Abstract
Many Interactive Storytelling (I-S) systems model story characters as quasi-autonomous agents that interact with one another under the guidance of another software component usually called the Drama Manager or Director. While the character agents for these systems have been constructed using a variety of conventional AI techniques, in this report we investigate whether such an agent can be constructed using a suitable cognitive architecture as its substrate.We investigated the feasibility of constructing character agents by adapting the cognitive architecture CLARION (Sun 2003) as the foundation for its AI. CLARION is a unique cognitive architecture that features both symbolic and sub-symbolic processing, as well as subsystems for handling agent motivation and metacognition tasks. CLARION itself was constructed in part to model multi-agent and social processes.
Our investigation suggests that these and other capabilities of CLARION make it uniquely suited to the challenge of constructing story agents, albeit not without a fair amount of modification. These architectural implications in the context of desiderata for storytelling agents in general.
This work is part of MISSI (Multi-Agent Interactive Storytelling Software Initiative), an informal research project at RPI for designing and building an Interactive Storytelling system. This project affords undergraduates in the Games and Simulations Arts and Sciences (GSAS) program opportunities for contributing original research as part of the GSAS program's requirements.
Bibliography
Lynch, M. (2008). Adapting the CLARION Cognitive Architecture for Interactive Storytelling Agents. (Submitted to ICA for publication.)Reiss, S. (2004). Multifaceted Nature of Intrinsic Motivation: The Theory of 16 Basic Desires. Review of General Psychology. 2004, Vol. 8, No. 3, 179–193.
Sun, R. (2003). Tutorial for Clarion 5.
Sun, R. (2005). The CLARION Cognitive Architecture: Extending Cognitive Modeling to Social Simulation.