Causal Machine Learning for Explainable and Interpretable Artificial Intelligence
DOI:
https://doi.org/10.63345/Keywords:
causal machine learning, explainable AI, interpretability, counterfactual explanations, causal inference, Shapley valuesAbstract
Post hoc explanation methods describe how a model’s output co-varies with its inputs, which is not the same as how the world responds to a change. This paper examines causal machine learning as a foundation for explainable and interpretable AI. It reviews the limits of associational attribution, surveys methods for causal discovery, effect estimation, counterfactual generation and causal attribution, and proposes CausalXAI, a modular pipeline that connects them. In a controlled simulation with known ground truth, standard permutation importance and Shapley-style attributions placed 29–41% of their importance on a feature with no causal effect, whereas the causal pipeline placed none, and its effect-estimation error was 0.027 against 0.54 at 2,000 training samples. Under a shift in the spurious feature’s behaviour, the causal-parent model held a constant error while the all-feature model’s error rose 2.7-fold. We discuss assumptions, limitations and ethical considerations.References
[1] C. Rudin, “Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead,” Nature Machine Intelligence, vol. 1, no. 5, pp. 206–215, 2019.
[2] F. Doshi-Velez and B. Kim, “Towards a rigorous science of interpretable machine learning,” arXiv:1702.08608, 2017.
[3] M. T. Ribeiro, S. Singh, and C. Guestrin, “‘Why should I trust you?’: Explaining the predictions of any classifier,” in Proc. 22nd ACM SIGKDD Int. Conf. Knowledge Discovery and Data Mining, 2016, pp. 1135–1144.
[4] S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” in Advances in Neural Information Processing Systems 30, 2017, pp. 4765–4774.
[5] J. Adebayo, J. Gilmer, M. Muelly, I. Goodfellow, M. Hardt, and B. Kim, “Sanity checks for saliency maps,” in Advances in Neural Information Processing Systems 31, 2018, pp. 9505–9515.
[6] J. Pearl, Causality: Models, Reasoning, and Inference, 2nd ed. Cambridge, U.K.: Cambridge Univ. Press, 2009.
[7] B. Schölkopf, F. Locatello, S. Bauer, N. R. Ke, N. Kalchbrenner, A. Goyal, and Y. Bengio, “Toward causal representation learning,” Proc. IEEE, vol. 109, no. 5, pp. 612–634, 2021.
[8] R. Moraffah, M. Karami, R. Guo, A. Raglin, and H. Liu, “Causal interpretability for machine learning – problems, methods and evaluation,” ACM SIGKDD Explorations Newsletter, vol. 22, no. 1, pp. 18–33, 2020.
[9] J. Kaddour, A. Lynch, Q. Liu, M. J. Kusner, and R. Silva, “Causal machine learning: A survey and open problems,” arXiv:2206.15475, 2022.
[10] D. B. Rubin, “Estimating causal effects of treatments in randomized and nonrandomized studies,” J. Educational Psychology, vol. 66, no. 5, pp. 688–701, 1974.
[11] M. A. Hernán and J. M. Robins, Causal Inference: What If. Boca Raton, FL, USA: Chapman & Hall/CRC, 2020.
[12] E. Bareinboim, J. D. Correa, D. Ibeling, and T. Icard, “On Pearl’s hierarchy and the foundations of causal inference,” in Probabilistic and Causal Inference: The Works of Judea Pearl. New York, NY, USA: ACM, 2022.
[13] J. Peters, D. Janzing, and B. Schölkopf, Elements of Causal Inference: Foundations and Learning Algorithms. Cambridge, MA, USA: MIT Press, 2017.
[14] D. Janzing, L. Minorics, and P. Blöbaum, “Feature relevance quantification in explainable AI: A causal problem,” in Proc. 23rd Int. Conf. Artificial Intelligence and Statistics (AISTATS), PMLR vol. 108, 2020, pp. 2907–2916.
[15] J. Peters, P. Bühlmann, and N. Meinshausen, “Causal inference by using invariant prediction: Identification and confidence intervals,” J. Royal Statistical Society: Series B, vol. 78, no. 5, pp. 947–1012, 2016.
[16] M. Arjovsky, L. Bottou, I. Gulrajani, and D. Lopez-Paz, “Invariant risk minimization,” arXiv:1907.02893, 2019.
[17] P. Spirtes, C. Glymour, and R. Scheines, Causation, Prediction, and Search, 2nd ed. Cambridge, MA, USA: MIT Press, 2000.
[18] X. Zheng, B. Aragam, P. Ravikumar, and E. P. Xing, “DAGs with NO TEARS: Continuous optimization for structure learning,” in Advances in Neural Information Processing Systems 31, 2018.
[19] V. Chernozhukov, D. Chetverikov, M. Demirer, E. Duflo, C. Hansen, W. Newey, and J. Robins, “Double/debiased machine learning for treatment and structural parameters,” The Econometrics Journal, vol. 21, no. 1, pp. C1–C68, 2018.
[20] S. R. Künzel, J. S. Sekhon, P. J. Bickel, and B. Yu, “Metalearners for estimating heterogeneous treatment effects using machine learning,” Proc. National Academy of Sciences, vol. 116, no. 10, pp. 4156–4165, 2019.
[21] S. Athey and G. Imbens, “Recursive partitioning for heterogeneous causal effects,” Proc. National Academy of Sciences, vol. 113, no. 27, pp. 7353–7360, 2016.
[22] S. Wager and S. Athey, “Estimation and inference of heterogeneous treatment effects using random forests,” J. American Statistical Association, vol. 113, no. 523, pp. 1228–1242, 2018.
[23] U. Shalit, F. D. Johansson, and D. Sontag, “Estimating individual treatment effect: Generalization bounds and algorithms,” in Proc. 34th Int. Conf. Machine Learning (ICML), PMLR vol. 70, 2017, pp. 3076–3085.
[24] C. Louizos, U. Shalit, J. Mooij, D. Sontag, R. Zemel, and M. Welling, “Causal effect inference with deep latent-variable models,” in Advances in Neural Information Processing Systems 30, 2017.
[25] S. Wachter, B. Mittelstadt, and C. Russell, “Counterfactual explanations without opening the black box: Automated decisions and the GDPR,” Harvard J. Law & Technology, vol. 31, no. 2, pp. 841–887, 2018.
[26] R. K. Mothilal, A. Sharma, and C. Tan, “Explaining machine learning classifiers through diverse counterfactual explanations,” in Proc. Conf. Fairness, Accountability, and Transparency (FAT*), 2020, pp. 607–617.
[27] A.-H. Karimi, B. Schölkopf, and I. Valera, “Algorithmic recourse: From counterfactual explanations to interventions,” in Proc. ACM Conf. Fairness, Accountability, and Transparency (FAccT), 2021, pp. 353–362.
[28] T. Heskes, E. Sijben, I. G. Bucur, and T. Claassen, “Causal Shapley values: Exploiting causal knowledge to explain individual predictions of complex models,” in Advances in Neural Information Processing Systems 33, 2020.
[29] C. Frye, C. Rowat, and I. Feige, “Asymmetric Shapley values: Incorporating causal knowledge into model-agnostic explainability,” in Advances in Neural Information Processing Systems 33, 2020.
[30] A. Chattopadhyay, P. Manupriya, A. Sarkar, and V. N. Balasubramanian, “Neural network attributions: A causal perspective,” in Proc. 36th Int. Conf. Machine Learning (ICML), PMLR vol. 97, 2019, pp. 981–990.
[31] Y. Goyal, A. Feder, U. Shalit, and B. Kim, “Explaining classifiers with causal concept effect (CaCE),” arXiv:1907.07165, 2019.
[32] A. Sharma and E. Kiciman, “DoWhy: An end-to-end library for causal inference,” arXiv:2011.04216, 2020.
[33] I. Bica, A. M. Alaa, J. Jordon, and M. van der Schaar, “Estimating counterfactual treatment outcomes over time through adversarially balanced representations,” in Int. Conf. Learning Representations (ICLR), 2020.
[34] Y.-L. Chou, C. Moreira, P. Bruza, C. Ouyang, and J. Jorge, “Counterfactuals and causability in explainable artificial intelligence: Theory, algorithms, and applications,” Information Fusion, vol. 81, pp. 59–83, 2022.
[35] F. Pedregosa et al., “Scikit-learn: Machine learning in Python,” J. Machine Learning Research, vol. 12, pp. 2825–2830, 2011.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 The journal retains copyright of all published articles, ensuring that authors have control over their work while allowing wide dissenmination.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
Articles are published under the Creative Commons Attribution NonCommercial 4.0 License (CC BY NC 4.0), allowing others to distribute, remix, adapt, and build upon the work for non-commercial purposes while crediting the original author.
