Erin May sits down with Larry Becker and Gavin Lew of Bold Insight to explore how researchers can prototype AI experiences that don't exist yet. They're joined by Katie Johnson, Global Head of Research for Panasonic Well. All share their methods for testing everything from voice interfaces for blind users to family wellness bots using low fidelity simulation techniques.
⭐️ Larry, Katie, and Gavin are presenting their work at EPIC 2025 in Helsinki ⭐️
The conversation reveals how AI research requires new approaches beyond traditional usability testing. Katie explains her work simulating family interactions with AI assistants, where researchers act as human filters for LLM responses to study relationship building over time. The team demonstrates how even small changes in AI personality can be detected by users, highlighting the sensitivity of human-AI relationships.
Gavin emphasizes the critical importance of getting AI right from the first interaction, unlike traditional MVP approaches. When AI fails to understand or respond appropriately, users often abandon it permanently, making prototype research essential for building trust before launch. The discussion provides practical frameworks for researchers working with emerging AI technologies and the evolving methods needed to understand human-AI interactions.
Highlights
- [06:25] Prototyping AI experiences mindset
- [20:53] Human simulation in AI prototyping
- [31:15] Human vs bot detection challenges
- [38:52] Context interaction and trust framework
- [49:33] Team collaboration in groundbreaking research
- [51:27] Trust themes in AI development
Resources
- The State of User Research 2025
- AI For User Research 101 Course
- 20+ AI Tools for User Researchers
