Thesis: Narrative Intelligence for Socially Situated Creatures

My thesis research involves rethinking autonomous agent architecture while taking seriously the idea that an agent is not simply a self-contained technology, but a form of communication between people.  There are several important components:

• analysis of Artificial Intelligence and its relationships to industrialization and psychiatric institutionalization;
• development of Socially Situated AI, an approach to agent design that situates agents not only in their physical environment, but also their social and cultural surroundings;
• design of an agent architecture, the Expressivator, which allows agent designers to more effectively control how their agent concept is being communicated.  Agents are made more comprehensible by structuring their behavior into narrative sequences.  The Expressivator involves 3 major components:
1. Behavior transitions, which connect behaviors into meaningful sequences
2. Sign management, which allow agents to be constructed with reference to and to reason about how they are being interpreted by the user
3. Meta-level controls, which provide the low-level hooks by which behaviors can coordinate their effect on the user;
• implementation of the Industrial Graveyard, a virtual environment illustrating the concepts of the thesis and testing its technical feasibility

Back to my work page.