CHIIR 2026 report: Proactive AI agents need calibrated initiative

Interactive information access is shifting from reactive search and query-response paradigms to proactive, agentic systems that do far more than answer. The 1st Workshop on Human-Centered Proactive and Personalized Agents for Interactive Information Access, held at CHIIR 2026, brought together researchers across information retrieval, human-computer interaction, dialogue systems, AI ethics, cognitive science, learning technologies, and human-centered AI. The resulting report, authored by Kirandeep Kaur, Vinayak Gupta, Tanya Roosta, Madhura Raju, Grace Hui Yang, and Chirag Shah, synthesizes presentations and discussions on how agents can personalize interaction, retain context, infer latent needs, recommend next steps, and initiate support.

Central to the workshop was the idea that proactivity is not merely acting earlier or predicting better, but exercising initiative in a way that is appropriately timed, transparent, contestable, and aligned with user goals. Participants tackled a broad set of design challenges: calibrated initiative that knows when to act, knowledge-gap navigation to guide learning, long-term memory for sustained context, value-sensitive design that respects user welfare, implicit personalization without intrusive inference, AI-mediated care, and proactive dialogue. Evaluation was another major concern, with calls to measure outcomes beyond task accuracy. The report frames autonomy, privacy, trust, and transparency as the critical tensions that designers must balance when building agents that take initiative, making it a useful reference for anyone developing proactive AI in interactive settings.