The feeling of good fit

I get angry sometimes when I chat with my AI agents. Right now I’m reviving an old idea, using a system from dance notation to describe character voice. It was going fine, but then I asked it to synthesize some research. Suddenly its response felt off, but I couldn’t articulate why. It felt wrong in a way that elicited anger. And, right now I’m feeling like I want to walk away. But I can’t articulate why.

Emotion acts as a common currency. I can think about a variety of things that will elicit different values – different levels of positive and negative feeling. With my AI chatbot, I’m using those negative or positive feelings to navigate a feature space. How do I know what I’m trying to say before I know what I’m trying to say? That is called Meno’s paradox. And I think the answer is the fit of features in a feature space that point to something you haven’t articulated.

The usual answer is to say we make a model and compare the model to reality. I don’t think the process is that articulate. A model implies we’ve constructed something whole but on a smaller scale. The design process does not start with a prototype.

This is a typical design problem; it has requirements which have to be met; and there are interactions between the requirements, which makes the requirements hard to meet.
– Christopher Alexander, Notes on the Synthesis of Form

The architect, Christopher Alexander, talks about the solution to a design problem as something that has “good fit”. But that good fit is defined less by its shape, and more by the feeling that all of the sharp edges have been removed.

When I’m angry with my AI, it’s because the fit is off. At first I can only say that I feel something is off. Articulation is the process of finding what the feeling has noticed.