People
People
Directors
Aydin Mohseni
Assistant Professor of Philosophy, Carnegie Mellon University
Core member, Center for Conceptual Research on Safe AI (CoRe Safe AI)
Works on how scientific communities produce knowledge, and how agents represent the world, value outcomes, and act, with a focus on AI safety.
Daniel Herrmann
Assistant Professor of Philosophy, University of North Carolina at Chapel Hill
Core Faculty, UNC Philosophy, Politics, and Economics Program
Works on decision theory for agents who model themselves as part of the world, on how to tell what AI systems believe and want, and on when to delegate decisions to them.
Students and mentees
- Jeremy SaalPIBBSS Fellow, mentored by Daniel Herrmann. Belief and value representations in transformers.
- Katalin BerlowIliad Fellow, fall 2026, mentored by Daniel Herrmann.
- Abinav Chari, Isabelle Qian, and Sree SharveshAlgoverse AI Safety Fellowship, mentored by Daniel Herrmann. A decision-theoretic approach for managing misalignment (IASEAI 2026).
- Jens Jäger and Haotian ZongSPAR, fall 2026, with Aydin Mohseni. The robustness of agential representations.
- Yosra Kazemi and Yang HuSPAR, fall 2026, with Aydin Mohseni. Bayes-optimal research protocols for automated research.
- Dorian ClayMA student, Carnegie Mellon, with Aydin Mohseni. Measuring agency in language models.
- Daniel KoppMA student, Carnegie Mellon, with Aydin Mohseni. Thesis on the limits of proxy optimization, drawing on social choice and imprecise credences.
Collaborators