Description:
We use the probabilistic graphical model (PGM) framework to explore human interaction dynamics in collaborative settings, focusing on false beliefs and team coordination. In studying false beliefs, we construct AI agents within a Minecraft-based urban search and rescue scenario to analyze human players’ beliefs and intentions grounded by evidence about what they see and do. This approach also supports studying the effect of interventions, which are vital if AI agents are to assist human teams. The second aspect of this work is introducing a novel framework to understand interpersonal coordination, viewed as a latent phenomenon explaining statistical temporal influence between multiple components in a system. For example, the state of one person affecting that of another at a later time, as indicated by their behavioral and biological factors (e.g., neural activity). Our models identified coordination patterns by analyzing speech, physiological,and neural data, demonstrating their applicability across different data modalities and time scales. Our proposed evaluation method, validated with synthetic data, effectively measures when coordination is helpful. It offers insights into its predictive relevance for team performance and works as a guide for assessing other latent hypotheses within a generative framework. Remarkably, our findings with real-world data from two datasets, substantially different in their experimental design, underline the model’s ability to explain observed data and predict team outcomes. Further experiments with shuffled data and artificial teams showcase that we can separate coordination due to spurious and random effects from that obtained from genuine interactions. We tested nine different coordination models, ranging from one modality to four, illustrating our framework’s flexibility. Integrating multimodal data enhanced our understanding of team dynamics, though it also suggested that adding more modalities does not linearly increase predictive accuracy of team performance.
Publisher:
The University of Arizona.
Contributors:
Barnard, Kobus ; Pacheco, Jason ; Morrison, Clayton ; Peterson, Mary
Year of Publication:
2024
Document Type:
Electronic Dissertation ; text ; [Doctoral and postdoctoral thesis]
Subjects:
False Belief ; Interpersonal Coordination ; Probabilistic Modeling
Rights:
Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author. ; http://rightsstatements.org/vocab/InC/1.0/
Content Provider:
The University of Arizona: UA Campus Repository
