Crowdsourcing Platform for Workplace Accessibility
Hironobu Takagi, Akihiro Kosugi, Shin Saito, Masayoshi Teraguchi · 2013 · Proceedings of the 10th International Cross-Disciplinary Conference on Web Accessibility (W4A) · doi:10.1145/2461121.2461135
Summary
This paper from IBM Research Tokyo proposes Crowd Card, an intra-organizational crowdsourcing platform designed to improve workplace accessibility while maintaining confidentiality of business materials. The core problem is that modern workplaces are filled with inaccessible content — uncaptioned meeting videos, undescribed diagrams in presentations, scanned documents without text alternatives — that creates barriers for employees with disabilities. While public crowdsourcing services like Amazon Mechanical Turk could generate captions and descriptions, confidential business materials cannot be exposed to external crowds. The solution uses employees as crowd workers, but addresses the challenge that employees have limited spare time by borrowing strategies from web advertising: microtask "cards" resembling web banners are embedded unobtrusively into intranet portals, web pages, social network feeds, and email client sidebars. The system crawls internal services (Q&A forums, captioning systems, document digitisation tools, presentation converters) to collect accessibility requests, converts them into uniform card formats through a process called "cardification," and distributes them to appropriate employees using recommendation algorithms.
Key findings
The authors report prior experience deploying crowdsourcing services within a company of over 100,000 employees, recruiting approximately 100 volunteers each for OCR error correction (86 workers) and video captioning (116 workers) — and actually struggling with a lack of tasks rather than a lack of workers. Corporate crowd workers proved highly motivated and produced high-quality work, partly because their lack of anonymity within the company discouraged malicious or low-effort submissions. The Crowd Card system implements four types of accessibility microtask cards: image/diagram description requests (where employees type a description directly in the card), general Q&A requests gathered from internal forums, video captioning call-for-action cards linking to the CCES collaborative captioning system, and OCR error correction cards for Japanese document digitisation (EBIS). Task distribution uses three algorithms: community-based access control (respecting existing confidentiality boundaries of internal social networks), context-based distribution (matching cards to the content an employee is currently viewing using k-nearest-neighbour), and interest/skill-based distribution using collaborative filtering. The prototype was deployed internally in 2012.
Relevance
This research addresses a critical but often overlooked aspect of digital accessibility: workplace information access for employees with disabilities. While most accessibility research focuses on public-facing websites, employees with disabilities face daily barriers from inaccessible internal content — presentation slides with undescribed charts, uncaptioned training videos, scanned documents without text equivalents — that impede their ability to perform their jobs and advance their careers. The Crowd Card approach offers a practical model for organisations: rather than creating dedicated accessibility remediation teams, distribute small tasks across the existing workforce by embedding them where employees already spend their time. The finding that enterprise workers are more motivated and reliable than public crowds suggests that internal crowdsourcing may actually be more effective than external services for accessibility tasks requiring domain expertise. For organisations today, the underlying principles remain highly relevant even as AI has advanced: combining automated recognition with human review through lightweight, contextually embedded microtasks can efficiently close accessibility gaps in workplace content while maintaining confidentiality.
Tags: crowdsourcing · workplace accessibility · image description · video captioning · document accessibility · assistive technology · intranet · confidentiality