Improving Low-Vision Chart Accessibility via On-Cursor Visual Context
Yotam Sechayk, Hennes Rave, Max Rädler, Mark Colley, Zhongyi Zhou, Ariel Shamir, Takeo Igarashi · 2026 · Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI '26) · doi:10.1145/3772318.3791165
Summary
Charts remain largely inaccessible to Low-Vision Individuals (LVI) because reading them requires holding both fine-grained data details and the broader visual context — axes, legend, grid lines, overview — in view at the same time, a task that magnification and restricted fields of vision make exhausting. The authors first ran a formative study with five LVI using the Mini-VLAT chart literacy test across twelve common chart types, observing how participants used residual vision together with screen magnifiers, browser zoom, color filters, and dark-mode themes. From that study they identified four contextual elements LVI rely on — axes, legend, grid lines, and the overall overview — and derived four design goals: make context accessible, reduce navigational effort, support diverse strategies, and adapt to fluctuating user needs. They then built two pointer-based prototypes intended to sit on top of existing assistive tools rather than replace them: Dynamic Context, a novel focus+context method that projects the axes, legend, and a crosshair grid-line indicator into a compact Overlay Area around the cursor; and Mini-map, an overview+detail adaptation that shows a minified chart with a position indicator. Both prototypes expose personalization settings (overlay size, border, outer dimming, axis ratio, crosshair ratio, mini-map scale) following a WYSIWYG pattern. The paper is written for practitioners as well as academics and frames visual access as a legitimate, often preferred, path for LVI alongside non-visual modalities.
Key findings
A within-subjects evaluation with 22 LVI compared baseline, Mini-map, and Dynamic Context using Mini-VLAT variants, pointer-trajectory analysis, NASA-TLX, SUS, and self-reported access. Accuracy and completion times did not differ significantly across conditions — a mixed quantitative result the authors attribute to the highly varied workflows participants brought with their AT. Subjective outcomes told a different story. SUS usability rose from 52.4 at baseline to 76.4 with Dynamic Context; NASA-TLX Effort and Physical Demand dropped significantly; and participants rated Dynamic Context significantly higher on sense of access to chart data, focus on data, personal fit, and likelihood to recommend. PERMANOVA with Dynamic Time Warping on pointer trajectories showed Dynamic Context significantly altered interaction patterns on stacked-area, stacked-bar, histogram, and area charts, concentrating activity near the chart center and reducing panning to the axes and legend. The projected crosshair acted as a stabilizing visual anchor that several participants said reduced drift and guesswork. Mini-map strengthened orientation but was judged too small to answer questions and was less preferred overall. Both methods clashed with full-screen color filters, high-contrast themes, and pinch-to-zoom, and Dynamic Context's compact layout hurt participants with double vision or very high magnification.
Relevance
This work matters for practitioners because it pushes back on the default assumption that LVI are best served by non-visual access such as screen readers or sonification. It shows that preserving visual agency — bringing context close to the pointer rather than forcing constant panning and zooming — can meaningfully reduce effort and improve perceived access even when raw accuracy gains are modest. For product teams building data-heavy interfaces, the three design insights (maximize visual access to context, reduce visual complexity through abstraction, adapt to fluctuating user needs) translate directly to decisions about sticky legends, tooltips, and overview widgets. The paper also surfaces a fragile integration layer: overlays must coexist with magnifiers, color filters, and themes, or they break — pointing toward ability-based design systems that auto-adapt to AT in use. Limits include a small sample, reliance on manual chart annotation, and no direct comparison with prior LVI-specific tools.
Tags: low vision · data visualization · chart accessibility · assistive technology · screen magnification · focus+context · overview+detail · user study · pointer-based interaction · visual access
Standards referenced: WCAG 2.1 · ARIA