Seeing UI: Improving Accessibility through Open-Source Data
Individuals who are blind or visually impaired face significant challenges in navigating the over 98.1% of the top 1 million home pages designed without adequate accessibility (WebAIM, Center for Persons with Disabilities, Utah State University, 2020). We, the 2020-2021 Microsoft SCOPE team, conducted extensive research into web accessibility by screen-reader users, specifically those who are fully blind, in order to identify those issues whose solutions would have the biggest impact on their lives. Through interviews and a survey with blind screen-reader users and accessibility experts, we have learned that issues that fully block progress are higher priority than more common issues that are slow, but do not fully derail usability. We have also determined that online shopping and banking sites are the highest priority website types for blind web-users, because they impact an individual’s essential needs such as procuring food and confidentially managing finances. We have published two whitepapers synthesizing our research: one introduces ten urgent, actionable interaction issues and system-level challenges, and the other summarizes our human-centered design insights. Furthermore, we have also produced functional specifications for three specific machine learning projects in distinct, high-impact areas: disorienting pop-up focus, hidden error handling messages, and absence of task-specific headings. These specifications include methods for open source dataset creation, to be pursued by a collaborating team at the University of Texas at Austin and by the wider research community.
Lynn Andrea Stein