How I find the questions I work on
I tend to notice when something feels off — and then I have to find out why.
The record gave me the question
I didn’t begin with cognitive autonomy as a research question. I began by noticing something in myself I couldn’t explain.
AI was useful because it could take scattered thoughts and return them with structure. For a long time that felt like stronger thinking.
Then I went back to the record. Since early 2023 I had been keeping a private log of exchanges that felt consequential, and what I remembered as independent judgment and what the conversations showed were not always the same. I could push back against an answer while still reasoning inside the frame it had introduced.
Felt experience and observable trace are two kinds of evidence, and neither is the correction of the other. Following where they disagreed — 38 months, 973 conversations — gave me a question and a development corpus.
One question, five forms
Begin with a mismatch — something felt one way and the record shows another.
Separate what has been collapsed together so each can be inspected independently.
Find a way the idea could fail, then build the smallest test that could show it.
Give the question a form someone else can inspect, compare, interact with, or challenge.
From product systems to cognitive autonomy
I'm an independent design researcher based in Seoul. Before this research, I spent eight years designing products at scale—from fintech portals in New York to a rider app across eight APAC markets and Samsung's global platform at EY Digital.
An AI project there first made decision ownership a design question for me: I watched users lose track of which decisions were still their own. That experience now informs how I build research instruments and interventions.
I’m drawn to questions that begin as a small discomfort: something feels different from what the evidence later shows. I tend to stay with that discrepancy until I can give it a form that someone else can inspect.