The article discusses a system that uses Large Language Models (LLMs) to facilitate collective decision-making in work contexts like meeting scheduling and project planning. The system extracts individual preferences and suggests options that satisfy most members. The authors applied this system to corporate meeting scheduling, creating synthetic employee profiles and simulating conversations at scale. The results show efficient coordination with reduced interactions between members and the LLM-based system. It also refines proposed options over time, ensuring their quality and equity. The system’s ability to aggregate preferences and reasoning was affirmed through a survey study involving human participants.
Publication date: 3 Nov 2023
Project Page: https://arxiv.org/abs/2311.04928
Paper: https://arxiv.org/pdf/2311.04928
