Investigating Group Decision-Making Mechanism in Decentralized Multi-Agent Collaboration - Nov 25, 2024
Speakers: Young-Min Cho
Biography of the speakers:
Young-Min Cho (known as Jeffrey) is a Ph.D. student in Computer and Information Science at the University of Pennsylvania, specializing in Natural Language Processing and Artificial Intelligence. As a member of Penn NLP, WWBP, and CSL, he is advised by Dr. Lyle Ungar and Dr. Sharath Chandra Guntuku. His research focuses on controlling conversations in Multi-Agent Systems and Conversational Agents, with an emphasis on achieving consensus in agent collaboration, managing conversational behavior traits, and generating effective clarifying questions. He is also exploring social and psychological insights through language-based assessments, including applications such as mental health chatbots and analyses of cultural differences in emotional expressions.
Abstract:
This study investigates the efficacy of Multi-Agent Systems in eliciting cross-agent communication and enhancing collective intelligence through group decision-making in a decentralized setting. Unlike centralized mechanisms, where a fixed hierarchy governs social choice, decentralized group decision-making allows agents to engage in joint deliberation. Our research focuses on the dynamics of communication and decision-making within various social choice methods. By applying different voting rules in various environments, we find that moderate decision flexibility yields better outcomes. Additionally, exploring the linguistic features of agent-to-agent conversations reveals indicators of effective collaboration, offering insights into communication patterns that facilitate or hinder collaboration. Finally, we propose various methods for determining the optimal stopping point in multi-agent collaborations based on linguistic cues. Our findings contribute to a deeper understanding of how decentralized decision-making and group conversation shape multi-agent collaboration, with implications for the design of more effective MAS environments.