Explore-exploit behavior in humans as a sequential sampling process
摘要
Animal foraging behavior often elucidates a tradeoff between exploitation and exploration. Exploitation is sticking with what is familiar to obtain a relatively good reward, whereas exploration is trying out something unfamiliar for potentially greater gains. Animals search their environments, accumulating evidence for either option as they do so. Here, we tested the notion that human decision-making among discrete options can be thought of as an abstract foraging process, where evidence for choosing the exploitative versus exploratory options accumulate according to a sequential sampling process. We use the classical drift diffusion model to test this notion on four separate existing datasets, all of which characterize different explore-exploit phenomena. By testing the drift diffusion model on a diverse set of explore-exploit data and a large sample size, we were able to provide supporting evidence for the notion that explore-exploit decision-making in humans resembles a sequential sampling process exhibited by animals physically.