Sense and Accessibility: Understanding People with Physical Disabilities' Experiences with Sensing Systems
Shaun K. Kane, Anhong Guo, Meredith Ringel Morris · 2020 · Proceedings of the 22nd International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2020) · doi:10.1145/3373625.3416990
Summary
This paper examines how sensing systems — the increasingly pervasive technologies that mediate our interactions with the digital and physical world — create accessibility barriers for people with physical disabilities. Through an online survey of 40 adults with physical disabilities (conditions including spinal cord injuries, cerebral palsy, multiple sclerosis, muscular dystrophy, spina bifida, and limb amputations; 70% wheelchair users), the researchers gathered detailed qualitative accounts of experiences with motion sensors, biometric sensors, speech input, touchscreens, and gesture systems. The study was motivated by the recognition that sensing technologies embed "ability assumptions" about users' bodies — assumptions about how fast people move, what range of motion they have, their body proportions, and their physical capabilities. As computing becomes more ubiquitous and invisible through smart devices, IoT, and AI-powered systems, these assumptions create a growing digital divide between those for whom sensors work and those for whom they do not. The survey covered 22 different sensing scenarios across home, workplace, and public settings, collecting both quantitative frequency data and rich free-text descriptions of specific experiences.
Key findings
The study identified ten key challenge categories. Premature timeouts were extremely common (72.5% experienced this), with automatic doors, elevators, ATMs, and kiosks closing or timing out before participants could complete interactions. Poor positioning affected 62.5% — sensors, buttons, and cameras placed at heights or angles inaccessible from a wheelchair, including doorbell cameras, security scanners, and thermostat sensors. Being "invisible" to sensors (55%) meant motion-activated lights, doors, and thermostats failed to detect wheelchair users entirely. Mismatched range of motion affected participants when sensors expected movements beyond their capabilities. Variable abilities was a critical finding — participants' abilities changed over time (disease progression, fatigue, medication effects) and even within a single day, yet sensing systems expected consistent normative inputs. Biometric failures (37.5%) included fingerprint scanners failing due to joint contractures, poor circulation, or inability to position fingers correctly, and facial recognition failing for wheelchair users. Incorrect inferences included fitness trackers not counting wheelchair movement as steps, speech systems misgendering users who use ventilators, and health apps rejecting atypical height-weight combinations as invalid data. Security vulnerabilities emerged when sensor failures forced participants to share passwords or rely on others. Participants responded with four strategies: seeking assistance (from family, strangers, or authorities), adaptation (exaggerated gestures, low-tech extensions like dressing sticks), avoidance (62.5% reported avoiding entire technology categories), and abandonment (giving up on technologies that consistently failed).
Relevance
This paper delivers a comprehensive and urgent message: the trend toward "smart," sensor-mediated environments is creating new and pervasive accessibility barriers that go far beyond traditional software and web accessibility. For accessibility practitioners, the findings highlight that physical environment accessibility now includes digital sensing infrastructure — automatic doors, biometric security, smart home controls, fitness trackers, and payment terminals all embed ability assumptions. The ten challenge categories provide a practical audit framework for evaluating sensing systems. The avoidance and abandonment findings (62.5% avoiding technologies) are particularly alarming, suggesting a widening digital divide. The paper also raises important tensions: between security and accessibility (biometrics), between universal design and personalization, and between independence and interdependence. Limitations include the exclusion of screen reader users (due to survey tool inaccessibility, an ironic barrier the authors acknowledge), the lack of intersectional analysis, and the potential for provided examples to prime participant responses.
Tags: physical disability · AI bias · AI fairness · sensors · ubiquitous computing · inclusive design · accessibility barriers · biometric authentication · digital divide · Internet of Things