You Can’t Outsource the Learning
AI tools can make the research process faster and more efficient, from defining research questions to recruiting participants, analyzing data, and writing reports. But in that process, there is a risk that the deeper learning teams gain by directly experiencing users’ experiences disappears.
Two Byproducts of Research
- Outputs: Themes, insights, recommendations, reports, and similar deliverables, which AI can replace quickly and almost perfectly.
- Learning: The shift in mindset and empathy for user experience among the people conducting the research, which AI cannot experience on their behalf.
Cognitive Science View: The Self-Generation Effect
- People remember and understand information much better when they think through and organize it themselves, rather than simply reading it.
- If a team only reads a polished report organized by AI, it can easily fall into the “illusion of knowing,” while actual understanding and memory remain weaker.
The Power of Direct Observation
- Seeing a metric like “40% of users failed the signup process” is completely different from watching a user struggle in front of the screen for over a minute and feeling that frustration in real time.
- Designers make design decisions based not only on percentages, but on the mental models they have directly felt and observed.
The Right Role for AI
- Principle: *“Automate the work that supports thinking, but protect the core work that creates thinking.”*
- Tools such as Otter.ai for transcription, Dovetail for data organization, and Claude for drafting research questions can be extremely useful as support tools, but the essential learning process of empathizing with, debating, and refining users’ pain points needs to be done by people directly.
One-Line Conclusion
- Research outputs are only the deliverables produced by research. The real essence of research is the learning itself: the process through which the team experiences users and changes because of it.
