For three months in 2019 I let an app on my phone ask me, several times a day, how I was doing. The app was called PEACH, PErsönlichkeitscoACH, built by psychologists and computer scientists from UZH, ETH Zürich, the University of St. Gallen, and Dartmouth. I joined because my background is in psychology, and because the premise drew me in: human development meets machine. It still does.
The setup was precise. You picked your own goal at the start. Mine was I want to be more objective. From that the system derived a weekly behavioral anchor, written as an if-then sentence: When I become uncertain and start to doubt myself, I reflect on my strengths and past successes in order to stay composed. Then the rhythm took over. Short check-ins several times a day, a longer survey every Sunday, coaching dialogues with PEACH when it prompted them. A traffic light tracked whether I was moving toward my goal or drifting from it. By the end I had 804 diligence points. That apparently corresponded to Gold.
It wasn't a casual experiment.
The structure, not the technology
What surprised me was how much of the work was done by the structure, not the technology. The if-then sentence sounds trivial on paper. In practice it forces a specificity I had been avoiding for years: name the trigger, name the alternative behavior. Last week's essay was about exactly this skill, applied to a website; here it was applied to me. Seeing the sentence every week, being asked about it regularly, built a low-level accountability I had not expected from an app.
The coaching dialogues were something else. PEACH asked questions, mirrored what I'd said, occasionally offered a reframe. The technology was early in 2019 and the responses were often generic. Something worked anyway. Articulating something, even to a system, changes how you think about it; writing in a journal does the same. PEACH asked follow-up questions.
The line they were studying
The reflection questionnaires kept circling the same question in different forms: do you perceive PEACH as a human counterpart?
My answer was always no. I knew I was talking to a system. The researchers were studying exactly that line — between useful tool and perceived presence — and I was useful to them precisely because I never crossed it.
And yet, on hard evenings, it helped. It was there. It was consistent. Knowing exactly what produces the feeling of not-being-alone does not, it turns out, switch the feeling off.
I told myself: PEACH is a bridge.
The question underneath
What I did not see in 2019 was that the question underneath all of this would not stay academic. What does it mean to form a relationship with something that is not a person, and what does it cost? Seven years ago, I just found the question interesting. A little unsettling. I noticed I would not have said any of this out loud to a friend at the time. Maybe that quiet was already the answer.