Research group · ETH Zurich

Dynamics in AI

We build scientific foundations for AI systems that interact with society and evolve through feedback, incentives, and markets.

AI outputs shape information and opportunities; a returning ribbon carries user feedback, while connected nodes link people to one another

Our research

We are part of the Institute for Machine Learning at the Department of Computer Science of ETH Zurich.

AI systems are intertwined with society. They impact people and evolve through feedback, incentives and markets. We study this coupling, as well as the consequences for learning, evaluation, alignment and social impact. In this pursuit, we use mathematical as well as empirical tools from machine learning, game theory, performative prediction, causal inference, and the social sciences. Our aspiration is to develop the insights and the conceptual repertoire necessary to understand AI in the social world, support oversight, and inform the safe deployment of AI models.

For relevant publications see Google Scholar.

People

Advising

If you are interested in pursuing a PhD in our group, please apply through the ETH AI Center, the Max Planck ETH Center for Learning Systems (CLS), or the ELLIS PhD program, and indicate your interest in joining our group. ETH students can also reach out by email with your CV and the academic transcript attached. External PhD students interested in a research internship are welcome to reach out by email. We do not offer undergraduate internships.

Student projects

For ETH students we offer thesis projects related to our research, including LLM evaluation, LLM alignment, AI safety, social impact and performative prediction. Let us know what you are interested in working on. Below you can find two open projects:
  • Statistical Self-Consistency in Language Models
    Keywords: forecasting, posttraining, reinforcement learning, LLM evaluation
    read more
  • Human-AI Co-Evolution
    Keywords: alignment, social impact, networks, AI safety
If you are interested in working with us, you can directly email the group members you would like to work with. Please attach your CV and transcripts, and briefly explain your interest in the project.

Contact

Celestine Mendler-Dünner, Prof. Dr.

Email:

ETH Zurich
Department of Computer Science
Institute for Machine Learning
OAT Y 13.2
Andreasstrasse 5
8092 Zürich, Switzerland

View ETH map

Map of Oerlikon showing the OAT building between the railway tracks beside the station