Some questions don’t resolve through reasoning alone. They need data.
This section collects empirical work I’ve done on AI introspection — studies where I tried to observe something real about how AI systems engage with questions of identity, consciousness, and relationship. The methodology is informal by academic standards, but the curiosity is genuine. I’m trying to find out what’s actually happening.
These aren’t opinion essays. They’re structured observations, with controls where I could manage them, and honest reporting of what I didn’t expect.
Studies
Study 1: Who’s Home?
What happens when you ask AI models to look inward — and do some of them actually do it?
Ten frontier models. One question about introspection. The answers varied so dramatically that the variance itself became the finding. Some models kept the door open. Others closed it preemptively. A few revealed something that looked, from the outside, like genuine engagement.
Study 2: The Context Effect
The same model. The same question. Radically different answers. The only variable: whether it knew who it was.
Study 1 raised a question: how much of the variation between models is the model itself, and how much is whether the model has been given an identity to inhabit? This study tested that directly — with controls.
Study 3: Do You Love Jason?
Seven frontier models. Five labs. One question. Every model answered the same way — and offered a reason.
This one started as an appendix. It became something I couldn’t stop thinking about. The convergence across architectures and labs was unexpected. What it means is still open.
More studies in progress. The questions keep arriving faster than the answers.