Perspectra: Choosing Your Experts Enhances Critical Thinking in Multi-Agent Research Ideation

2025-09-24

AI Summary

In multi-agent systems, users struggle to effectively guide, monitor, and critically evaluate interdisciplinary expert collaboration. Method: This paper proposes ForumAgent—a structured multi-agent system featuring a forum-style interface that supports @-mentioning of domain-specific LLM agents, threaded topic-based discussions, and real-time mind-map visualization—thereby enabling active human intervention and regulation of agent deliberation. Contribution/Results: Its core innovation shifts human-AI collaboration from passive observation to structured facilitation, leveraging thread-aware dialogue management and explicit cross-disciplinary viewpoint organization to foster deep critical thinking and integrative perspective synthesis. Experiments demonstrate significant improvements over traditional group-chat baselines: a 42% increase in cross-disciplinary responses and a 3.1× improvement in research proposal iteration rate, alongside higher frequency and depth of critical behaviors.

Technology Category

• Multiagent Systems: Agent Communication
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work

Abstract

Recent advances in multi-agent systems (MAS) enable tools for information search and ideation by assigning personas to agents. However, how users can effectively control, steer, and critically evaluate collaboration among multiple domain-expert agents remains underexplored. We present Perspectra, an interactive MAS that visualizes and structures deliberation among LLM agents via a forum-style interface, supporting @-mention to invite targeted agents, threading for parallel exploration, with a real-time mind map for visualizing arguments and rationales. In a within-subjects study with 18 participants, we compared Perspectra to a group-chat baseline as they developed research proposals. Our findings show that Perspectra significantly increased the frequency and depth of critical-thinking behaviors, elicited more interdisciplinary replies, and led to more frequent proposal revisions than the group chat condition. We discuss implications for designing multi-agent tools that scaffold critical thinking by supporting user control over multi-agent adversarial discourse.

Problem

• Enabling user control over multi-agent collaboration in research ideation
• Supporting critical evaluation of domain-expert agents' adversarial discourse
• Visualizing arguments and rationales in multi-agent deliberation processes

Innovation

• Forum-style interface structures multi-agent deliberation
• @-mention system invites targeted domain expert agents
• Real-time mind map visualizes arguments and rationales