The Multi-agent System based on LLM for Online Discussions

The considerable improvement on the Internet and the corresponding applications leads to the result of online discussions becoming far more popular and significant than any other method for people to communicate with each other and reach a consensus. Meanwhile, the incredible improvement in Large Language Models (LLM) has promoted the performance of LLM-based agents in text understanding and content generation capabilities. The research objective of the PhD thesis is to build democratic discussion environments, with three main issues existing right now: 1) Large-scale discussions tend to be complicated, 2) Rumours and misinformation bring negative effects to the discussions, and 3) Direct democratic discussions are complex and time-consuming. This extended abstract introduces the efforts that have been made to address those issues, with the introduction of the potential directions in the future.