# Will Mayner

> Will Mayner is a computational neuroscientist and AI interpretability researcher. Over a decade he built the mathematical core of Integrated Information Theory (IIT) and the field's standard open-source software, and he now focuses on mechanistic interpretability and AI safety.

## About

Will Mayner is a researcher at the Center for Sleep and Consciousness at the University of Wisconsin–Madison, where he developed the mathematical formalism of Integrated Information Theory (IIT) alongside Giulio Tononi. He has a decade of experience formalizing IIT, building scientific software, and analyzing large-scale neural data, and is now focused on mechanistic interpretability and AI safety — work he considers the most impactful use of his skills.

He created PyPhi (https://github.com/wmayner/pyphi), the standard open-source library for IIT research (150+ citations). He has 13 peer-reviewed publications, including first-author work in PLOS Computational Biology, Entropy, and eNeuro, and co-authorship in Nature Neuroscience. He was a finalist for the Anthropic Fellows program (top 130 of ~5,000 applicants).

He holds a Ph.D. in Neuroscience from UW–Madison (advised by Giulio Tononi) and a Sc.B. in Mathematics–Computer Science from Brown University.

## Research interests

Mechanistic interpretability; AI safety; machine consciousness and model welfare; neural geometry; integrated information theory (IIT); computational neuroscience.

## Writing

- [Decomposing introspection in LLMs: representation and report](https://willmayner.com/blog/decomposing-introspection/): Decomposes concept-injection introspection in open models (Gemma-3 12B, Qwen-2.5 32B) into separable representation and report components, reconciling apparent conflicts across prior protocols.
- [Belief manifolds, and how to steer along them](https://willmayner.com/blog/shape-of-beliefs/): A reproduction of Sarfati et al.'s "The Shape of Beliefs" — how LLMs encode in-context beliefs as curved manifolds, and how manifold-aware steering changes them with fewer side effects than linear steering.

## Pages

- [Home](https://willmayner.com/)
- [CV](https://willmayner.com/cv/)
- [Publications](https://willmayner.com/publications/)
- [Resume](https://willmayner.com/resume/)

## Links

- GitHub: https://github.com/wmayner
- Google Scholar: https://scholar.google.com/citations?user=BWRciS8AAAAJ&hl=en
- LinkedIn: https://www.linkedin.com/in/will-mayner
- X: https://x.com/wmayner
- Email: wmayner@gmail.com