Preference Learning for AI Alignment: a Causal Perspective

Kasia Kobalczyk, Mihaela Van Der Schaar
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:31063-31083, 2025.

Abstract

Reward modelling from preference data is a crucial step in aligning large language models (LLMs) with human values, requiring robust generalisation to novel prompt-response pairs. In this work, we propose to frame this problem in a causal paradigm, providing the rich toolbox of causality to identify the persistent challenges, such as causal misidentification, preference heterogeneity, and confounding due to user-specific factors. Inheriting from the literature of casual inference, we identify key assumptions necessary for reliable generalisation and contrast them with common data collection practices. We illustrate failure modes of naive reward models and demonstrate how causally-inspired approaches can improve model robustness. Finally, we outline desiderata for future research and practices, advocating targeted interventions to address inherent limitations of observational data.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-kobalczyk25a, title = {Preference Learning for {AI} Alignment: a Causal Perspective}, author = {Kobalczyk, Kasia and Van Der Schaar, Mihaela}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {31063--31083}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/kobalczyk25a/kobalczyk25a.pdf}, url = {https://proceedings.mlr.press/v267/kobalczyk25a.html}, abstract = {Reward modelling from preference data is a crucial step in aligning large language models (LLMs) with human values, requiring robust generalisation to novel prompt-response pairs. In this work, we propose to frame this problem in a causal paradigm, providing the rich toolbox of causality to identify the persistent challenges, such as causal misidentification, preference heterogeneity, and confounding due to user-specific factors. Inheriting from the literature of casual inference, we identify key assumptions necessary for reliable generalisation and contrast them with common data collection practices. We illustrate failure modes of naive reward models and demonstrate how causally-inspired approaches can improve model robustness. Finally, we outline desiderata for future research and practices, advocating targeted interventions to address inherent limitations of observational data.} }
Endnote
%0 Conference Paper %T Preference Learning for AI Alignment: a Causal Perspective %A Kasia Kobalczyk %A Mihaela Van Der Schaar %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-kobalczyk25a %I PMLR %P 31063--31083 %U https://proceedings.mlr.press/v267/kobalczyk25a.html %V 267 %X Reward modelling from preference data is a crucial step in aligning large language models (LLMs) with human values, requiring robust generalisation to novel prompt-response pairs. In this work, we propose to frame this problem in a causal paradigm, providing the rich toolbox of causality to identify the persistent challenges, such as causal misidentification, preference heterogeneity, and confounding due to user-specific factors. Inheriting from the literature of casual inference, we identify key assumptions necessary for reliable generalisation and contrast them with common data collection practices. We illustrate failure modes of naive reward models and demonstrate how causally-inspired approaches can improve model robustness. Finally, we outline desiderata for future research and practices, advocating targeted interventions to address inherent limitations of observational data.
APA
Kobalczyk, K. & Van Der Schaar, M.. (2025). Preference Learning for AI Alignment: a Causal Perspective. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:31063-31083 Available from https://proceedings.mlr.press/v267/kobalczyk25a.html.

Related Material