New 'Proactive Computing' paradigm aims to replace reactive AI

Joonhee Lee from the University of California, Irvine, has published a 28-page survey on arXiv outlining the emerging paradigm of *Proactive Computing*—a system-level shift from reactive tools to AI that senses, predicts, and acts *before* users explicitly request it.

The paper defines proactive computing as a fusion of sensing (wearables, ambient devices), foundation models (for context inference), distributed edge infrastructure (for low-latency processing), and physical actuation (e.g., adjusting smart home devices). Lee distinguishes this from reactive, context-aware, or predictive computing by emphasizing *system-level integration* across five layers: sensing, understanding, decision-making, action, and governance. Key challenges include uncertainty-aware triggering (knowing *when* to act), the prediction-to-action gap (bridging forecasts with real-time decisions), and socio-technical hurdles like privacy, accountability, and user trust. The core research question isn’t just better predictions—it’s determining *whether, when, and how* systems should act autonomously on behalf of users.