Collaborative Web Accessibility Improvement: Challenges and Possibilities
Hironobu Takagi, Shinya Kawanaka, Masatomo Kobayashi, Daisuke Sato, Chieko Asakawa · 2009 · Proceedings of the 11th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '09) · doi:10.1145/1639642.1639677
Summary
This paper reports on the Social Accessibility Project, an experimental service developed by IBM Research in Tokyo that uses collaborative metadata authoring to improve web accessibility. The system works through a three-part architecture: end users (primarily screen reader users) report accessibility problems they encounter on websites; volunteer supporters create external metadata to fix those problems (such as alternative text for images, heading tags, and labels for form inputs); and a public repository stores this metadata so it benefits all users visiting the same pages. Over 10 months of operation starting July 2008, approximately 18,000 pieces of metadata were created for 2,930 webpages through collaboration. The system uses web transcoding technology to apply the metadata fixes without requiring any changes to the original websites. The paper describes three case studies: a hospital website where a supporter created 10 alternative texts and 121 heading tags for 22 pages within 7 hours; a radio station website where two supporters resolved the request in 20 hours; and a sightseeing information site that used an embedded scrollable map widget which proved too technically complex to fix. Two major interface improvements were introduced during the pilot: a screen-reader-independent end-user tool (supporting both JAWS and a native Windows application) and visual metadata authoring interfaces including a WYSIWYG-style page map editor and a quick fix wizard for batch-processing inaccessible images.
Key findings
Supporter productivity far exceeded expectations — 45% of requests were resolved within one day (24 hours) and 50% by 34 hours, dramatically faster than traditional source-code modification approaches which typically require weeks. The most common metadata type was alternative text (11,969 items, 65%), followed by heading tags (6,069 items, 33%). Of 323 user requests, 85.1% were resolved and only 5.9% remained unanswered. The most striking and concerning finding was that screen reader users were often unaware of the nature of the accessibility problems they faced. About 30 blind computer users in a seminar discussion reported that most had tried general websites but found them "totally confusing and time consuming" and gave up, accessing only selected sites for the blind. Users could report that a page was difficult but often could not articulate what specifically was wrong — they didn't know what they didn't know. Another significant awareness gap involved Flash content: around 50% of Flash content is hidden from screen readers, yet users were only aware of about half of Flash-related problems. The page-map visual authoring tool dramatically lowered the skill barrier for supporters, and supporter activity increased sharply after its release.
Relevance
This research pioneers a fundamentally different model for addressing web accessibility — rather than waiting for website owners to fix problems, it empowers a community of volunteers to create external fixes that are applied in real time through transcoding. The approach is especially relevant given the persistent gap between accessibility guidelines and actual website compliance. The finding about users' lack of awareness of their own accessibility barriers is profoundly important: it suggests that the true scale of web accessibility problems is significantly underreported because the people most affected cannot fully perceive what they are missing. This has implications for how we conduct user research, interpret complaint-driven accessibility audits, and design screen reader interfaces that might better surface hidden content or inaccessible elements. The paper also raises important questions about sustainability — the collaborative model depends on a motivated volunteer community, and the authors discuss strategies including partnerships with nonprofit organisations and integration with site-wide authoring tools. The concept of external accessibility metadata that can be applied without website owner cooperation remains highly relevant to modern browser extension-based accessibility tools.
Tags: web accessibility · collaborative accessibility · crowdsourcing · metadata · screen readers · social computing · web transcoding · alternative text
Standards referenced: WCAG