Retrofitting Synthetic Dialog Agents to Game Characters for Lifestyle Risk Training

Susann Luperfoy

This paper describes a game-based Intelligent Tutoring System (ITS) that engages learners in interactive role-play as a form of indirect health communication about lifestyle risks. Successful instruction must not only inform patients of correlations and risks but also equip them to overcome well-known cognitive obstacles to wise lifestyle choices. Through the game, HEDONIST, we mean to address these sub-rational decision processes that are inaccessible to conventional didactic instruction methods. This is an experiment in capturing the pedagogical value of experiential learning, especially cognitive dissonance, learning by teaching others, and the social support value of ad hoc social networks that form spontaneously inside the game and persist outside. For instance, to excel in the game HEDONIST the player must use dialog to persuade Non-Player Characters (NPCs) to modify their simple decision functions by teaching them about the consequences of substance abuse, unsafe sex, eating disorders, smoking, etc. The ITS integrates the commercial game engine NeverWinterNights with intelligent dialog and instructional components that convert 3D graphical game engine characters into intelligent dialog agents. The focus of this paper is the reusable dialog agent architecture and its interface to the backend game-engine system.


This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.