Cognitive Computational Model Reveals Repetition Bias in a Sequential Decision-Making Task

Abstract

Humans tend to repeat past actions due to rewarding outcomes. Recent computational models propose that the probability of selecting a specific action is also, in part, based on how often this action was selected before, independent of previous outcomes or reward. However, these new models so far lack empirical support. Here, we present evidence of a repetition bias using a novel sequential decision-making task and computational modeling to reveal the influence of choice frequency on human value-based choices. Specifically, we find that value-based decisions can be best explained by concurrent influence of both goal-directed reward seeking and a repetition bias. We also show that participants differ substantially in their repetition bias strength, and relate these measures to task performance. The new task enables a novel way to measure the influence of choice repetition on decision-making. These findings can serve as a basis for further experimental studies on the interplay between rewards and choice history in human decision-making.