Prawns and Probability: Model comparison approaches in collective behaviour
9 March 2012
Presented by: Dr. Richard Mann (Uppsala University)
How do groups of animals or humans act cohesively and make collective decisions? How do complex patterns of collective motion emerge in groups of individually simple organisms? Simulation studies show that simple interaction rules between individuals and their local neighbours are sufficient to produce complex group behaviours, but the identity of these rules remains unknown. Recent studieshave focused on inferring interaction rules by analysing regularities in thegroup configuration and the looking for correlations between the motion of individuals and the local configuration of their neighbours, a so-called 'data-driven' approach.
An alternative methodology is theory-driven model comparison. Rather than looking for significant correlations between measured responses and putative cues we phrase alternative hypothesised interactions as models which predict the behaviour of individuals, using a Bayesian model comparison to select between competing theories. Crucially we ask for more than just a rejection of the null hypothesis - we must show that our theories describe the data better. In this talk I will show examples of this approach in the context of the collective motion of glass prawns; co-navigation in homing pigeons; interactions between mosquito fish; decision making in damsel fish; and theonset and cessation of audience applause and I will discuss how theory, model comparison and data-driven research can be combined to best effect.
Time: 4pm followed by drinks in the Macleay foyer from 5pm
Location: Quadrangle Building, History Room S223