Proceedings of the Annual Meeting of the Cognitive Science Society

When communicating our conceptual knowledge to others, we can either provide examples, or provide the defining rule of the concept directly. Here, we propose a normative computational theory of when people should use each mode of transmission. Our model is based on the idea that teachers who are uncertain about a target concept prefer to induce a similar structure of uncertainty in the learner's mind. Simulations show that example-based teaching more effectively communicates conceptual uncertainty, whereas rule-based teaching is more efficient when teachers are confident. A pre-registered study using a "number game" paradigm (n=100) supports this account: participants assigned the role of teachers made choices that align with the model predictions when deciding whether to communicate examples or rules. Subsequent participants (n=100) that learned from these messages showed uncertainty patterns that sensitive to the mode of communication, and were similar to the uncertainty patterns of their teachers.