A learned decision rule that selects actions based on the current state of the system (here, accumulated evidence).