A Large Scale, Multi Factor Approach to Understanding and Improving Mobile Application Accessibility
Loading...
Date
Authors
Ross, Anne Spencer
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Accessibility failures in mobile applications (apps) create barriers for disabled people and people who use assistive technologies. Given the growing role of apps in everyone’s daily life, equitable access is imperative. Toward this goal, I created a conceptual framework for understanding and improving app accessibility at scale inspired by epidemiology. My epidemiology‑inspired framework poses app accessibility failures as “diseases” in a “population” of apps. This perspective forefronts a population-level perspective within an ecosystem of factors that impact app accessibility (e.g., developer tools, guidelines, company culture, and many more). In this dissertation, I demonstrate my thesis that applying my epidemiology‑inspired framework, which emphasizes large‑scale and multi‑factor approaches, (1) can reveal population‑level trends of accessibility failures, (2) can aid in identifying a range of factors that impact app accessibility, and (3) can inform the design of tools for identifying and repairing accessibility failures in apps. To enhance our understanding of the state of app inaccessibility, I performed the first large‑scale analyses of Android app accessibility. My results measured the prevalence of accessibility failures across apps and identified classes of elements that frequently had accessibility failures. Missing labels was one of the most prevalent failures; 23% of the 8,901 apps had more than 90% of their image-based elements missing labels. Reflecting a less frequent but severe accessibility failure, 8% of 9,999 tested apps were completely unusable with many assistive technologies, such as screen readers. Apps with such failures disproportionately came from the Education category. Using a multi‑factor assessment, partially guided by my large‑scale analyses, I identified accessibility shortcomings in environmental factors such as programming tools, developer guidelines, and inter-team dynamics that could contribute to app inaccessibility at scale. For example, I identified a set of game engines and cross-platform tools (e.g., Unity, Adobe Air) that frequently produced apps that were unusable with many assistive technologies. Another factor I assessed was Android’s implementation documentation, finding many instances of missing labels in the example code snippets. Toward improving app accessibility, I explored techniques for enhancing developer tools, testing tools, and third‑party repairs. In developer tools, I present novel designs and techniques that aim to improve the efficiency, effectiveness, and education of developers by tightening the runtime-implementation feedback loop, leveraging screen context to provide more specific repairs, and using new visualization and interaction tool interface designs. Within testing tools, I prototyped an extension to Google’s Accessibility Scanner to allow human annotation of automated results. Professional accessibility testers found the tool promising and discussed factors beyond tools, such as knowledge gaps and social dynamics with the developer teams, which affected their testing. Addressing repairs after an app is released, I present a proof-of-concept for third-party repair that people who use screen readers found useful while highlighting factors around trust, repair-production infrastructure, and co-creation that would impact real-word deployment. Throughout my dissertation, I leverage my epidemiology‑inspired concepts and language to highlight the value of my research and place it within the larger space of work on app accessibility. Together, this research aims to expand our understanding of app accessibility at scale and inform efforts to improve it.
Description
Thesis (Ph.D.)--University of Washington, 2021
Keywords
Accessible computing, Human-computer interaction, Large-Scale Analysis, Mobile application, Tools, Computer science, Design, Disability studies