Proceedings of Intelligent Virtual Agents 2012, Springer, page 15-27.
Abstract
Over the past couple of decades, virtual humans have been
attracting more and more attention. Many applications including, video
games, movies, and various training and tutoring systems have benefited
from work in this area. While the visual quality of virtual agents has
improved dramatically, their intelligence and socialization still needs improvement. In this paper, we present work towards endowing agents with
social roles and exploiting Explanation-Based Learning (EBL) to enable them to acquire additional, contextual behaviors from other agents.
These virtual humans are capable of learning and applying role related actions from multiple agents and only adopt behaviors that have been
explained to them, meaning that their denition of a role may be a subset from one or more agents. This results in emergent behaviors in
heterogeneous populations.