Visual Communication for Low-Vision Information Access

When I magnify a chart enough to read one bar, the axis and legend can disappear from my view. I can see the bar, but I lose what it means.

I have albinism and low vision, and I use screen magnification every day. That gap between detail and context connects my research. Writing my doctoral symposium paper meant putting the connection into words.

My paper “Visual Communication for Low-Vision Information Access: From Personalized Highlights to AI Agents” (Sechayk, 2026) was accepted to the UIST 2026 Doctoral Symposium. 🎓 In November, I’ll present it in Detroit to a panel of senior researchers.

What is a doctoral symposium? PhD students present their dissertation plans to senior researchers and get feedback while there is still time to change them. Think of it as a friendly stress test for the research.

Seeing a fragment at a time

Magnification makes things bigger by showing only part of the screen at a time. That tradeoff matters when information is spread across a page.

In a lecture video, an instructor can mark a sentence outside my magnified view. In a chart, I can see the bar under my pointer while its axes and legend sit elsewhere. Sometimes I don’t even know what I missed.

Magnification makes what I'm reading big and clear, but only a piece of the sentence fits in view at once. The heading, chart, and legend around it are out of sight.

Using the vision we have

Many people with low vision prefer to use the vision they have. Speech, sound, and touch can also make visual information accessible, but they don’t directly support that way of seeing.

So what if we build on the vision people already use?

My research follows ability-based design: start with what a person can do, then shape the tool around their abilities. For me, that means working alongside tools such as magnification, allowing people to adjust the visual support, and designing with people with low vision. Their needs and strategies differ, and their vision can change with the situation.

Visual communication, in two directions

The thread that ties my projects together is visual communication in two directions. A tool can show me where to look with a highlight or bring missing context into my view. I can show a tool what I’m looking at by sharing my magnified view and where I point. What happens when both directions work together?

A four-column timeline of my research, read left to right. VeasyGuide (ASSETS 2025) highlights an instructor's mark on a lecture slide. On-Cursor Visual Context (CHI 2026) shows two chart interactions, Dynamic Context and Mini-map. Visual Context for AI (ongoing) shows a user pointing at a bar in a chart while an AI assistant answers with its value. Visual Communication with AI Agents (future) asks whether visual dialogue with AI builds independence or creates reliance, and whether AI should tell, highlight, guide, or show. Labels along the bottom read: how tools can communicate visually with low-vision users, how low-vision users communicate visually with AI, and how bidirectional visual communication impacts agency.

My research trajectory. It starts with tools that communicate visually to low-vision users, moves to low-vision users who communicate visually with AI, and heads toward visual communication in both directions.

From tools that show to tools that see

My dissertation follows this idea step by step. Each project taught me something, and each lesson shaped the next one.

VeasyGuide: a highlight is a message

In a lecture video, an instructor may point to a part of a slide without naming it. VeasyGuide (Sechayk et al., 2025) makes that action easier to find with a highlight that can appear before the action and with magnification that can follow it. My co-design partners helped decide how the highlight should look and behave. You can read how we designed VeasyGuide or try it in your browser.

Three panels of VeasyGuide. Panel a, the VeasyGuide player: a lecture slide titled Machine Learning, where a red bounding box with a translucent yellow fill highlights the phrase the instructor is marking, next to a hand-shaped pointer icon. Panel b, zoomed-in view: pressing the Z key shows a magnified view of the highlighted slide text. Panel c, personalization panels: settings for the fill color and opacity, border color and width, pointer shape and scale, filters named Invert, Thicker, and Thicker dark, and the highlight's shape and animation.

VeasyGuide marks an instructor's action with a personalized highlight and pointer icon (a), can zoom in to follow the action (b), and lets each learner tune how the highlight looks (c).

Takeaway: a highlight has to be findable, predictable, and adjustable to communicate anything. It should also leave the learner free to decide whether to follow it.

But a chart poses a different problem: the context I need is spread across axes, legends, and data points.

On-Cursor Visual Context: bring context into view

On-Cursor Visual Context (Sechayk et al., 2026) explores two ways to keep chart context available under magnification. Dynamic Context brings axes and legend information near the pointer, with a crosshair to help trace values. Mini-map keeps a small overview of the chart in view.

The crosshair came from a participant’s workaround: they used the edge of their docked magnifier as a ruler because the chart’s grid lines were too faint.

Top: an illustration of a person leaning close to a monitor to read a bar chart, next to an annotated bar chart titled Car Counts by Number of Gears, with callouts labeling its overview, legend, grid lines, and axes. Bottom: two monitors showing visual context interactions under screen magnification. On the left, Dynamic Context shows a magnified part of the chart with the Count axis values, the Transmission legend, the Number of Gears axis label, and a crosshair gathered in a compact area around the pointer. On the right, Mini-map shows a magnified part of the chart with a small overview of the whole chart in the corner and a dot marking where the pointer is.

The context a chart reader needs (top), and two ways to bring it into the magnified view (bottom): Dynamic Context gathers the axes, legend, and a crosshair around the pointer, and Mini-map keeps an overview of the whole chart in view.

Takeaway: context can move into the magnified view, and people’s workarounds can show us how. But a fixed display can’t answer the next question a reader has about the chart.

Visual context for AI: let the tool see what I see

What if an AI assistant could see where I’m looking in a chart? In my ongoing work, I’m exploring an assistant that receives the user’s magnified view and pointer location as context. That lets a person point and ask without first describing exactly where they are in the chart.

The next question: can the assistant answer in a way I can see? The user can show the assistant their visual context, but the assistant answers in words. It can tell me where to look, but it can’t show me. That brings me back to the problem VeasyGuide started with.

Next: both directions

I want to study an assistant that can also answer visually: highlight a region, guide my attention along a path, or adjust the view. Then the person and the tool could each show the other what they mean. I also want to explore this beyond charts, starting with information seeking on web pages and eventually extending to documents and the physical world.

Put together, the projects form one arc. VeasyGuide and On-Cursor Visual Context show that tools can communicate visually with people with low vision. The AI work turns this around, so people can communicate visually with their tools. The next step closes the loop.

Who decides where to look?

Should the tool move my view, highlight something and let me look, or just answer? The right response may depend on the person and the task.

Tell, highlight, guide, or adjust the view. These are different kinds of help, not a ranking, and the right one depends on the person and the moment.

More assistance is not always more access. Being able to look for myself is part of what I value. I don’t want a tool that sees for me. I want one that helps me see.

My dissertation asks how people with low vision and their tools can show each other what they mean, extending the sight we already use rather than replacing it. Through all of it, the person stays in control of what they see.

See you in Detroit

At the symposium, I want feedback on how to measure visual access and agency, and how to bring these projects together into one framework. I’m also looking forward to the panel’s advice on building a research career as a disabled researcher.

If you’re at UIST 2026, come say hi!

If you want to read more about why I care so much about designing with the vision we have, check out my post on why my needs are not “special”.

I’m grateful to my advisors, Takeo Igarashi and Ariel Shamir, and to everyone who shared their experiences in these studies.

If you found this useful, please cite the paper:

Sechayk, Y. (2026). Visual Communication for Low-Vision Information Access: From Personalized Highlights to AI Agents. In Adjunct Proceedings of the 39th Annual ACM Symposium on User Interface Software and Technology (UIST Adjunct ‘26). ACM. https://doi.org/10.1145/3830397.3841706

or as a BibTeX entry:

@inproceedings{sechayk2026visualcommunication,
  author    = {Sechayk, Yotam},
  title     = {Visual Communication for Low-Vision Information Access: From Personalized Highlights to AI Agents},
  booktitle = {Adjunct Proceedings of the 39th Annual ACM Symposium on User Interface Software and Technology},
  series    = {UIST Adjunct '26},
  year      = {2026},
  publisher = {Association for Computing Machinery},
  address   = {Detroit, MI, USA},
  doi       = {10.1145/3830397.3841706}
}

Charts, slides, and web pages remain difficult to access for the estimated 295 million people globally with low vision (LV). Much accessibility research translates visual content into speech, sound, or touch, but LV people often prefer to use their residual vision. My dissertation follows ability-based design to create tools motivated by residual vision use. My ongoing work extends this to ability-based AI systems that adapt AI interaction to LV users’ abilities. The central method is visual communication between tools and LV users. My completed work explores tool-to-user communication: VeasyGuide highlights instructor actions for LV learners in videos, while on-cursor interactions restore chart context hidden by magnification. Both systems improved access in studies with LV participants. My ongoing work examines user-to-tool communication through an AI assistant that receives the user’s magnified view and point of interest during chart exploration. My planned work asks how AI can answer visually, combine both directions, and balance assistance with user agency. I plan to extend this communication beyond charts to the web, documents, and eventually the physical world. Ultimately, my dissertation advances personalized AI that uses bidirectional visual communication to support the independence and agency of LV people.

Instructors often rely on visual actions such as pointing, marking, and sketching to convey information in educational presentation videos. These subtle visual cues often lack verbal descriptions, forcing low-vision (LV) learners to search for visual indicators or rely solely on audio, which can lead to missed information and increased cognitive load. To address this challenge, we conducted a co-design study with three LV participants and developed VeasyGuide, a tool that uses motion detection to identify instructor actions and dynamically highlight and magnify them. VeasyGuide produces familiar visual highlights that convey spatial context and adapt to diverse learners and content through extensive personalization and real-time visual feedback. VeasyGuide reduces visual search effort by clarifying what to look for and where to look. In an evaluation with 8 LV participants, learners demonstrated a significant improvement in detecting instructor actions, with faster response times and significantly reduced cognitive load. A separate evaluation with 8 sighted participants showed that VeasyGuide also enhanced engagement and attentiveness, suggesting its potential as a universally beneficial tool.

Despite widespread use, charts remain largely inaccessible for Low-Vision Individuals (LVI). Reading charts requires viewing data points within a global context, which is difficult for LVI who may rely on magnification or experience a partial field of vision. We aim to improve exploration by providing visual access to critical context. To inform this, we conducted a formative study with five LVI. We identified four fundamental contextual elements common across chart types: axes, legend, grid lines, and the overview. We propose two pointer-based interaction methods to provide this context: Dynamic Context, a novel focus+context interaction, and Mini-map, which adapts overview+detail principles for LVI. In a study with N=22 LVI, we compared both methods and evaluated their integration to current tools. Our results show that Dynamic Context had significant positive impact on access, usability, and effort reduction; however, worsened visual load. Mini-map strengthened spatial understanding, but was less preferred for this task. We offer design insights to guide the development of future systems that support LVI with visual context while balancing visual load.