Social accessibility: the challenge of improving web accessibility through collaboration
Daisuke Sato, Masatomo Kobayashi, Hironobu Takagi, Chieko Asakawa · 2010 · Proceedings of the 2010 International Cross Disciplinary Conference on Web Accessibility (W4A) · doi:10.1145/1805986.1806024
Summary
This paper presents Social Accessibility, a collaborative project from IBM Research Tokyo that enlists volunteers to fix web accessibility problems reported by visually impaired users. The system works through a three-step process: blind users encounter and report accessibility barriers on web pages, volunteers then author external metadata to fix those problems (such as adding missing alternative text for images), and the fixes are applied to the page without requiring the site owner to modify their content. The project builds on earlier collaborative approaches like the ALT-server (1997), which pioneered collaborative image labeling, and WebInsight, which combined crowdsourcing with automatic OCR for image descriptions. After operating for over eighteen months as a pilot, the project produced surprising findings about the dynamics of collaborative accessibility work. The authors argue that the scale of web accessibility problems — affecting billions of people including those with disabilities, elderly users, and people with low literacy — is too vast to be addressed by web developers alone, making crowd-based approaches a necessary complement to traditional developer-side solutions.
Key findings
The most significant finding was counterintuitive: volunteers were far more active and effective than expected, with almost half of reported problems resolved within 24 hours, but the target users (blind people) participated less actively than anticipated. The primary barrier was problem awareness — many blind users were not fully aware of the accessibility issues they were encountering. Seminar discussions with about 30 blind computer users revealed that screen readers silently skip inaccessible objects like unlinked images without alt text or opaque Flash content, leaving users completely unaware these elements exist. Additionally, accessibility problems tend to repeat across pages with similar layouts, making users uncertain about when to submit requests. To address this, the team developed accessibility checking functions to help users discover gaps between visual and non-visual content. Tool usability proved critical for volunteer retention — after releasing an easier metadata authoring tool, volunteer activity increased dramatically and new participants were more likely to continue contributing. A site-wide metadata feature allowed one volunteer to fix 3,500 validation errors across 82 pages of a government website in just one hour by creating 142 metadata records.
Relevance
This research pioneered the crowdsourced accessibility model that has since influenced projects like Be My Eyes and other collaborative assistive platforms. The core insight — that blind users often cannot identify accessibility problems because screen readers silently skip inaccessible content — has profound implications for how we think about user testing and accessibility feedback loops. If users cannot perceive the barriers, relying solely on user-reported issues will always underestimate the true scope of problems. The external metadata approach offers a pragmatic middle ground between waiting for site owners to fix their pages and accepting inaccessible content. For organizations today, the finding that tool usability directly drives volunteer participation rates provides a clear lesson: accessibility remediation tools must be as frictionless as possible to sustain community engagement. The project also demonstrated that collaborative accessibility can extend beyond visual impairment to address captioning for deaf users, audio description, and reading difficulties — a vision that has increasingly materialized through modern crowdsourcing platforms.
Tags: crowdsourcing · screen readers · alternative text · visual impairment · social computing · collaborative accessibility · external metadata