The Good, the Bad, and the Irrelevant: Neural Mechanisms of Learning Real and Hypothetical Rewards and Effort

摘要: Natural environments are complex, and a single choice can lead to multiple outcomes. Agents should learn which outcomes due their choices therefore relevant for future decisions stochastic in ways common all irrelevant between options. We designed an experiment human participants learned the varying reward effort magnitudes of two options repeatedly chose them. The associated with was randomly real or hypothetical (i.e., only sometimes received magnitude chosen option). real/hypothetical nature on any one trial was, however, learning longer-term values choices, ought have focused informational content outcome disregarded whether it reward. However, we found that showed irrational bias, preferring had previously led, by chance, last trial. Amygdala ventromedial prefrontal activity related way participants9 were biased receipt. By contrast, dorsal anterior cingulate cortex, frontal operculum/anterior insula, especially lateral cortex degree resisted this bias effectively manner guided aspects more sustained relationships particular suppressing information optimal decision making. SIGNIFICANCE STATEMENT In complex natural environments, Human agents from not without such relationship. measure about environment other features random no relationship choice. that, although people could magnitudes, they nevertheless irrationally toward repeating certain as function presence absence features. Activity different brain regions either reflected resistance bias.