Computational Modeling of Pedagogical Reasoning

Much of human learning goes on in social situations. Among these situations, pedagogical situations stand out as potentially the most important. In pedagogical situations, a person (a teacher) chooses data for the purpose of helping another person learn a concept. Shafto will present a computational model of pedagogical data sampling, which formalizes the problem as complementary inferences on the parts of both the teacher and learner. Shafto will present a series of experiments that test the model's predictions about reasoning with adults and children. He will conclude by sketching a larger picture, in which pedagogical sampling is a special case of reasoning about intentionally sampled data, and outline directions of future research in this context.