Epidemiology as a Framework for Large-Scale Mobile Application Accessibility Assessment
Anne Spencer Ross, Xiaoyi Zhang, James Fogarty, Jacob O. Wobbrock · 2017 · Proceedings of the 19th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3132525.3132547
Summary
This paper introduces a novel conceptual framework for understanding mobile app accessibility by drawing an extended metaphor from epidemiology — the study of disease patterns, causes, and effects in populations. The core insight is that app accessibility is typically considered at the level of individual apps, but rarely examined at the ecosystem level encompassing app stores, developers, companies, toolkits, design patterns, and user influences. Under this framework, "inaccessibility" is treated as a set of diseases that can be analyzed through epidemiological concepts like risk factors (elements that increase the likelihood of barriers), protective factors (elements that reduce barriers), prevalence (how widespread a barrier is), incidence (new cases over time), transmission (how barriers spread through code reuse and shared toolkits), and treatments (preventative and therapeutic interventions). The researchers from the University of Washington map dozens of epidemiological terms to app accessibility equivalents across four categories: terms describing a single app (health, disease, host, infectious agent, determinant), terms describing a disease (reservoir, contagiousness, prevalence, lethality, transmission), population-level terms (census, high-risk group, outbreak, herd immunity, health indicator), and terms for taking action (public health, treatment, prevention, therapy, universal precautions). Two detailed real-world examples illustrate the framework: Android Studio's icon button widget, which generates code without content descriptions, acts as an "infectious agent" transmitting the "inaccessible button disease" to every app that uses it; and Android's Floating Action Button design pattern, which introduces navigation order and state-tracking barriers through its design guidelines — the "reservoir" of the infection. The framework also presents two "natural history" models: one for app development (from conception through design, implementation, testing, release, updates, to abandonment/death) and one for app usage (from finding and downloading through first use, encountering barriers, workarounds, to usage abandonment), identifying where preventative and therapeutic treatments can be applied.
Key findings
To exercise the framework, the researchers assessed a stratified sample of 100 popular free Android apps (10 from each of 10 Google Play Store categories) using Google's Accessibility Scanner, examining nine types of accessibility errors. The results were stark: 100% of apps had at least one accessibility error. The most prevalent determinants were Touch Target (too small, 95% of apps), Item Label (missing, 94%), and Text Contrast (insufficient, 94%). Slightly less prevalent were Item Descriptions (85%), Image Contrast (85%), and Clickable Items (overlapping, 57%). Less common were Item Type Label (redundant, 20%), Editable Item Label (10%), and Link errors (1%). The distribution of co-occurring errors was notable: 72% of apps had five or six of the nine error types, with 36% having exactly five and 36% having exactly six. This clustering — rather than a uniform distribution — suggests interaction among different errors or underlying common factors driving multiple accessibility failures simultaneously. The three apps with only a single error type each had their entire interface rendered as a single canvas element, making the whole screen inaccessible to the accessibility API while technically presenting only one "error." This highlights a significant limitation of automated scanning tools: they may undercount the severity of accessibility problems.
Relevance
This framework represents a paradigm shift in how we think about app accessibility — moving from treating it as an individual app problem to understanding it as a systemic, population-level phenomenon with identifiable vectors of transmission, risk factors, and opportunities for intervention at scale. For accessibility practitioners and organizations, the most actionable insights come from the "Chain of Infection" model: identifying that toolkits, design guidelines, and code repositories serve as "reservoirs" for accessibility barriers suggests that fixing barriers at the source (e.g., making Android Studio generate accessible code by default) would have far greater impact than fixing individual apps. The 100% prevalence rate of accessibility errors among popular apps is a powerful data point for advocacy, demonstrating that even the most downloaded, well-resourced apps fail basic accessibility checks. The framework also motivates longitudinal tracking — monitoring whether interventions like improved developer tools actually reduce the prevalence of specific barriers over time — which the accessibility field has largely not done systematically.
Tags: mobile accessibility · accessibility testing · automated testing · Android · accessibility frameworks · population-level analysis · accessibility barriers · app accessibility
Standards referenced: WCAG 2.0