Much of what we learn, we learn from others. What guides learners' choice of informants? Research suggests that learners resist informants who provide incorrect information or insufficient information for accurate inference. Here we propose that learners’ choices of informants are rationally guided by the extent to which evidence supports accurate inference, rather than the sheer amount of evidence provided. Extending recent research formalizing pedagogical reasoning, we propose a computational model of efficient teaching. We present an experiment on adults testing three different hypotheses about learners’ preferred level of the amount of data. The results suggest that learners care about the inductive sufficiency of evidence, rather than the amount of evidence provided. We conclude by discussing the implications of these findings for cognition and cognitive development.