Smoke and mirrors in causal downstream tasks

conference paper Smoke and mirrors in causal downstream tasks Advances in Neural Information Processing Systems published yes Riccardo Cadei author 0fa8b76f-72f0-11ef-b75a-a5da96e5ad6b Lukas Lindorfer author 85f0e6d3-06b3-11ec-8982-8c5049fa4455 Sylvia Cremer author 2F64EC8C-F248-11E8-B48F-1D18A9856A870000-0002-2193-3868 Cordelia Schmid author Francesco Locatello author 26cfd52f-2483-11ee-8040-88983bcc06d40000-0002-4850-0683 SyCr department FrLo department NeurIPS: Neural Information Processing Systems Machine Learning and AI have the potential to transform data-driven scientific discovery, enabling accurate predictions for several scientific phenomena. As many scientific questions are inherently causal, this paper looks at the causal inference task of treatment effect estimation, where the outcome of interest is recorded in high-dimensional observations in a Randomized Controlled Trial (RCT). Despite being the simplest possible causal setting and a perfect fit for deep learning, we theoretically find that many common choices in the literature may lead to biased estimates. To test the practical impact of these considerations, we recorded ISTAnt, the first real-world benchmark for causal inference downstream tasks on high-dimensional observations as an RCT studying how garden ants (Lasius neglectus) respond to microparticles applied onto their colony members by hygienic grooming. Comparing 6 480 models fine-tuned from state-of-the-art visual backbones, we find that the sampling and modeling choices significantly affect the accuracy of the causal estimate, and that classification accuracy is not a proxy thereof. We further validated the analysis, repeating it on a synthetically generated visual data set controlling the causal model. Our results suggest that future benchmarks should carefully consider real downstream scientific questions, especially causal ones. Further, we highlight guidelines for representation learning methods to help answer causal questions in the sciences. Code: https://github.com/CausalLearningAI/ISTAnt Data: https://doi.org/10.6084/m9.figshare.26484934.v2 Neural Information Processing Systems Foundation2024Vancouver, Canada eng 38th Conference on Neural Information Processing Systems 1049-5258 2405.1715110.52202/079017-0820 37 https://research-explorer.ista.ac.at/record/18847 https://github.com/CausalLearningAI/ISTAnt Cadei, Riccardo, Lukas Lindorfer, Sylvia Cremer, Cordelia Schmid, and Francesco Locatello. “Smoke and Mirrors in Causal Downstream Tasks.” In 38th Conference on Neural Information Processing Systems, Vol. 37. Neural Information Processing Systems Foundation, 2024. <a href="https://doi.org/10.52202/079017-0820">https://doi.org/10.52202/079017-0820</a>. R. Cadei, L. Lindorfer, S. Cremer, C. Schmid, and F. Locatello, “Smoke and mirrors in causal downstream tasks,” in 38th Conference on Neural Information Processing Systems, Vancouver, Canada, 2024, vol. 37. R. Cadei, L. Lindorfer, S. Cremer, C. Schmid, F. Locatello, in:, 38th Conference on Neural Information Processing Systems, Neural Information Processing Systems Foundation, 2024. Cadei, Riccardo, et al. “Smoke and Mirrors in Causal Downstream Tasks.” 38th Conference on Neural Information Processing Systems, vol. 37, Neural Information Processing Systems Foundation, 2024, doi:<a href="https://doi.org/10.52202/079017-0820">10.52202/079017-0820</a>. Cadei R, Lindorfer L, Cremer S, Schmid C, Locatello F. 2024. Smoke and mirrors in causal downstream tasks. 38th Conference on Neural Information Processing Systems. NeurIPS: Neural Information Processing Systems, Advances in Neural Information Processing Systems, vol. 37. Cadei R, Lindorfer L, Cremer S, Schmid C, Locatello F. Smoke and mirrors in causal downstream tasks. In: 38th Conference on Neural Information Processing Systems. Vol 37. Neural Information Processing Systems Foundation; 2024. doi:<a href="https://doi.org/10.52202/079017-0820">10.52202/079017-0820</a> Cadei, R., Lindorfer, L., Cremer, S., Schmid, C., & Locatello, F. (2024). Smoke and mirrors in causal downstream tasks. In 38th Conference on Neural Information Processing Systems (Vol. 37). Vancouver, Canada: Neural Information Processing Systems Foundation. <a href="https://doi.org/10.52202/079017-0820">https://doi.org/10.52202/079017-0820</a> 195092025-04-06T22:01:32Z2026-10-21T06:45:33Z