EchoMinds Case Study
EchoMinds began as a human-centered design exploration into how machine learning could support people who are blind or low vision in everyday note-taking and information retrieval. Through summer fieldwork at the Carroll Center for the Blind and MABVI, the team learned that employment, personal organization, and access to information all depend on tools that are flexible, context-aware, and easy to use without sight.
The project evolved from a broad set of ideas into a prototype for a notetaking app that helps users search for notes by meaning rather than exact keywords. The final experience reflects a core lesson from the design process: technology should support the user’s mental model, not force them to adapt to an opaque system.
Fieldwork
The team began by observing life at the Carroll Center and MABVI, attending presentations on disability justice and access technology, and practicing interviews and fieldwork techniques. These experiences helped the team move from abstract ideas about AI to concrete questions about how real users manage information, organize work, and navigate technology without sight.
Ideation
- Interviews with community partners and students to understand how notes are created and retrieved
- Design activities around a machine-learning-powered resume builder
- Roleplay sessions for a screen-reader-oriented web navigation tool
- Exploration of a notetaking app that could retrieve notes by semantic similarity
Prototyping
After evaluating several concepts, the team focused on a notetaking app that would retrieve notes by semantic similarity. The prototype combined a simple mobile interface with a sentence-transformers-based backend to test whether machine learning could support meaningful retrieval in everyday tasks. The design process showed that the biggest challenges were not only technical, but also about making the system understandable to the user.
Testing
The team tested the prototype internally and then with community partners from the Carroll Center, Olin, and MABVI. The results confirmed that users want tools that make it obvious how information is being organized and retrieved. The most important design lessons were to keep note granularity understandable, support retrieval by concept, and design for real-world contexts rather than abstract lab scenarios.