Mimo — Making Visual Learning Accessible

Every diagram has a story. Mimo helps blind and low-vision students explore it through voice, touch, and AI that finds missing explanations—so no learner misses the bigger picture.

Comment

Mimo — Every Diagram Has a Story

Inspiration: What if learning depends on something you cannot see?

Imagine sitting in a biology classroom.

Your teacher says, "Look at the diagram."

Everyone starts learning. But what happens when you cannot see it?

A screen reader might tell you, "This is an animal cell." But it may not explain where the nucleus is, how the organelles connect, or why those relationships matter.

A diagram is more than a collection of labels. It is a story of how things work together.

We built Mimo because students deserve access to that whole story.

What it does

Mimo is an AI-powered learning companion that turns visual diagrams into interactive, accessible lessons.

Instead of simply describing what is in an image, Mimo helps students understand what it means.

Our experience is designed around four ideas:

  • Explore: Navigate a diagram and discover its different parts.
  • Listen: Hear friendly, simple explanations instead of dense textbook descriptions.
  • Understand: Learn how different parts connect and why they matter.
  • Discover what's missing: Use our Missing Knowledge Detector, called Gap Check, to suggest important explanations that may have been left out.

For example, a description might say:

"The mitochondria is an organelle inside the cell."

But it may never explain that mitochondria help produce energy the cell can use.

Mimo's goal is to notice that missing connection and help complete the lesson.

How we built it

We designed Mimo around Google's Gemini models and an accessibility-first learning experience.

Our approach combines:

  1. Diagram understanding: Analyze educational images and their descriptions.
  2. Accessible explanations: Turn complex visual information into understandable language.
  3. Interactive exploration: Help students learn one part at a time.
  4. Gap Check: Compare the available explanation against important diagram concepts and suggest missing information.
  5. Verification: Evaluate whether the suggested information is accurate and genuinely useful.

Rather than treating AI-generated descriptions as automatically complete, Mimo asks a more important question:

"What does the student still need to know?"

What makes Mimo different?

Most image-description tools focus on explaining what is visible.

Mimo explores a different challenge: finding what an explanation leaves out.

We tested Gap Check using Gemini 3.5 Flash-Lite on four biology diagrams. Each diagram had an original description and a deliberately shortened version.

Here is what we observed:

  • 6 accepted suggestions with the original descriptions.
  • 21 accepted suggestions when information was deliberately removed.

When descriptions contained less information, Gap Check suggested more explanations.

This is promising early evidence that the system responds to missing information. However, these results do not yet prove that every suggestion is correct or improves learning. Human evaluation is our next step.

Challenges we faced

1. Describing is not the same as teaching.

Knowing that a nucleus exists is different from understanding its role inside a cell. We had to think beyond labels and focus on meaningful relationships.

2. AI can sound confident while missing important information.

We designed Gap Check to identify possible omissions, while recognizing that suggested explanations still need verification.

3. Accessibility requires thoughtful design.

A useful experience must be understandable, navigable, and comfortable for students—not simply powered by AI.

What we learned

We learned that more words do not always mean more understanding.

The most useful explanation is one that helps a student build a clear mental picture of the concept.

We also learned that testing AI requires more than an impressive demo. We need to measure what it identifies, what it misses, and whether its suggestions genuinely help learners.

What's next?

We want to expand Mimo beyond biology diagrams, test it with blind and low-vision learners, improve the accuracy of Gap Check, and evaluate how much students actually understand after using it.

Our long-term vision is to make complex visual learning more accessible across science, technology, engineering, and mathematics.

Because accessibility should not stop at describing what is there. It should help every student understand what matters.

Mimo — Don't just describe the diagram. Unlock the learning inside it.

Built With

Share this project:

Updates