Agentic AI Requires More Explanation, Not Less Interaction

A new paper from arXiv argues that as AI systems become more agentic—acting autonomously on behalf of users—the need for user interaction shifts from routine back-and-forth to increased communication for oversight and explanation. The authors propose that agentic AI must incorporate three types of explanations: action-process, uncertainty, and coordination. Grounded in communication theory, the paper examines how users perceive AI's communicative roles (as source or channel) and how this perception affects trust. The study highlights risks when agentic AI plays multiple roles, complicating source perception. The preprint (arXiv:2605.01610) was announced on May 1, 2025.