Human Verification in the Age of AI Agents
Description:
As AI systems move from producing individual predictions toward performing extended and increasingly autonomous tasks, the role of humans may shift from directly performing work to overseeing, verifying, and occasionally intervening in AI-generated work. This shift raises new questions for visualization and human-AI interaction: What information should an AI system expose so that a human can efficiently verify its work? How should interfaces support transitions between observation, verification, intervention, and takeover? When explanations are costly to inspect, what makes an AI system's behavior sufficiently legible for human oversight? And how might visualization help humans identify where attention and verification are most needed? This meetup will bring together researchers and practitioners interested in visualization, visual analytics, explainable AI, human-AI interaction, uncertainty, provenance, and AI agents. Rather than focusing on a particular system or solution, the meetup will provide an informal forum for identifying emerging challenges, sharing relevant experiences and methods, and discussing a research agenda for visualization-supported human verification of increasingly agentic AI systems. Prospective audience: VIS researchers and practitioners working in visualization, visual analytics, human-AI interaction, explainable AI, AI-assisted decision-making, uncertainty, provenance, and agentic AI. Participants from adjacent HCI and AI communities would also be welcome.
Jump to DiscussionPresentations in this session
-