Do Real-World Datasets Contain Natural Experiments? An Empirical Study Using Causal Feature Selection
The study investigates whether real-world datasets contain natural experiments, which are implicit interventions affecting certain groups. The authors utilize causal discovery to analyze datasets and determine if treating them as interventional improves model performance. Their findings suggest that real-world datasets do indeed contain natural experiments, offering a new avenue for enhancing causal inference in data analysis.
- ▪Natural experiments are events that affect some individuals or groups but not others, serving as implicit interventions.
- ▪The study employs causal discovery to recover causal graphs and perform feature selection based on causal links.
- ▪Results indicate that real-world datasets contain natural experiments, which can improve model performance when treated as interventional.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2606.03251 (cs) COVID-19 e-print Important: e-prints posted on arXiv are not peer-reviewed by arXiv; they should not be relied upon without context to guide clinical practice or health-related behavior and should not be reported in news media as established information without consulting multiple experts in the field. [Submitted on 2 Jun 2026] Title:Do Real-World Datasets Contain Natural Experiments? An Empirical Study Using Causal Feature Selection Authors:Gautam Gare, John Galeotti, Michael Mozer, Deva Ramanan, Nan Rosemary Ke View a PDF of the paper titled Do Real-World Datasets Contain Natural Experiments? An Empirical Study Using Causal Feature Selection, by Gautam Gare and 4 other authors View PDF Abstract:In nature, events that…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
