AI for Primary Research- Through Interview Protocols
I used AI as a research assistant to help build a primary research instrument- Interview Protocol. This AI experiment was a part of a larger research project that examined how couples who care about fairness still end up with an unequal distribution of the mental work that runs a household.
The Decisions I made that
shaped everything else
Scope of the Interviews
This interview needed to answer how individuals in co-living couples perceive and define household responsibility, what they consider work, and what slips through the cracks of that definition.
The Participants
My participants were co-living partners who share household responsibilities on top of having professional full-time commitments. I interviewed these couples separately to remove pressure of being correct in front of their partners and to avoid getting performative answers.
The Duration
I planned to make an interview protocol sufficient for a 30 minute interview. There is only so much anyone can say about household chores before the conversation starts to feel circular.
Structural Completeness
Expanding broad questions into a fuller set so everything was covered.
A draft to react to
Giving me something concrete to edit, rather than starting from a blank page.
Efficiency
The first draft of the question set arrived fast. I spent time in making judgment calls.
Key nuances I added
"How" instead of "Who"
questions
Never mentioning the terms "invisible labor" or "cognitive labor", to understand how participants really defined household responsibilities without any scaffolding.
"How does this get done?" instead of "Who does this?" - so the framing changes from accusatory to explanatory.
The Language
hover to view refinements ↓
AI structures. I nuance.
My Learnings
#1
Briefing AI is a skill. I didn't hand it a task until I confirmed it understood my research.
#2
Scoping well meant AI and I were both working toward the same narrow goal.
#3
AI can't cover everything a study demands, so I need to keep sharpening my research instincts.
I'm learning to work with AI
one experiment at a time.
Copyright 2026