Collective wisdom in animal groups

Abstract

There is much evidence that sociality can grant animals a wide variety of fitness benefits. A less well explored benefit is the potential ability of social interactions to integrate diverse observations of the environment and perform computations in order to improve collective decision accuracy. However, existing theory has largely been developed with human organizations in mind, and new theory needs to be developed, taking into account the specific abilities and limitations of non-human animals. In this thesis, I explore collective wisdom in animal groups by considering how animals groups make consensus decisions, the internal structure of animal groups, how animals learn about cues in their environment, and the complexity of natural environments. I find that the traditional wisdom of crowds, whereby collective accuracy increases monotonically with group size, is not the norm in many environments. Instead, it may be more typical that small groups maximize decision accuracy. Internal group structure, such as clusters and hierarchies, can lead to information loss within groups, such that large groups behave, in terms of decision-making ability, like a smaller group with no internal structure. This may be a mechanism by which large groups can recover decision accuracy and gain robustness in unpredictable environments. Furthermore, by using simple learning rules within a context in which a group makes consensus decisions, individuals can readily learn a behavior that maximizes not individual, but collective accuracy. I also demonstrate through an experiment with humans that the rules individuals follow to integrate social information with their personal opinions can improve both individual and collective accuracy simultaneously. Together, these results demonstrate that despite cognitive and communicative limitations, there exist simple and robust mechanisms by which group-living animals can gain the benefits of collective wisdom.