Hanna Dettki

I am currently a Research Associate (full-time) at the Flatiron Institute, where I develop psyphy, an open-source framework for cognitive modeling and adaptive experiments.

As a Fulbright scholar at NYU, under the supervision of Prof. Brenden Lake and Prof. Bob Rehder, I investigated human–AI alignment through the lens of causal reasoning. This work culminated in causAIign, an open-source benchmark and analysis framework for comparing human and LLM reasoning on the same causal tasks. Broadly, I am interested in narrowing the gap between human and machine cognition — working toward AI systems that, especially in sensitive domains such as healthcare, can represent users’ goals and intentions and interact in ways that are cooperative, trustworthy, and aligned with human values.

Outside of work, I enjoy cycling, board games, and trying new sports.

Education

M.Sc. Machine Learning, University of Tübingen, Germany — exp. 02/2026
- Thesis: Reasoning Strategies and Robustness in Language Models: A Cognitive View
  (advised by Prof. Charley Wu, Prof. Bob Rehder, and Prof. Brenden Lake)
M.A. Psychology, New York University (NYU) — 2025
- Fulbright scholarship covering tuition and living expenses
- Advised by Prof. Brenden Lake and Prof. Bob Rehder
B.Sc. Cognitive Science, University of Tübingen — 2022
- Semesters abroad: National Cheng Kung University, Taiwan; The University of Hong Kong
- Thesis: Probabilistic Modeling of Infectious Disease Dynamics (advised by Prof. Philipp Hennig)

Selected Publications

Do Large Language Models Reason Causally Like Us? Even Better?

H. M. Dettki, Brenden M. Lake, Charley M. Wu, and Bob Rehder

In Proceedings of the Annual Meeting of the Cognitive Science Society (CogSci), 2025

Abs Bib PDF Code

Causal reasoning is a core component of intelligence. Large language models (LLMs) have shown impressive capabilities in generating human-like text, raising questions about whether their responses reflect true understanding or statistical patterns. We compared causal reasoning in humans and four LLMs using tasks based on collider graphs, rating the likelihood of a query variable occurring given evidence from other variables. We find that LLMs reason causally along a spectrum from human-like to normative inference, with alignment shifting based on model, context, and task. Overall, GPT-4o and Claude showed the most normative behavior, including "explaining away", whereas Gemini-Pro and GPT-3.5 did not. Although all agents deviated from the expected independence of causes - Claude the least - they exhibited strong associative reasoning and predictive inference when assessing the likelihood of the effect given its causes. These findings underscore the need to assess AI biases as they increasingly assist human decision-making.
@inproceedings{dettki2025largelanguagemodelsreason, title = {Do Large Language Models Reason Causally Like Us? Even Better?}, author = {Dettki, H. M. and Lake, Brenden M. and Wu, Charley M. and Rehder, Bob}, booktitle = {Proceedings of the Annual Meeting of the Cognitive Science Society (CogSci)}, year = {2025}, }