Strategies using recent feedback lead to matching or maximising behaviours

Abstract

One challenge facing humans (and nonhuman animal) is that some options that appear attractive locally may not turn out best in the long run. To analyse this human learning problem, we explore human performance in a dynamic decision-making task that places local and global rewards in conflict. We found that experiences that included previous choices and rewards are not easily incorporated into people’s strategy to enhance their performance. Our results suggest that humans are easily driven by concerns about recent feedback, and that choice of a suboptimal behaviour option may be overcome by providing informative cues that indicate a clear immediate outcome for a better option.