VizChitra 2025: India's first community-driven Data Visualization conference

Imagine trying to read an entire spreadsheet cell-by-cell through a straw; that’s how many blind people describe “looking” at a bar chart today. Some have even failed classes because the graph on the midterm was invisible to their screen reader.

The Problem:
Data visualizations compress thousands of data points into a glance, unless you rely on audio. Even with ARIA labels, a screen reader still exposes information linearly, piling cognitive load until the user feels, in their words, “like I’m seeing the world through a microscope.” The result: lost time, lost grades, lost opportunities.

Our Hunch:
What if AI could be your eyes and let you hear a data visuslisation.

The Journey:
We started our project with very rough concepts. Early tests with a blunt prompt, “Describe this image” flopped. Users sifted through color callouts and axis trivia to find the one sentence they needed. So we got personal, conducting iterative sessions and ripping out every adjective that didn’t drive insight.
The result is, I call it the Onion Peel Theory:
Outer Layer - Immediate takeaway, what is this visualization & what direction is it pointing?
Middle Layer – Key Numbers, biggest rises and falls, notable Clusters
Core – Natural-language queries such as “Which city outperformed the average in Q3?” answered on demand.
Many ideas die in the graveyard of big tech possibilities. To push an idea to reality you require a little bit of ‘delusion’. I pushed this idea across forums more than a year. To finally this year see the V1 of it come to life: https://blogs.windows.com/windows-insider/2025/04/11/announcing-windows-11-insider-preview-build-26120-3863-beta-channel/#:~:text=Making visual content more accessible with image descriptions in Narrator

To Be a Real Lover, You Have to be a Hater:
Loving the vision means realising the pitfalls

1. Bias in, bias out: A dataset is simplified using a graph, on which AI is applied to get visual descriptions, there is data loss at every step. Therefore using AI to interpret a chart can amplify its designer’s bias. Long-term, we need a way to work on the raw data used to derive the visual.
2. Complexity breeds hallucination. Accuracy rocks on single-series line charts, wobbles on multi-facet scatterplots. Trust is non-negotiable for accessibility. Therefore application of AI to generate visual descriptions requires better benchmarking.
3. Everything is data viz: Our defination of data viz often stays limited to dashboard or trend reports. But high information density UI like a data picker on a Booking’s site is also a data viz. A seat maps while booking a flight has similar problem and so does a a calendar view of your meetings. As quotes by my brilliant PM Nidhi Jain “While you can with in a second which slot on your day is empty to book a meeting, I have to hear each cell and then figure out oh I have my afternoon free”
4. Design for the future: Pace at which technology is shifting is faster than the speed at which software reaches market. When I started this project, AI models that were publicly available were average at reading visual information and were also taking a lot of time to generate response. However within 3 months, all that was ancient history.

Disability is a minority we all can be one day:
When you design for the outlier you design for everyone. A piece of technology that works for someone with low mobility, also works for someone who has an injuired hand or a parent holding a child. This idea of leveraging AI to make data viz accessible while designed for people with visual impairment, can also be extending to people with low data literacy or citizen in rural india for civic enagement.
When a graph can talk, more people can listen.

{Actionable methods to make data viz accessible using established guidelines and emerging practices.
How AI can transform data accessibility for visually impaired users, bridging critical information gaps.}

{Young, mid-career, UI, UX, Product Designers, Product Managers}

{I am Sabhyata Jain, a product designer and visual artist. By day, I create AI-powered productivity tools at Microsoft, having previously designed Accessibility Tools for Windows 11 that enhance digital experiences for blind and visually impaired users. By night, I lead design & research at Bharat Digital, a non-profit reimagining the role of design in gov-tech in India. With a background in Visual Communication from the NID, Ahmedabad, I explore the intricacies of womanhood and solitude through my personal project “Bade Shehr Ki Ladki”.}