Kyle Fitzgerald, Department of Applied Mathematics, University of Colorado Boulder

A Coupled Fokker-Planck Model for Collective Decision-Making

One way to model decision-making in a variable environment is via stochastic evidence accumulation. Specifically, we use coupled Fokker-Planck equations to model noisy evidence accumulation in a two-alternative forced choice (TAFC) task, where an undecided agent immediately receives a discontinuous "kick" towards the decision threshold at which the first agent has decided. The question we then investigate is: if each agent receives a reward for making a "correct" decision, how might each agent set (i) its own decision thresholds and (ii) the weight it places on previous agents' decisions, in order to maximize the associated collective reward rate? And how does the optimal strategy change as the amount of reward given for a correct decision changes? Of particular interest are the implications of asymmetric decision thresholds on decision-making, in which case the agent requires less evidence to choose one choice over the other, thus representing an inherent bias in the agent's preferences.