Affect-Aware Student Models for Robot Tutors

摘要: Computational tutoring systems, such as educational software or interactive robots, have the potential for great societal benefit. Such systems track and assess students' knowledge via inferential methods, popular Bayesian Knowledge Tracing (BKT) algorithm. However, these methods do not typically draw on affective signals that human teachers use to knowledge, indications of discomfort, engagement, frustration.In this paper we present a novel extension BKT model uses data, derived autonomously from video records children playing an story-telling game with robot, infer student reading skills. We find that, compared control group who played only tablet, interacted embodied social robot generated stronger data engagement enjoyment during interaction. then show incorporating into training improves quality learned inference models.These results suggest physically embodied, affect-aware tutors can provide more effective empathic experiences children, advance both algorithmic human-centered motivations further development tightly integrate affect understanding complex models interactive, robots.