It began as a discrepancy I couldn't explain and hadn't gone looking for, followed until I measured it.
Since early 2023 I had been keeping a private record of my own AI use, just the exchanges that felt like something had shifted, and at some point I noticed a pattern in what I was writing down that the interaction itself didn't explain. I was pushing back on the AI more than I used to, which read like my judgment getting sharper, but nothing in the record told me whether it actually had.
I believed that reading for a long time before I checked it. Going back through the record, I could not get it to confirm the feeling either way, so I started treating them as two separate readings, what I felt about an exchange and what the exchange actually showed.
That gap bothered me enough to follow it. I asked whether it was only me, then whether there was a pattern, then whether the pattern could be measured, across 38 months and 973 conversations. By the time I could answer that, I had built the first instruments for it, a way to score what an interaction does to a person's judgment and tools that make the shift visible while it happens.
I went through this same sequence with each of the projects here, though the Ben Study is the only one where I kept enough of the intermediate work to actually show it. The others started the same way:
The Ben Study is where I named the felt–trace gap, the distance between how independent your judgment feels and where your reasoning actually stays. After it, everything I built tries to measure that same gap in someone other than me, because my own record is only one case.
Felt vs Trace is the one measuring instrument. It stages the gap as a first-person experience, the only place the construct is measured directly. A framework operationalizes it for scoring, MCAB (three states, four dimensions), drawn from the record. And two tools intervene on it, Intuition Canvas and Friction Family, which hold or provoke judgment.
The diagram is a branch, not a chain. The framework and the tools sit at the same distance from the record because they do different work with it, one naming the effect and the other designing against it, while Felt vs Trace sits on its own layer as the only one of them that measures the gap directly.
I'm an independent design researcher in Seoul, studying what long-term AI use does to people's judgment and building the instruments that let me measure it.
Before this I spent eight years designing products, moving from fintech portals in New York to a rider app that ran across eight APAC markets and then to Samsung's global platform at EY Digital, and it was on an AI project there that I first watched users lose track of which decisions were still their own. Because I had been keeping a daily record of my own AI use since early 2023, I could go back and read my way through the same shift from the inside, which is where this research actually began.