Abstract
Machine learning is used more and more in scientific contexts, from the recent breakthroughs with AlphaFold2 in protein fold prediction to the use of ML in parametrization for large climate/astronomy models. Yet it is unclear whether we can obtain scientific explanations from such models. I argue that when machine learning is used to conduct causal inference we can give a new positive answer to this question. However, these ML models are purpose-built models and there are technical results showing that standard machine learning models cannot be used for the same type of causal inference. Instead, there is a pathway to causal explanations from predictive ML models through new explainability techniques; specifically, new methods to extract structural equation models from such ML models. The extracted models are likely to suffer from issues though: they will often fail to account for confounders and colliders, as well as deliver simply incorrect causal graphs due to ML models tendency to violate physical laws such as the conservation of energy. In this case, extracted graphs are a starting point for new explanations, but predictive accuracy is no guarantee for good explanations.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Data Availability
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Notes
It should be noted here that the term ’machine learning’ has both narrow and broad interpretations. In a broad interpretation it is any computer method that solves a problem by fitting a function to data. In that case, simple models such as those based on linear regression count as machine learning. I follow the narrower definition of machine learning common in the literature discussed here, where the term is only applied to methods such as deep neural networks and random forest algorithms, which are distinguished by their use of a large number of parameters and non-linearity.
To be precise, the method looks at neural networks j for each variable X (indexed 1 to d). The parameters (i.e. weights) of these neural networks are represented in vector \(\phi _{(j)}\). The maximum likelihood problem solved over all these neural networks is then the equation \(max_\phi {\mathbb {E}}_{X~P_X} \sum _{j=1}^d \log p_j(X_j | X_{\pi ^\phi _j}; \phi _{(j)})\), where \(X_{\pi ^\phi _j}\) is the set of parents of node j in graph \({\mathcal {G}}_\phi \). Essentially, the idea is that one optimized the predictive accuracy of all these neural networks together, where each neural network aims to predict the value of variable \(X_j\) in terms of the values of all the other variables.
Note that these are not, as in Sect. 3, values of treatment effects but rather are values of variables figuring in the explanation. The causal graph thus remains the same, it is only instantiated in a particular way based on the outcomes of PkANN.
References
Agarwal, S., Abdalla, F. B., Feldman, H. A., Lahav, O., & Thomas, S. A. (2012). Pkann—I. Non-linear matter power spectrum interpolation through artificial neural networks. Monthly Notices of the Royal Astronomical Society, 424(2), 1409–1418.
Baiardi, A., & Naghi, A. (2021). The value added of machine learning to causal inference: Evidence from revisited studies. arXiv preprintarXiv:2101.00878.
Batterman, R. W. (1992). Explanatory instability. Nous, 26(3), 325–348.
Beckers, S. (2022). Causal explanations and XAI. In Conference on causal learning and reasoning (pp. 90–109). PMLR.
Beckers, S., & Halpern, J. Y. (2019). Abstracting causal models. In Proceedings of the AAAI conference on artificial intelligence (Vol. 33, pp. 2678–2685).
Bellot, A., & van der Schaar, M. (2019). Conditional independence testing using generative adversarial networks. Advances in Neural Information Processing Systems, 32, 1–10.
Biswas, S., Corti, L., Buijsman, S., & Yang, J. (2022). Chime: Causal human-in-the-loop model explanations. In Proceedings of the AAAI conference on human computation and crowdsourcing (Vol. 10, pp. 27–39).
Buijsman, S. (2022). Defining explanation and explanatory depth in XAI. Minds and Machines, 32(3), 563–584.
Cao, Y., Kang, Q., Zhang, B., Zhu, Z., Dong, G., Cai, Q., Lee, K., & Chen, B. (2022). Machine learning-aided causal inference for unraveling chemical dispersant and salinity effects on crude oil biodegradation. Bioresource Technology, 345, 126468.
Caruana, R., Lou, Y., Gehrke, J., Koch, P., Sturm, M., & Elhadad, N. (2015). Intelligible models for healthcare: Predicting pneumonia risk and hospital 30-day readmission. In Proceedings of the 21th ACM SIGKDD international conference on knowledge discovery and data mining (pp. 1721–1730).
Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., & Newey, W. (2017). Double/debiased/Neyman machine learning of treatment effects. American Economic Review, 107(5), 261–265.
Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W., & Robins, J. (2018). Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21(1), C1–C68.
Das, A., & Rad, P. (2020). Opportunities and challenges in explainable artificial intelligence (XAI): A survey. arXiv preprintarXiv:2006.11371
Duncan, W. D. (2017). Ontological distinctions between hardware and software. Applied Ontology, 12(1), 5–32.
Geiger, A., Lu, H., Icard, T., & Potts, C. (2021). Causal abstractions of neural networks. Advances in Neural Information Processing Systems, 34, 9574–9586.
Geiger, A., Potts, C., & Icard, T. (2023). Causal abstraction for faithful model interpretation. arXiv preprintarXiv:2301.04709
Glymour, C., Zhang, K., & Spirtes, P. (2019). Review of causal discovery methods based on graphical models. Frontiers in Genetics, 10, 524.
Halpern, J. Y., & Pearl, J. (2005). Causes and explanations: A structural-model approach. Part II: Explanations. The British Journal for the Philosophy of Science, 56(4), 889–911.
Jebeile, J., Lam, V., & Räz, T. (2021). Understanding climate change with statistical downscaling and machine learning. Synthese, 199(1), 1877–1897.
Jumper, J., Evans, R., Pritzel, A., Green, T., Figurnov, M., Ronneberger, O., Tunyasuvunakool, K., Bates, R., Žídek, A., Potapenko, A., et al. (2021). Highly accurate protein structure prediction with alphafold. Nature, 596(7873), 583–589.
Kalainathan, D., Goudet, O., Guyon, I., Lopez-Paz, D., & Sebag, M. (2018). Structural agnostic modeling: Adversarial learning of causal graphs. arXiv preprintarXiv:1803.04929
Kawamleh, S. (2021). Can machines learn how clouds work? The epistemic implications of machine learning methods in climate science. Philosophy of Science, 88(5), 1008–1020.
Knüsel, B., & Baumberger, C. (2020). Understanding climate phenomena with data-driven models. Studies in History and Philosophy of Science Part A, 84, 46–56.
Lachapelle, S., Brouillard, P., Deleu, T., & Lacoste-Julien, S. (2019). Gradient-based neural dag learning. arXiv preprintarXiv:1906.02226
López-Rubio, E., & Ratti, E. (2021). Data science and molecular biology: Prediction and mechanistic explanation. Synthese, 198(4), 3131–3156.
Meskhidze, H. (2023). Can machine learning provide understanding? How cosmologists use machine learning to understand observations of the universe. Erkenntnis, 88, 1895–1909.
Milkowski, M. (2013). Explaining the computational mind. MIT Press.
Pearl, J. (2009). Causal inference in statistics: An overview. Statistics Surveys, 3, 96–146.
Piccinini, G. (2010). The mind as neural software? understanding functionalism, computationalism, and computational functionalism. Philosophy and Phenomenological Research, 81(2), 269–311.
Pietsch, W. (2016). The causal nature of modeling with big data. Philosophy & Technology, 29, 137–171.
Rasp, S., Pritchard, M. S., & Gentine, P. (2018). Deep learning to represent subgrid processes in climate models. Proceedings of the National Academy of Sciences, 115(39), 9684–9689.
Räz, T., & Beisbart, C. (2022). The importance of understanding deep learning. Erkenntnis, 1–18.
Schmidt, J., Marques, M. R., Botti, S., & Marques, M. A. (2019). Recent advances and applications of machine learning in solid-state materials science. npj Computational Materials, 5(1), 1–36.
Sen, R., Suresh, A. T., Shanmugam, K., Dimakis, A. G., & Shakkottai, S. (2017). Model-powered conditional independence test. In Advances in neural information processing systems, 30, 1–11.
Shah, R. D., & Peters, J. (2020). The hardness of conditional independence testing and the generalised covariance measure. The Annals of Statistics, 48(3), 1514–1538.
Shi, C., Xu, T., Bergsma, W., & Li, L. (2020). Double generative adversarial networks for conditional independence testing. arXiv preprintarXiv:2006.02615
Spirtes, P., Glymour, C. N., Scheines, R., & Heckerman, D. (2000). Causation, prediction, and search. MIT Press.
Srećković, S., Berber, A., & Filipović, N. (2021). The automated Laplacean demon: How ML challenges our views on prediction and explanation. Minds and Machines, 32, 159–183.
Stinson, C. (2018). Explanation and connectionist models. In The Routledge handbook of the computational mind. Routledge.
Sullivan, E. (2019). Understanding from machine learning models. The British Journal for the Philosophy of Science, 73(1), 109–133.
Turner, R. (2011). Specification. Minds and Machines, 21, 135–152.
Woodward, J. (2005). Making things happen: A theory of causal explanation. Oxford University Press.
Wu, Z., D’Oosterlinck, K., Geiger, A., Zur, A., & Potts, C. (2023). Causal proxy models for concept-based model explanations. In International conference on machine learning (pp. 37313–37334). PMLR.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author(s) declare that there are no conflicts of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Buijsman, S. Causal scientific explanations from machine learning. Synthese 202, 202 (2023). https://doi.org/10.1007/s11229-023-04429-3
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11229-023-04429-3