Omitted Labels Induce Nontransitive Paradoxes in Causality

Bijan Mazaheri, Siddharth Jain, Matthew Cook, Jehoshua Bruck
Proceedings of the Fourth Conference on Causal Learning and Reasoning, PMLR 275:818-833, 2025.

Abstract

We explore "omitted label contexts," in which training data is limited to a subset of the possible labels. This setting is standard among specialized human experts or specific focused studies. By studying Simpson’s paradox, we observe that "correct" adjustments sometimes require non-exchangeable treatment and control groups. A generalization of Simpson’s paradox leads us to study networks of conclusions drawn from different contexts, within which a paradox of nontransitivity arises. We prove that the space of possible nontransitive structures in these networks exactly corresponds to structures that form from aggregating ranked-choice votes.

Cite this Paper

BibTeX
@InProceedings{pmlr-v275-mazaheri25a, title = {Omitted Labels Induce Nontransitive Paradoxes in Causality}, author = {Mazaheri, Bijan and Jain, Siddharth and Cook, Matthew and Bruck, Jehoshua}, booktitle = {Proceedings of the Fourth Conference on Causal Learning and Reasoning}, pages = {818--833}, year = {2025}, editor = {Huang, Biwei and Drton, Mathias}, volume = {275}, series = {Proceedings of Machine Learning Research}, month = {07--09 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v275/main/assets/mazaheri25a/mazaheri25a.pdf}, url = {https://proceedings.mlr.press/v275/mazaheri25a.html}, abstract = {We explore "omitted label contexts," in which training data is limited to a subset of the possible labels. This setting is standard among specialized human experts or specific focused studies. By studying Simpson’s paradox, we observe that "correct" adjustments sometimes require non-exchangeable treatment and control groups. A generalization of Simpson’s paradox leads us to study networks of conclusions drawn from different contexts, within which a paradox of nontransitivity arises. We prove that the space of possible nontransitive structures in these networks exactly corresponds to structures that form from aggregating ranked-choice votes.} }
Endnote
%0 Conference Paper %T Omitted Labels Induce Nontransitive Paradoxes in Causality %A Bijan Mazaheri %A Siddharth Jain %A Matthew Cook %A Jehoshua Bruck %B Proceedings of the Fourth Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2025 %E Biwei Huang %E Mathias Drton %F pmlr-v275-mazaheri25a %I PMLR %P 818--833 %U https://proceedings.mlr.press/v275/mazaheri25a.html %V 275 %X We explore "omitted label contexts," in which training data is limited to a subset of the possible labels. This setting is standard among specialized human experts or specific focused studies. By studying Simpson’s paradox, we observe that "correct" adjustments sometimes require non-exchangeable treatment and control groups. A generalization of Simpson’s paradox leads us to study networks of conclusions drawn from different contexts, within which a paradox of nontransitivity arises. We prove that the space of possible nontransitive structures in these networks exactly corresponds to structures that form from aggregating ranked-choice votes.
APA
Mazaheri, B., Jain, S., Cook, M. & Bruck, J.. (2025). Omitted Labels Induce Nontransitive Paradoxes in Causality. Proceedings of the Fourth Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 275:818-833 Available from https://proceedings.mlr.press/v275/mazaheri25a.html.

Related Material