Understanding and Reducing the Challenges Faced by Creators of Accessible Online Data Visualizations
Ather Sharif, Joo Gyeong Kim, Jessie Zijia Xu, Jacob O. Wobbrock · 2024 · Proceedings of the 26th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '24) · doi:10.1145/3663548.3675625
Summary
This paper investigates the challenges visualization creators face in making online data visualizations accessible to screen-reader users and develops interventions to address these challenges. Data visualizations (charts, graphs, maps, dashboards) are increasingly prevalent on the web but are overwhelmingly visual, making them largely inaccessible to blind and low-vision users who rely on screen readers. The researchers conducted a three-phase study: first, a formative survey of 57 visualization creators to understand their challenges, knowledge gaps, and prioritisation of accessibility; second, semi-structured interviews with 12 creators to develop specifications for five proposed interventions; and third, a task-based user study with 10 creators evaluating enhanced versions of these interventions integrated into VoxLens, an open-source JavaScript library the authors previously developed. VoxLens makes data visualizations accessible through three modalities: a natural language summary of the data, sonification (mapping data values to audio tones so users can hear trends and patterns), and a voice-based question-and-answer interface where users can query the data conversationally. The five interventions identified were: Workshops (educational modules on visualization accessibility), Emulators (tools that simulate how screen readers interact with visualizations), Evaluators (automated accessibility checkers for visualizations), Feedback Collectors (tools for gathering accessibility feedback from screen-reader users), and Multi-Modal Automated Tools (libraries like VoxLens that automatically add non-visual access).
Key findings
The survey revealed that while 84% of creators considered visualization accessibility important, only 30% had formal knowledge of how to implement it, and only 19% regularly tested their visualizations with screen readers. The most commonly cited challenges were lack of knowledge about screen-reader interaction patterns, difficulty understanding what information blind users need from a visualization, uncertainty about ARIA roles and properties for SVG-based charts, and the absence of automated tools comparable to WAVE or axe for visualization-specific accessibility. In the interviews, creators expressed frustration that existing accessibility guidelines (including WCAG) provide insufficient guidance specific to data visualizations — they address general web content but not the unique challenges of conveying data relationships, trends, and patterns non-visually. The enhanced VoxLens with all five interventions produced significant improvements in creator outcomes: a 44% improvement in understanding screen-reader users' challenges, 17% improvement in accessibility knowledge, and 12% improvement in perceived usefulness of the tool. The Emulator was particularly impactful — creators reported that experiencing their own visualizations through a simulated screen reader was transformative for understanding the barriers. The Feedback Collector used a "mini-survey" format that proved effective for gathering structured accessibility feedback from screen-reader users without imposing excessive burden. Creators also valued the sonification modality, noting that audio representation of data trends was sometimes more immediately intuitive than textual descriptions for conveying patterns like upward trends or outliers.
Relevance
This paper is directly relevant to web developers and data analysts who publish visualizations online. Data visualizations represent one of the largest remaining gaps in web accessibility — most charting libraries (D3.js, Chart.js, Highcharts) produce visually rich but semantically opaque SVG or canvas elements that screen readers cannot interpret. The five interventions provide a practical roadmap for organisations wanting to improve visualization accessibility: educate creators, let them experience screen-reader interaction firsthand, automate what can be automated, and collect feedback from actual users. VoxLens itself is open-source and can be integrated into existing web applications with minimal code changes, making it an immediately actionable tool. The sonification approach is particularly noteworthy — it demonstrates that some data relationships are actually conveyed more efficiently through audio than through text descriptions, suggesting that accessible alternatives can sometimes be superior to visual-only presentation. Limitations include the relatively small user study sample and the focus on common chart types (bar, line, scatter) rather than more complex visualizations like networks or geographic maps, but the framework and interventions are extensible.
Tags: data visualization · screen readers · web accessibility · sonification · blind and low vision · developer tools · ARIA
Standards referenced: WCAG 2.1 · ARIA