Towards Trustworthy Explanation: On Causal Rationalization

Wenbo Zhang, Tong Wu, Yunlong Wang, Yong Cai, Hengrui Cai
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:41715-41736, 2023.

Abstract

With recent advances in natural language processing, rationalization becomes an essential self-explaining diagram to disentangle the black box by selecting a subset of input texts to account for the major variation in prediction. Yet, existing association-based approaches on rationalization cannot identify true rationales when two or more snippets are highly inter-correlated and thus provide a similar contribution to prediction accuracy, so-called spuriousness. To address this limitation, we novelly leverage two causal desiderata, non-spuriousness and efficiency, into rationalization from the causal inference perspective. We formally define a series of probabilities of causation based on a newly proposed structural causal model of rationalization, with its theoretical identification established as the main component of learning necessary and sufficient rationales. The superior performance of the proposed causal rationalization is demonstrated on real-world review and medical datasets with extensive experiments compared to state-of-the-art methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-zhang23ap, title = {Towards Trustworthy Explanation: On Causal Rationalization}, author = {Zhang, Wenbo and Wu, Tong and Wang, Yunlong and Cai, Yong and Cai, Hengrui}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {41715--41736}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/zhang23ap/zhang23ap.pdf}, url = {https://proceedings.mlr.press/v202/zhang23ap.html}, abstract = {With recent advances in natural language processing, rationalization becomes an essential self-explaining diagram to disentangle the black box by selecting a subset of input texts to account for the major variation in prediction. Yet, existing association-based approaches on rationalization cannot identify true rationales when two or more snippets are highly inter-correlated and thus provide a similar contribution to prediction accuracy, so-called spuriousness. To address this limitation, we novelly leverage two causal desiderata, non-spuriousness and efficiency, into rationalization from the causal inference perspective. We formally define a series of probabilities of causation based on a newly proposed structural causal model of rationalization, with its theoretical identification established as the main component of learning necessary and sufficient rationales. The superior performance of the proposed causal rationalization is demonstrated on real-world review and medical datasets with extensive experiments compared to state-of-the-art methods.} }
Endnote
%0 Conference Paper %T Towards Trustworthy Explanation: On Causal Rationalization %A Wenbo Zhang %A Tong Wu %A Yunlong Wang %A Yong Cai %A Hengrui Cai %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-zhang23ap %I PMLR %P 41715--41736 %U https://proceedings.mlr.press/v202/zhang23ap.html %V 202 %X With recent advances in natural language processing, rationalization becomes an essential self-explaining diagram to disentangle the black box by selecting a subset of input texts to account for the major variation in prediction. Yet, existing association-based approaches on rationalization cannot identify true rationales when two or more snippets are highly inter-correlated and thus provide a similar contribution to prediction accuracy, so-called spuriousness. To address this limitation, we novelly leverage two causal desiderata, non-spuriousness and efficiency, into rationalization from the causal inference perspective. We formally define a series of probabilities of causation based on a newly proposed structural causal model of rationalization, with its theoretical identification established as the main component of learning necessary and sufficient rationales. The superior performance of the proposed causal rationalization is demonstrated on real-world review and medical datasets with extensive experiments compared to state-of-the-art methods.
APA
Zhang, W., Wu, T., Wang, Y., Cai, Y. & Cai, H.. (2023). Towards Trustworthy Explanation: On Causal Rationalization. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:41715-41736 Available from https://proceedings.mlr.press/v202/zhang23ap.html.

Related Material