to reach a consensus decision in favour of the only stimulus preferred by both training groups (green striped, in this case) and thus exhibit collective intelligence beyond that of any individual group member. Using computerized tracking, we investigate the mechanisms by which fish achieve this. These results have potential implications for our understanding of analogous behavior in human crowds, and for the design of autonomous robotic swarms.