Designing for Individuals: Usable Touch-Screen Interaction through Shared User Models
Kyle Montague, Vicki L. Hanson, Andy Cobley · 2012 · Proceedings of the 14th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/2384916.2384943
Summary
This paper introduces the Shared User Modelling (SUM) Framework, a system that collects touch interaction data during normal application use to build a user model that can be shared across multiple applications and devices via web services. Unlike traditional accessibility settings that require users to manually configure each application, SUM automatically adapts interfaces based on observed abilities — specifically touch accuracy, precision, and duration patterns. The framework was evaluated with 12 adults (ages 21-71) with diverse visual and motor impairments including tremors, muscle spasms, spinal injury, macular degeneration, dystonia, ataxia, and hypermobility syndrome. Three iPod Touch apps were developed: Target Practice (a calibration game generating baseline ability data by having users tap green targets), Indoor Navigation (wayfinding instructions with adapted text and target sizes), and TV Guide (browsing TV listings with adapted grids and programme details). Each participant used all three apps across two sessions, with half receiving adaptive interfaces informed by the SUM model and half receiving static default interfaces. The adaptive interface adjusted target sizes based on each user's precision (calculated as mean distance from target centroid), set minimum and maximum touch duration bounds to filter involuntary taps, and scaled text based on Snellen eye test results.
Key findings
The SUM adaptive interfaces produced significantly fewer touch errors than static interfaces in the target practice task (mean 18.83 errors vs. 27.67, t(11)=1.977, p<.05, d=.632). Touch error heat maps showed the adaptive condition had uniformly lower error rates across all screen regions. An important touch pattern finding was that participants consistently touched to the right of targets (68.3% of horizontal touches landed right of center), likely because they held the device in their left hand and tapped with their right, positioning their interaction closer to the right side. Vertically, touches were roughly evenly split above and below targets. Three participants with tremors experienced frequent unintentional taps on the highly sensitive capacitive screen — one participant abandoned her iPod Touch entirely in favor of an LG phone with "much lower sensitivity." The SUM Framework addressed this by setting touch duration bounds to filter involuntary inputs. Free scrolling was problematic for several participants: it moved too fast to read content, and tightening grip during swiping caused unintentional touches on the bezel-less device. SUS usability scores showed a small, non-significant improvement for adaptive interfaces (3.33 vs. 3.10), likely because limited exposure time prevented participants from fully experiencing the benefits.
Relevance
This research addresses a fundamental problem in mobile accessibility: the burden of configuring accessibility settings falls on users who may be least equipped to navigate complex preference menus, and settings must be reconfigured for each new application. The SUM Framework's approach of inferring abilities from actual interaction data and sharing adaptations across apps represents a significant improvement over manual configuration. For accessibility practitioners and mobile developers, several findings are directly applicable: touch targets should account for the systematic rightward bias observed in one-handed right-hand use; minimum touch duration thresholds can filter involuntary inputs from users with tremors; capacitive screen sensitivity (near-zero pressure detection) is a barrier, not a feature, for users with motor impairments; and shrinking device bezels eliminate the non-interactive grip area that users with disabilities rely on. The cross-application sharing model anticipates modern platform-level accessibility APIs while going further by adapting based on observed ability rather than stated preference.
Tags: touch screen accessibility · adaptive interface · user model · mobile accessibility · visual impairment · motor impairment · personalization