Options
2026
Conference Paper
Title
A Human-Centered Interaction Framework for Multi-agent AI Reasoning in Enterprise Decision-Making
Abstract
Enterprise decision-making requires processing diverse information, comparing alternatives, and integrating multiple perspectives under uncertainty. These demands have increased interest in AI tools that can summarize information and support complex reasoning. However, most existing systems rely on chat-based or single-perspective interactions that offer limited structure, transparency, or user control. Although recent advances in generative AI and multi-agent LLM reasoning offer richer analytical capabilities, the interaction design needed to make such systems understandable and usable in organizational settings remains underexplored. This paper presents a human-centered interaction framework for multi-perspective AI reasoning, investigating how users prefer to interact with multi-agent systems to enhance trust and robustness. The framework is derived from a requirements study with 16 practitioners and refined through iterative design. The framework structures role-based analysis, staged reasoning flows, layered transparency, and human oversight into a coherent model for engaging with multi-perspective AI output. A high-fidelity prototype demonstrates how the framework can be instantiated, and an assessment with 10 decision-makers indicates clear value in improving clarity, traceability, and cross-functional alignment. The paper concludes with design implications for future enterprise AI systems and argues that human-centered interaction frameworks will be essential as multi-agent AI becomes integrated into real-world decision support.
Author(s)