Distinguishing Users By Pointing Performance in Laboratory and Real-World Tasks

Insights on Assistive Orientation and Mobility of People with Visual Impairment Based on Large-Scale Longitudinal Data

Assistive applications for orientation and mobility promote independence for people with visual impairment (PVI). While typical design and evaluation of such applications involves small-sample iterative studies, we analyze large-scale longitudinal data from a geographically diverse population. Our publicly released dataset from i Move, a mobile app supporting orientation of PVI, contains millions of interactions by thousands of users over a year.

Our analysis (i) examines common functionalities, settings, assistive features, and movement modalities in i Move dataset and (ii) discovers user communities based on interaction patterns. We find that the most popular interaction mode is passive, where users receive more notifications, often verbose, while in motion and perform fewer actions. The use of built-in assistive features such as enlarged text indicate a high presence of users with residual sight. Users fall into three distinct groups: (C1) users interested in surrounding points of interest, (C2) users interacting in short bursts to inquire about current location, and (C3) users with long active sessions while in motion. i Move was designed with C3 in mind, and one strength of our contribution is providing meaningful semantics for unanticipated groups, C1 and C2. Our analysis reveals insights that can be generalized to other assistive orientation and mobility applications.