The Design of Human-Powered Access Technology
Jeffrey P. Bigham, Richard E. Ladner, Yevgen Borodin · 2011 · Proceedings of the 13th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS 2011) · doi:10.1145/2049536.2049540
Summary
This paper presents a comprehensive framework of 13 design dimensions for evaluating and comparing human-powered access technology — systems that use human assistance, facilitated by technology, to overcome accessibility barriers that automation alone cannot solve. The authors trace the historical roots of such technology from informal community assistance (blind people finding readers, deaf people recruiting volunteer interpreters) through formalized services (Video Relay Services, CART captioning, reading services) to modern crowdsourcing platforms (VizWiz, Mechanical Turk-based systems). They catalog 15 diverse examples of human-powered access technology spanning the spectrum from asynchronous community contributions (ASL-STEM Forum, Bookshare) to real-time interactive services (VRS, remote captioning, VizWiz). From these examples, the authors distill 13 design dimensions organized around key questions any human-powered access system must address: who initiates the assistance (Initiative), where workers come from (Source), what drives participation (Motivation), how it is funded (Financing), what skills workers need (Competence), how fast responses arrive (Latency), how quality is ensured (Accuracy), how dependable the service is (Reliability), what users know about the assistance process (Provenance), how privacy is protected (Confidentiality/Privacy/Anonymity), who else is affected (Broader Context), whether work products can be shared (Broader Applicability), and what disability is served (Target Disability).
Key findings
The framework reveals several important patterns and gaps across existing human-powered access technologies. The comparative analysis of all 15 systems shows that most target sensory disabilities (blindness and deafness), with very few addressing cognitive or motor disabilities — identifying a significant gap for future research. The authors find that different worker sources (professionals, crowd workers, volunteers, organized volunteers) each carry distinct trade-offs in cost, reliability, competence, and confidentiality that cannot be optimized simultaneously. A key insight is that most systems focus exclusively on the end user's needs while ignoring broader context — workers who are asked to read medication labels may face ethical dilemmas, bystanders may appear unwittingly in photos sent to crowds, and the implications of human assistance on third parties are rarely considered. The paper also identifies a gap in sustainability models: some services rely on government funding (VRS), others on research grants (VizWiz, GoBraille), and others on volunteer goodwill (Bookshare, Social Accessibility), but no single model works for all contexts. The analysis demonstrates that systems like VizWiz and Solona may appear similar (both describe images for blind people) but differ fundamentally in worker source, latency, reliability, and confidentiality guarantees.
Relevance
This framework paper remains one of the most cited and influential works in human-powered accessibility, providing vocabulary and structure for discussing a rapidly evolving field. The 13 design dimensions continue to be relevant for evaluating modern accessibility services — including AI-powered tools like Be My Eyes with GPT-4, which can be analyzed along the same dimensions (initiative, latency, accuracy, privacy) but with "artificial intelligence" as a new worker source alongside human options. For accessibility practitioners, the framework is valuable for making informed decisions about which human-powered services to deploy: understanding the trade-offs between professional captionists (high accuracy, high cost, low availability) and crowd captioning (lower cost, variable accuracy, higher availability) requires exactly the kind of structured comparison this paper enables. The identification of broader context as an under-explored dimension was prescient — questions about worker wellbeing, bystander privacy, and the ethics of crowdsourced accessibility work have become increasingly important as these systems scale.
Tags: human computation · crowdsourcing · assistive technology · accessibility framework · design principles