CAN: Composable Accessibility Infrastructure via Data-Driven Crowdsourcing
Yun Huang, Brian Dobreski, Bijay Bhaskar Deo, Jiahang Xin, Natã Miccael Barbosa, Yang Wang, Jeffrey P. Bigham · 2015 · Proceedings of the 12th International Web for All Conference (W4A) · doi:10.1145/2745555.2746651
Summary
This paper presents CAN (Composable Accessibility Infrastructure), a crowdsourcing platform that collects web accessibility issues from real websites and their open-source fixes, dynamically composes solutions on-the-fly, and delivers the crowdsourced content as both immediate accessibility fixes for end users and teaching materials for web developers. CAN serves three user groups: end users with disabilities who install a browser plugin that automatically detects and fixes accessibility issues during normal browsing (also implicitly contributing issue data); open-source developers who contribute modular fix components called Accessible Components (ACs); and web developers who use a separate developer plugin to diagnose accessibility issues and learn from real-world examples and their solutions. The system architecture includes a Diagnostic Engine that leverages existing tools like AChecker and WAVE while also handling complex issues those tools miss (e.g., inaccessible dropdown menus), an Issue Repository that collects crowdsourced accessibility problems, an AC Container storing contributed fixes, a Validation component for quality-checking contributed code, and an AC Composition engine that assembles the right combination of fixes for each webpage. The platform was built on Node.js with ACs written in JavaScript.
Key findings
Testing CAN against the top 100 popular websites (as of November 2014) revealed that 93 of them needed at least one AC, with 19 needing all four ACs developed. Specific findings: 35 sites had missing alternative text for images, 32 had color contrast issues, 22 had dropdown menus but only 2 were keyboard accessible, and most sites had login forms that were difficult for screen reader users to reach. The Alternative Image Text AC (AC-1) fixed issues on 25 of the 35 affected sites. The Color Contrast AC (AC-2) fixed 30 of 32 sites with contrast problems. The Dropdown Menu ACs demonstrated that the same symptom (keyboard-inaccessible menus) could stem from different implementations requiring different fixes. A Login First AC successfully prioritized login forms on 73 of 89 testable websites by cloning the form to appear first in the DOM. A user study with 9 web developers of varying experience found that none could immediately fix the dropdown menu accessibility problem, but all found CAN an effective learning platform. After using CAN, participants reported increased awareness of and concern about accessibility. The study highlighted that many developers learn through self-directed methods, and CAN's real-world examples provided more engaging material than abstract guidelines.
Relevance
CAN addresses a persistent gap in web accessibility: developers often lack both awareness and practical knowledge of how to fix accessibility issues. Rather than relying solely on guidelines or automated checkers that identify problems without solving them, CAN provides actual working fixes alongside educational context. The crowdsourcing model is particularly valuable because it distributes the burden of accessibility work — end users contribute issue data passively, open-source developers contribute fixes, and web developers learn from real examples. The finding that 93 of the top 100 websites had accessibility issues underscores how pervasive these problems remain even on high-traffic sites. For accessibility practitioners, CAN demonstrates that the same user-facing problem (e.g., an inaccessible dropdown menu) can have multiple underlying implementations requiring different solutions, which is why automated tools alone cannot solve all accessibility issues. The platform's educational component is especially relevant given research showing that developer training directly impacts website accessibility outcomes.
Tags: crowdsourcing · web accessibility · accessibility education · automated testing · screen readers · accessibility remediation · developer tools · WCAG
Standards referenced: WCAG 2.0 · WAI-ARIA · UAAG 2.0