Title:  Computational Sensemaking for Embodied Co-Creative Artificial Intelligence
Date:  November 11th, 2024
Location: TSRB 523A  & Teams
Link: Microsoft Teams Meeting Link
Committee Chair: Dr. Brian Magerko
Committee:  Dr. Anne Sullivan, Dr. Richmond Wong, Dr. Mark Riedl, Dr. Tiffany Knearem
Abstract:  

This research lies at the intersection of human-AI interaction, creative collaboration, embodied cognition, and social cognition. Its central aim is to advance the understanding of how AI agents can actively participate in creative processes traditionally dominated by human-human interactions, such as dance and drawing. Grounded in theories of embodiment and intersubjectivity—which prioritize sensory engagement and interaction over abstract cognition—this work explores the complexities of co-creativity and social cognition through the lens of sensemaking within AI systems.

The research investigates how AI systems can engage in co-creative processes by leveraging sensemaking patterns, both descriptively and generatively. It explores how theories of embodiment and sensemaking enhance our understanding of co-creativity, how sensemaking patterns can be analyzed and integrated into co-creative systems, and how design considerations for future co-creative AI systems can be developed. The study employs a mixed-methods approach, combining qualitative and quantitative analyses through empirical studies, interviews, video coding, thematic analysis, surveys, speculative design, and self-reflective exercises.

Key contributions include a new perspective on computational co-creativity that integrates theories from human-computer interaction, embodiment, and social cognition. The dissertation introduces the Observable Creative Sensemaking (OCSM) framework, a method for quantifying sensemaking in embodied creative improvisation. It demonstrates how OCSM can be used descriptively to compare different co-creative interactions and applied as a generative model to guide real-time improvisation. Additionally, the development of two co-creative AI systems—Drawcto, a multi-agent drawing application, and LuminAI, an embodied improvisational dance system—highlights the practical application of these theoretical frameworks and models in real-world AI systems.