J.A. Jacko (Ed.): Human-Computer Interaction, Part IV, HCII 2011, LNCS 6764, pp. 580–589, 2011. © Springer-Verlag Berlin Heidelberg 2011 The Interaction of Children’s Concepts about Agents and Their Ability to Use an Agent-Based Tutoring System Alicia M. Hymel 1 , Daniel T. Levin 1 , Jonathan Barrett 2 , Megan Saylor 1 , Gautam Biswas 2,3 1 Department of Psychology and Human Development, Vanderbilt University, Nashville, TN 37240, USA 2 Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN 37240, USA 3 Institute for Software Integrated Systems, Vanderbilt University, Nashville, TN 37240, USA {Alicia.M.Hymel,Daniel.T.Levin,Jonathan.I.Barrett, Megan.Saylor}@vanderbilt.edu, Biswas@eecsmail.vuse.vanderbilt.edu Abstract. Computer-based teachable agents are a promising compliment to classroom instruction. However, little is known about how children think about these artificial agents. In this study, we investigated children’s concepts about the intentionality of a software agent they had interacted with and tested whether these concepts would change in response to exposure to the agent. We also tested whether individual differences in concepts about agent intentionality would affect children’s ability to learn from the agent. After repeated exposure to a teachable agent, students did not make more intentional attributions for the agent than a computer, but a general understanding of agency predicted success in learning from the agent. Understanding basic concepts about agency appears to be an important part of the successful design, implementation, and effectiveness of computer-based learning environments. Keywords: learning, artificial agents, theory of mind. 1 Introduction Computer-based learning environments have proven to be a valuable resource for both students and educators. Of particular promise are environments featuring teachable agents – graphical representations of characters that students can teach using speech, text, or visual representations (e.g., [1]). A central assumption underlying these systems is that the agents exhibit behaviors that are similar to human characters, which may invoke cognitive processes for learning and monitoring of one’s learning more effectively than less social systems. However, little is known about the specific cognitions that underlie this putative benefit, and almost nothing is known about the relationship between students’ specific beliefs about these agents and their more general understanding of concepts such as agency. Although students