A Human-Like Intelligent Agent with Applications in Intelligent Tutoring
CMU
Abstract
Building an intelligent agent that simulates human learning of problem solving could benefit both cognitive science, by contributing to the understanding of human learning, and artificial intelligence, by advancing the goal of creating human-level intelligence. Furthermore, such an intelligent agent can be used for automated creation of intelligent systems such as intelligent tutors.
However, constructing such a learning agent currently requires time-consuming manual encoding of prior domain knowledge. Previous research has shown that one of the key factors that differentiates experts and novices is their different representations of knowledge. In this work, I propose an efficient algorithm that acquires representation knowledge in the form of hierarchical perceptual features, and integrate this algorithm into a learning agent, SimStudent. I show that integrating representation learning into skill learning reduces the requirements for knowledge engineering, and that the extended SimStudent can be used to discover more accurate student models.
Bio: Nan Li is a PhD candidate in the Computer Science Department at Carnegie Mellon University. Her advisors are Ken Koedinger and William Cohen. Her research interests include artificial intelligence, machine learning, human computer interaction, and their applications to educational technologies. More specifically, she has been focusing on building intelligent agents that simulate human-level learning of complex problem solving using machine learning techniques.
