Brian Magerko
Georgia Tech
Computational approaches to creating theatrical improvisational agents have relied on representing a single aspect of how improvisation is taught, eschewing a deeper understanding of what improvisers actually do on stage as individuals or as a group. We contend that by studying how people improvise on stage we may come to a better understanding of how to create AI for synthetic characters in narrative-based situations, computer games, assistive technology, etc. This talk will present the motivation, experimental design, and current findings for the Digital Improv project, which is dedicated to empirically study the cognition underlying improvisational behaviors in theatre. We will discuss our current findings on improvisation and cognition related to narrative development and cognitive consensus and present video examples of our current computational "micro-agent" representations of these findings. We will also provide a meta-discussion of the role of subject matter in cognitive science research and how a "fun" topic has been useful in engaging and retaining student involvement at the undergraduate and graduate levels.