What We Leave Behind: Delegating Sensemaking in the Age of Generative AI

Agentic data analytics promises better and more accurate findings in less time and effort, but even if we accept these promises at face value, surely they cannot come for free? In this talk, I will examine the emerging trend of agentic AI for sensemaking with an eye not only for its benefits, but also for its costs. When an LLM coding agent builds an interactive dashboard for me, what of my control over the analytical decisions? When an automated design assistant picks the best chart for me, what of my own aesthetic and personal taste? When a conversational agent prompts me for my reactions, what of my own initiative? To answer these questions and more, I will present several design probes where my collaborators and I explore mechanisms for agent-based visualization and analysis. I will close with ideas for measuring phenomena such as delegation, fluid agency, and skill atrophy, and with a framework for modeling and designing human-centered AI tools for data analysis.