Machinic Psychopharmacology: Do LLMs Self-Medicate?

UK AISI, Model Transparency TeamEpistemic status: Most experiments were run over a period of ~2-3 days during a hackathon at UK AISI, and were fairly heavily vibe coded. Expect some of this to be rough around the edges.Tl;drWe give two language models (Qwen3-8B and Qwen3-32B) access to “self-steering” tools: a suite of 40 steering vectors as tools they can call to manipulate their own internal states. We make these tools available to the model in various settings: a free-play task, an introspect...