Report on The 1st Workshop on Human-Centered Proactive and Personalized Agents for Interactive Information Access at CHIIR 2026

Proactive AI systems require careful calibration of when and how they take initiative. The workshop emphasized that simply making AI systems act earlier or predict better is insufficient; proactivity must be "appropriately timed, transparent, contestable, and aligned with user goals" to serve users effectively rather than manipulate them.

Transparency and user control are essential for building trust in personalized agents. As these systems infer latent needs, retain long-term memory, and make recommendations, the workshop identified transparency and the ability for users to contest AI decisions as critical safeguards against systems that operate as "black boxes."

Personalization raises significant privacy and autonomy concerns that require intentional design choices. The workshop highlighted "implicit personalization" and "value-sensitive design" as key topics, suggesting that personalized systems must be deliberately designed to protect user privacy and autonomy rather than treating these as afterthoughts.

Current evaluation methods focused only on task accuracy are inadequate for assessing proactive agents. The workshop surfaced the need for "evaluation beyond task accuracy," indicating that responsible AI assessment must measure whether systems respect user welfare, maintain appropriate autonomy, and align with human values—not just whether they complete tasks efficiently.

Designing helpful AI requires interdisciplinary collaboration across ethics, human-computer interaction, and technical fields. The workshop brought together experts from information retrieval, AI ethics, cognitive science, and human-centered AI, suggesting that responsible proactive agents cannot be developed by technologists alone.