Human value learning and representation reflect rational adaptation to task demands.

Humans and other animals routinely make choices between goods of different values. Choices are often made within identifiable contexts, such that an efficient learner may represent values relative to their local context. However, if goods occur across multiple contexts, a relative value code can lead to irrational choices. In this case, an absolute context-independent value is preferable to a relative code. Here we test the hypothesis that value representation is not fixed but rationally adapted to context expectations. In two experiments, we manipulated participants' expectations about whether item values learned within local contexts would need to be subsequently compared across contexts. Despite identical learning experiences, the group whose expectations included choices across local contexts went on to learn more absolute-like representation than the group whose expectations covered only fixed local contexts. Human value representation is thus neither relative nor absolute but efficiently and rationally tuned to task demands.