Engendering believable communicative behaviors in synthetic entities for tactical language training: An interim report

In this paper we describe an ongoing research and development effort that supports the representation of coordinated verbal and nonverbal behaviors in synthetic characters. The ultimate goal of this effort is to allow a user to populate a virtual environment with characters that exhibit believable communicative competence. In this way, we will make computerized language training more engaging and more effective. Our work is focused on developing a middleware that supports two levels of abstraction. The first allows the user to specify a character's communicative intent and the relevant contextual attributes that underlie dialog in the synthetic environment. The second allows the user to specify how those intents might be variously realized in the virtual environment at run time. Although our efforts are geared toward improving the realism of a particular simulation-based training system, this work also demonstrates the reciprocal ability of modeling and simulation to advance theory; insofar as the development of a high-level specification constrains the behavior of synthetic entities, it also lays bare theoretical assumptions about what we take to be the salient aspects of those behaviors and puts them to the test in a synthetic environment.