Technological advancements are creating new opportunities for teams of AI agents, such as robots and unmanned ground/aerial vehicles, to be deployed alongside humans to support and augment labor-intensive and/or dangerous manual work. The vision is for robots to do more than free up humans so they can focus on the tasks that they are skilled at, like planning and dexterous manipulation, but also work collaboratively with them to solve problems that would be difficult, dangerous, or otherwise impossible.
A version of this vision is being realized across many sectors, including warehouse management, mission planning, and disaster response. However, AI agents exhibit brittle autonomy and poor adaptability to changing human behaviors and strategies, which are commonly undefined/unknown to AI agents and can also be altered substantially by the presence of AI agents.
Over the next four years, Electrical and Computer Engineering Professor Tian Lan will be leading the project, “CHASE: Cultivating Human-AI Synergy via Decentralized Elicitation and Learning,” to tackle this challenging problem of decentralized preference elicitation and learning. The study is supported by a $1.5 million grant from the Office of Naval Research’s Science of AI program.
Alongside co-investigators Professor Taeyoung Lee from the Department of Mechanical and Aerospace Engineering and Professor Mahdi Imani from Northeastern University, the team will investigate novel AI/ML capabilities of decentralized preference elicitation and learning to make AI agents better understand human behaviors/strategies in shared environments, thus supporting teams of human and AI agents working together to solve complex problems.
“Very excited to kick off this interdisciplinary project,” Lan stated.
Their findings are expected to have a deep impact on many real-world scenarios where the collaborative team of human and AI agents takes advance planning and immediate actions, such as assistive autonomy, crisis response, search and rescue, and mission planning. Ultimately, this will help human-AI teams address issues such as natural and man-made disasters, crises, urgent incidents, and critical events in ways not possible today.