Models for Individual Decision-Making with Social Feedback
Speaker: Andrea Nedic
Series: Final Public Orals
Location: Engineering Quadrangle B327
Date/Time: Tuesday, October 11, 2011, 2:00 p.m. - 4:00 p.m.
To investigate the influence of input from fellow group members in a constrained decisionmaking context, we develop four 2-armed bandit tasks in which subjects freely select one of two options (A or B) and are informed of the resulting reward following each choice. Rewards are determined by the fraction x of past A choices by two functions fA(x), fB(x)(unknown to the subject) which intersect at a matching point ¯ x that does not generally represent globally-optimal behavior. Each task is designed to probe a different type of behavior, and subjects work in groups of five with feedback of other group members’ choices, of their rewards, of both, or with no knowledge of others’ behavior. We employ a soft-max choice model that emerges from a drift-diffusion process, commonly used to model perceptual decision making with noisy stimuli. Here the stimuli are replaced by estimates of expected rewards produced by a temporal-difference reinforcement-learning algorithm, augmented to include appropriate feedback terms. Models are fitted for each task and feedback condition, and we use them to compare choice allocations averaged across subjects and individual choice sequences to highlight differences between tasks and inter-subject differences. The most complex model, involving both choice and reward feedback, contains only four parameters, but nonetheless reveals significant differences in individual strategies. Strikingly, we find that rewards feedback can be either detrimental or advantageous to performance, depending upon the task.
To further investigate social effects and disassociate the behaviors motivated by the reward structure itself from the behaviors caused by social influence, we investigate data from our second experiment: a two-dimensional spatial exploration task in which rewards received are determined by a spatially-dependent schedule whose mean varies along one dimension, with no change in rewards, on average, along the other direction. We examine how rewards may be inferred over the space being explored, and then consider how this reward-inference model may elucidate behavioral changes and different propensities for exploration or exploitation arising from various types of social feedback.