ITO: Real-time Browser Tab Orchestration Through Intent Detection

Abstract

People routinely interleave activities while browsing the web, often simultaneously and with overlapping boundaries. Yet organizational primitives in modern browsers treat every tab uniformly, offering no structural awareness of which items serve which purpose. While task-based organization approaches exist, they typically require users to manually organize or invoke reorganization features, and quickly fall out of sync as user intents evolve. To address this, we present Ito, a mixed-initiative approach that infers user intent in real time and organizes browsing activities into dynamic collections called Flows. Ito continuously reads unfolding user context, determines moment-to-moment focus, and creates, restructures, hibernates, and awakens Flows in real time while preserving user control through a mixed-initiative loop of proposals and corrections. In a controlled lab study (N=12) and a two-week field study (N=11), results suggest that Ito's intent-driven Flows aligned more closely with participants' mental models, facilitated task switching and navigation more effectively, and reduced management burden without loss of perceived control.