Insights on Assistive Orientation and Mobility of People with Visual Impairment Based on Large-Scale Longitudinal Data
Hernisa Kacorri, Sergio Mascetti, Andrea Gerino, Dragan Ahmetovic, Valeria Alampi, Hironobu Takagi, Chieko Asakawa · 2018 · ACM Transactions on Accessible Computing (TACCESS) · doi:10.1145/3178853
Summary
This paper analyses large-scale longitudinal usage data from iMove, a mobile app supporting outdoor orientation for people with visual impairment (PVI), examining millions of interactions from approximately 15,000 users worldwide over 15 months. This is a departure from the typical small-sample iterative studies in assistive technology research — after filtering out "incidental" users (those who downloaded the app but showed no sustained engagement), the analysis covers 1.5 million log records from a geographically diverse population. iMove provides location notifications (current address, speed, heading), nearby points of interest (POI) information, and route-based audio and text geonotes, all accessible through VoiceOver. The study proceeds along three analytical dimensions: exploratory analysis of app functions, settings, and accessibility features; inferential analysis of movement modalities and notification patterns; and unsupervised clustering to discover distinct user communities. A key unexpected finding was the high presence of users with residual sight — at least three-quarters of the most active users interacted with visual interface elements (enlarged text was the second most activated accessibility feature after VoiceOver), indicating that iMove served a much broader spectrum of visual impairment than the fully blind users it was primarily designed for.
Key findings
Three distinct user communities emerged from k-means clustering of approximately 1,400 active users. C1 users ("POI explorers") primarily used the app to discover surrounding points of interest, often while stationary — they were interested in knowing what was around them rather than navigating to a destination. C2 users ("location checkers") interacted in short bursts to inquire about their current location — brief sessions centered on the question "where am I?" C3 users ("active navigators") had long active sessions while in motion, the use case iMove was originally designed for. Only C3 matched the designers' intended user model; C1 and C2 represented unanticipated but meaningful use patterns that together accounted for approximately 75% of active users. The most popular interaction mode across all groups was passive — users preferred to receive automatic notifications while in motion rather than actively querying the app, likely because their hands and attention are occupied by canes or guide dogs while walking. Users showed high variability in preferred notification proximity and frequency settings, reinforcing the need for early customization options. Notification verbosity was preferred high by default, with users wanting all available information (address, speed, heading, city) rather than minimal output.
Relevance
This research demonstrates the value of large-scale observational data for understanding real-world assistive technology use, overcoming limitations of small lab studies that may not capture the diversity of actual usage patterns. For developers of navigation and orientation apps for PVI, the three user communities provide actionable design personas: apps designed only for active turn-by-turn navigation miss the majority of users who want environmental awareness or quick location checks. The finding that most users prefer passive interaction while in motion has clear design implications — hands-free, automatic notifications should be the default rather than requiring active input. The high proportion of users with residual sight challenges the binary blind/sighted design paradigm and suggests that visual interface elements (maps, enlarged text) remain valuable even in apps designed for visual impairment. The publicly released dataset enables other researchers to build on these findings without needing to deploy their own large-scale app, contributing to the accessibility dataset ecosystem alongside resources like WeAllWalk.
Tags: orientation and mobility · visual impairment · large-scale data · longitudinal study · user clustering · mobile navigation · points of interest · usage analytics · residual vision