Derivative Causal Models: Modeling Causality at Mixed Scales of Observation

27 Sept 2024 (modified: 22 Nov 2024)ICLR 2025 Conference Withdrawn SubmissionEveryoneRevisionsBibTeXCC BY 4.0
Keywords: derivative causal models, causal modeling, constraint causal models
TL;DR: We propose Derivative Causal Models, a novel compact causal representation that handles mixed scales of observables in real-world systems.
Abstract:

Causal relations can materialize in many different ways. In their most simple form --typically assumed in classical causal models and discovery approaches--, similar variations of a cause lead to similar variations of an effect. However, this `smoothness' requires an observation of cause and effect just at the right scales. Unfortunately, this conflicts with records often encountered in the real-world, mixing continuous measurements with once-in-a-while observations of sparse events. Compactly modeling the causal effects between (discrete) events and continuous states is hard to achieve with classical causal models. To ease this situation, we leverage transformations that derive different scales of observables, respectively, to decompose relations and allow for compact causal representations, called Derivative Causal Models (DCM). We instantiate them using integral and derivative transforms and demonstrate that the resulting Differential Causal Models ($\partial$CM) can be discovered automatically from data.

Primary Area: causal reasoning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 9470
Loading

OpenReview is a long-term project to advance science through improved peer review with legal nonprofit status. We gratefully acknowledge the support of the OpenReview Sponsors. © 2025 OpenReview