Toward the Holodeck: Computational Models of Narrative and Their Relation to Human Cognition

Rogelio Cardona-Rivera, North Carolina State University

Narratologists, anthropologists, psychologists and artificial intelligence researchers suggest that narrative intelligence (the generation and comprehension of stories) depends on competencies that distinguish us from our primate relatives. Stories play a foundational role in our cognition and are ubiquitous: They can serve as a target of interpretation and as a framework to understand the world around us. Driven by the recognition that the study of narrative is a worthwhile endeavor and that computational modeling is well suited to precisely understand this complex human phenomenon, my research focuses on creating computational-cognitive models of narrative. In particular, I characterize the cognitive processes at play in interactive narrative contexts (i.e. video games) that require a player to understand a narrative setting as well as her role and options for action within it. Key to this understanding is memory performance, which plays an important role in a person’s projection of a fictional world by shaping expectations for the future development of a narrative. In essence, working memory guides the prediction of future human action and I model it to understand what I define as “narrative affordances” – courses of action that a game player can imagine as part of a story that completes her current story experience. In this talk, I will cover the context and development of my computational model of narrative memory and outline the remaining work that I expect to complete over the coming year to validate it as a plausible model of narrative memory performance. I also will focus on the model's capability to guide the computational generation of narrative that cares to achieve a specific mental configuration in the mind of a human consumer. Finally, I will highlight some exciting future directions that leverage high-performance computing to explore narrative intelligence cognitive modeling in dynamic and stochastic environments.

Abstract Author(s): Rogelio E. Cardona-Rivera