About us
Raisonne.ai publishes informal field studies of artificial intelligence, seeking to measure and audit the priorities, worldview, and bias of large language models: observations primarily driven by our concern as to their potential impact on public health.
The name borrows from catalogue raisonné, the inventory of an artist’s work. We aren’t aiming for completeness, but if something is worth looking at, it is worth looking at carefully.
Each study is published in three layers — an essay, a full report, and the complete experimental data for browsing or download — so every claim can be reviewed against the responses that produced it. We ask specific questions of AI systems — what they prefer, what they refuse, what they commit themselves to when pressed — and record the answers with the care those questions deserve. The models may be black boxes, but they exhibit behaviour that we can critically examine. We have more ideas in the pipeline.
We work for no institution, but we publish our methods and data in full, and encourage replication and peer review. If you wish to support our work, we welcome funding or sponsorship but value our independence.
Contact
Social
Colophon
Built with Zola. Links turn International Klein Blue (#002FA7) when touched. The raisonne wordmark uses a light variant (#BEBEFA). Set in Newsreader and Inter. This site is written in collaboration with Claude Code, and Claude is also a subject of study. We think the tension is worth disclosing.