Causal Fairness for Outcome Control

Authors: Drago Plecko, Elias Bareinboim

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We apply the causal framework of outcome control to the problem of allocating mechanical ventilation in intensive care units (ICUs)...To investigate this issue using the tools developed in this paper, we use the data from the MIMIC-IV dataset [17, 16]...The learning rate was fixed at η = 0.1, and the optimal number of rounds was chosen via 10-fold cross-validation. We then use the obtained model to generate predictions...The results for the probability of treatment given a fixed decile are shown in Fig. 6b.
Researcher Affiliation Academia Drago Plecko and Elias Bareinboim Department of Computer Science Columbia University dp3144@columbia.edu, eb@cs.columbia.edu
Pseudocode Yes Algorithm 1 Decision-Making with Benefit Fairness; Algorithm 2 Benefit Fairness Causal Explanation; Algorithm 3 Causal Discrimination Removal for Outcome Control
Open Source Code Yes The source code for reproducing all the experiments can be found in our code repository. Furthermore, the vignette accompanying the main text can be found here.
Open Datasets Yes To investigate this issue using the tools developed in this paper, we use the data from the MIMIC-IV dataset [17, 16] that originates from the Beth Israel Deaconess Medical Center in Boston, Massachusetts...[17] A. E. Johnson, L. Bulgarelli, L. Shen, A. Gayles, A. Shammout, S. Horng, T. J. Pollard, B. Moody, B. Gow, L.-w. H. Lehman, et al. Mimic-iv, a freely accessible electronic health record dataset. Scientific data, 10(1):1, 2023.
Dataset Splits Yes The learning rate was fixed at η = 0.1, and the optimal number of rounds was chosen via 10-fold cross-validation.
Hardware Specification No The paper does not mention any specific hardware specifications (e.g., CPU, GPU models, memory, cloud instances) used for running the experiments.
Software Dependencies No The paper mentions using an
Experiment Setup Yes We fit an xgboost model which regresses Y on D, X, Z, and W, to obtain the fit b Y . The learning rate was fixed at η = 0.1, and the optimal number of rounds was chosen via 10-fold cross-validation.