COVID-19 Highlights the Issues Facing Blind and Visually Impaired People in Accessing Data on the Web
Alexa F. Siu, Danyang Fan, Gene S-H Kim, Hrishikesh V. Rao, Xavier Vazquez, Sile O'Modhrain, Sean Follmer · 2021 · Proceedings of the 18th International Web for All Conference (W4A) · doi:10.1145/3430263.3452432
Summary
This mixed-methods study examines how blind and visually impaired (BVI) people experienced accessing data-driven information on the web during the COVID-19 pandemic — a period when data visualizations such as infection curves, maps of hotspots, and trend charts became central to public health communication and personal decision-making. The researchers conducted an online survey (n=127) between June and July 2020 followed by contextual inquiry sessions (n=12) between August and September 2020. Survey participants ranged in age from 18 to 84 (mean 42), with 42.5% describing themselves as totally blind, 31.5% as legally blind, and the remainder having some shape, light perception, or very low vision. All participants used screen readers as their primary assistive technology. The contextual inquiry used three accessible COVID-19 websites that employed different data representation modalities — CVStats.net (tabular data), covid.ski (sonification), and a researcher-created site following NCAM guidelines for data graphic descriptions (alt text with tables). Participants explored these sites using a think-aloud protocol, making decisions about pandemic severity across different US states. The study provides both quantitative evidence of the accessibility gap from the survey and rich qualitative observations of how BVI users actually interact with, interpret, and draw insights from different accessible data representations.
Key findings
The survey revealed that 94% of respondents (117/125) had concerns about accessing accurate COVID-19 data in a timely manner, and 73% disagreed that data-driven media they encountered was typically accessible with their assistive technology. Respondents ranked local trends and global trends as the most commonly inaccessible types of COVID-19 information. At the early onset of the pandemic (March-April 2020), participants described encountering broad inaccessibility across data sources, with accessibility improving over time partly due to community advocacy efforts. BVI users employed diverse strategies to overcome barriers: 28.8% sought out accessibility-branded COVID-19 dashboards, 25% learned sonification tools, 21.2% relied on podcasts and news, and 15.4% depended on sighted relatives or live visual interpretation services like AIRA. In the contextual inquiry, the researchers found that no single data representation worked best for all tasks or all users — tables made value retrieval easier but did not facilitate understanding of broader trends, sonification conveyed trends but gaps in knowledge led to misinterpretation, and text descriptions painted a mental picture but depended on the quality of human authoring. Almost all participants (P3-P12) leveraged insights from multiple representations to complement each other and fill information gaps. Data literacy and prior experience significantly affected how participants interpreted representations, particularly sonification where users without prior knowledge drew false comparisons. Confidence was a major factor: participants were less confident when relying on others' subjective interpretations (such as alt text authored by someone else) and more confident when they could access raw data directly.
Relevance
This paper powerfully demonstrates how the inaccessibility of data visualizations creates an information equity gap that has real consequences — during a pandemic, inability to access local infection trends and testing locations directly affects personal safety decisions. For web developers and data publishers, the key takeaway is that providing multiple complementary representations (tables, descriptions, sonification) is more effective than any single accessible format, as different modalities serve different data tasks and user preferences. The finding that alt text and descriptions are only as good as their human authors — and that users are aware of and skeptical about this subjectivity — underscores the need for standardized description templates and direct access to underlying data tables. The study also highlights systemic issues: accessibility-branded community sites filled gaps that mainstream government and news sites failed to address, and BVI users had to actively advocate to get state health department sites made accessible. For organizations publishing public data, this research argues that accessible data representation must be built into the publishing workflow from the start, not retrofitted after advocacy pressure during a crisis.
Tags: data visualization · blind and low vision · web accessibility · COVID-19 · screen readers · sonification · data literacy · accessible data · information access
Standards referenced: WCAG 2.1 · Section 508