Meaningful Causal Aggregation and Paradoxical Confounding

Yuchen Zhu, Kailash Budhathoki, Jonas M. Kübler, Dominik Janzing
Proceedings of the Third Conference on Causal Learning and Reasoning, PMLR 236:1192-1217, 2024.

Abstract

In aggregated variables the impact of interventions is typically ill-defined because different micro-realizations of the same macro-intervention can result in different changes of downstream macro-variables. We show that this ill-definedness of causality on aggregated variables can turn unconfounded causal relations into confounded ones and vice versa, depending on the respective micro-realization. We argue that it is practically infeasible to only use aggregated causal systems when we are free from this ill-definedness. Instead, we need to accept that macro causal relations are typically defined only with reference to the micro states. On the positive side, we show that cause-effect relations can be aggregated when the macro interventions are such that the distribution of micro states is the same as in the observational distribution; we term this natural macro interventions. We also discuss generalizations of this observation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v236-zhu24a, title = {Meaningful Causal Aggregation and Paradoxical Confounding}, author = {Zhu, Yuchen and Budhathoki, Kailash and K\"ubler, Jonas M. and Janzing, Dominik}, booktitle = {Proceedings of the Third Conference on Causal Learning and Reasoning}, pages = {1192--1217}, year = {2024}, editor = {Locatello, Francesco and Didelez, Vanessa}, volume = {236}, series = {Proceedings of Machine Learning Research}, month = {01--03 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v236/zhu24a/zhu24a.pdf}, url = {https://proceedings.mlr.press/v236/zhu24a.html}, abstract = {In aggregated variables the impact of interventions is typically ill-defined because different micro-realizations of the same macro-intervention can result in different changes of downstream macro-variables. We show that this ill-definedness of causality on aggregated variables can turn unconfounded causal relations into confounded ones and vice versa, depending on the respective micro-realization. We argue that it is practically infeasible to only use aggregated causal systems when we are free from this ill-definedness. Instead, we need to accept that macro causal relations are typically defined only with reference to the micro states. On the positive side, we show that cause-effect relations can be aggregated when the macro interventions are such that the distribution of micro states is the same as in the observational distribution; we term this natural macro interventions. We also discuss generalizations of this observation.} }
Endnote
%0 Conference Paper %T Meaningful Causal Aggregation and Paradoxical Confounding %A Yuchen Zhu %A Kailash Budhathoki %A Jonas M. Kübler %A Dominik Janzing %B Proceedings of the Third Conference on Causal Learning and Reasoning %C Proceedings of Machine Learning Research %D 2024 %E Francesco Locatello %E Vanessa Didelez %F pmlr-v236-zhu24a %I PMLR %P 1192--1217 %U https://proceedings.mlr.press/v236/zhu24a.html %V 236 %X In aggregated variables the impact of interventions is typically ill-defined because different micro-realizations of the same macro-intervention can result in different changes of downstream macro-variables. We show that this ill-definedness of causality on aggregated variables can turn unconfounded causal relations into confounded ones and vice versa, depending on the respective micro-realization. We argue that it is practically infeasible to only use aggregated causal systems when we are free from this ill-definedness. Instead, we need to accept that macro causal relations are typically defined only with reference to the micro states. On the positive side, we show that cause-effect relations can be aggregated when the macro interventions are such that the distribution of micro states is the same as in the observational distribution; we term this natural macro interventions. We also discuss generalizations of this observation.
APA
Zhu, Y., Budhathoki, K., Kübler, J.M. & Janzing, D.. (2024). Meaningful Causal Aggregation and Paradoxical Confounding. Proceedings of the Third Conference on Causal Learning and Reasoning, in Proceedings of Machine Learning Research 236:1192-1217 Available from https://proceedings.mlr.press/v236/zhu24a.html.

Related Material