SAN2020

Recently, a Bayesian approximation was proposed for understanding pedagogical situations (Shafto et al, 2008)*. In this context, different models have evaluated how a teacher chooses limited examples with the objective of having a learner infer (learn) a concept. But these don’t explore the relevance that knowing your teacher may have on learning. Hence, we wondered how prior information about the teacher would affect learners’ inferences of new concepts.
Extending the Little Teachers project, we present a new paradigm, in which a teacher helps a learner to find a secret box on the screen using a limited set of cues. First, young adults were placed in the learners’ role knowing (or not) that they were receiving cues chosen by 2nd, 4th or 6th graders, and then they were placed in the teachers’ role. Our ongoing results indicate that (1) performance increases when playing consecutively with the same teacher and (2) knowing the teachers’ age has a complex impact: It improves performance at first, but it seems to overshadow the teachers’ particular strategies. In fact, it had a significant negative effect on performance for learners whose teachers were 2nd graders. Interestingly, adults in the teachers’ role chose examples highly biased by the strategies used by their teachers, even when they reported those as “bad”. These results suggest that prior knowledge about teachers can affect learners’ performance and teaching strategies. * Equal Contribution