A reader magnifies a chart, then asks about a point. ChartAEye sends its AI two images: the complete chart, marked with the visible area and selected point, and a crop matching the reader’s view.[1]
A 2024 study of 11 low-vision adults helps explain why position matters. Participants sometimes memorised legends to avoid repeatedly panning between labels and marks. Others traced a mark towards an axis with the cursor. Enlarging a detail could make it readable while moving its reference information out of sight.[2]
In an ASSETS 2026 study, twelve adults tried two assistant versions in counterbalanced order, using two charts during one session. Both received an overview and crop; one also had instructions to explain spatial relationships and visual features. Eight participants preferred that tailored guidance, but questionnaire comparisons did not clearly establish a difference.[1]
The guidance could be concrete: directions helped one participant pan to Japan just outside the visible area. Errors also included confusing Panama with Costa Rica. The system works from images, without the chart’s underlying data.[1]
The design question is where an answer directs a reader’s next look. ChartAEye gives that answer a location within the chart they are already exploring.[1]
Sources & further reading
- See What I Mean: A Technology Probe of AI Assistants for Chart Exploration with Low Vision Yotam Sechayk, Lucy Jiang, Yuhang Zhao and Takeo Igarashi / ASSETS 2026 · Accessed 29 Oct 2026
- How Do Low-Vision Individuals Experience Information Visualization? Yanan Wang, Yuhang Zhao and Yea-Seul Kim / CHI 2024 · Accessed 29 Oct 2026