Double Machine Learning for Causal Inference under Shared-State Interference

Authors: Chris Hays, Manish Raghavan

ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In each instantiation, we provide simulations validating that our method produces estimators that concentrate around the true treatment effects in finite samples. We also show our consistent variance estimators can be used to construct confidence intervals with the desired coverage probability.
Researcher Affiliation Academia Chris Hays 1 Manish Raghavan 1 1MIT. Correspondence to: Chris Hays <EMAIL>.
Pseudocode Yes Algorithm 1 DML for shared-state interference
Open Source Code Yes Full details of the simulations are available in Appendix F and the code used to generate them is available at github.com/johnchrishays/dml4ssi.
Open Datasets No In our simulations, we generated data according to a smooth function of both Dt, Xt and of Ht. For all machine learning predictors ˆf, ˆm and f , we used random forests with default parameters trained on auxiliary data of size T sampled independently of the data used for inference.
Dataset Splits No The paper generates synthetic data for simulations and does not use or describe standard training/test/validation splits for a fixed dataset.
Hardware Specification No The paper does not explicitly mention any specific hardware (e.g., GPU models, CPU types, cloud instances) used for running the experiments.
Software Dependencies No For all machine learning predictors ˆf, ˆm and f , we used random forests with default parameters trained on auxiliary data of size T sampled independently of the data used for inference.
Experiment Setup No The paper mentions using 'random forests with default parameters' for machine learning predictors, but does not provide specific hyperparameter values (e.g., learning rate, batch size) for these models. It provides parameters for the synthetic data generation and simulation length, but not model training hyperparameters.