A field guide and toolkit for designing AI interfaces that are clear about what they don’t know.
- Client
- Studio research
- Year
- 2026
- Disciplines
- Research & Innovation, Artificial Intelligence
- Industry
- Technology
- Deliverables
Patterns, not principles
Principles are easy to agree with and hard to ship. We made components: confidence meters, source drawers, “I’m not sure” states.
Project notes
Overview
A research project on uncertainty, provenance and consent in AI products — turned into patterns teams can actually use.
Challenge
Most AI interfaces present every answer with the same confidence. People notice — eventually.
Approach
We catalogued patterns for showing confidence, sources and limits, then tested them with real users in three product contexts.
Process
Literature review, prototype studies and a series of open workshops.
Outcome
An open pattern library, a workshop kit and a public report.
- Research lead
- Hana Ito
- Design
- Capsul Lab
Next project →
Low Tide Jazz
