Date: Tuesday, May 5th, 2026

Time: 11am - 1pm EST

Location: Tech Square Research Building Room 223

Virtual (Zoom): https://gatech.zoom.us/j/96614909235

Qiao Zhang

Ph.D. Student in Computer Science

School of Interactive Computing

College of Computing

Georgia Institute of Technology

Committee:

Dr. Christopher J. MacLellan (Advisor) – School of Interactive Computing, Georgia Institute of Technology

Dr. Ashok Goel – School of Interactive Computing, Georgia Institute of Technology

Dr. Mark Riedl – School of Interactive Computing, Georgia Institute of Technology

Dr. Erik Harpstead –  Human-Computer Interaction Institute, Carnegie Mellon University

Abstract

Human-AI Teaming (HAT) is an increasingly important area of research as AI systems are integrated into complex, collaborative environments. AI agents were developed and applied in multiple real-life cases, including chatbots and personal assistants for single, well-defined tasks, embodied agents that operate in a self-contained environment, and AI teammates that support collaborative reasoning and decision-making. Although games have been widely used for agent training and evaluation, much existing work has focused primarily on improving agent capabilities or benchmarking performance against human experts. Fewer studies explored human-AI interaction in game settings, positioning AI agents as teammates rather than competitors. Similarly, research on human-AI and human-robot collaboration often positions AI as an assistant rather than a teammate, limiting AI’s autonomy and agency compared to human partners. These limitations create research gaps and opportunities to identify tasks that require true collaboration, arrange team composition, understand teaming dynamics, and examine strategy alignment.

In this thesis, I use Dice Adventure, a customizable multiplayer cooperative game developed in-house, to investigate three key aspects of human-AI teaming: how team structure influences performance, how communication conventions emerge in human teams, and how humans adapt when collaborating with AI under varying degrees of strategy alignment. My work aims to generate insights that inform the design of more adaptive and human-compatible AI teammates.