ZoomPursuit: Smooth Pursuit-based Gaze Interaction for General-Purpose GUIs

Abstract

Gaze interaction in general-purpose GUIs is limited by its inaccuracy and calibration requirements. Smooth pursuit-based methods offer a calibration-free alternative but often demand modified user interfaces. We present ZoomPursuit, a technique that adapts smooth pursuit for unmodified GUIs by repurposing system-level magnification functionality.

A baseline evaluation (N=16) established that under simulated calibration drift, ZoomPursuit achieved the accuracy of a manual refinement method with the speed of a simpler automated zoom-based technique. However, the study also revealed a conflict between its imperceptible automation and users' sense of control. A follow-up study (N=7) resolved this with a two-stage model integrating automated correction with manual refinement. This design achieved a final pointing error of 0.10° while dynamically calibrating on the fly from an uncalibrated state.

This work details the design evolution of a pursuit-based interaction technique for general GUIs. The final design demonstrates a model that is both technically effective and preserves user agency.