Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Beyond Individualized Recourse: Interpretable and Interactive Summaries of Actionable Recourses
Authors: Kaivalya Rawal, Himabindu Lakkaraju
NeurIPS 2020 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental evaluation with real world datasets and user studies demonstrate that our framework can provide decision makers with a comprehensive overview of recourses corresponding to any black box model, and consequently help detect undesirable model biases and discrimination. |
| Researcher Affiliation | Academia | Kaivalya Rawal Harvard University EMAIL; Himabindu Lakkaraju Harvard University EMAIL |
| Pseudocode | Yes | Pseudocode for this procedure is provided in Appendix. |
| Open Source Code | No | The paper does not provide an explicit statement or link for the open-source code of the methodology described. |
| Open Datasets | Yes | Our first dataset is the COMPAS dataset which was collected by Pro Publica [18]. Our second dataset is the German Credit dataset [9]. Our third dataset is the bail decisions dataset [16]. |
| Dataset Splits | No | We split our datasets randomly into train (50%) and test sets (50%). The paper does not specify a validation set split. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running experiments. |
| Software Dependencies | No | The paper mentions using various machine learning models and baselines (e.g., LIME, SHAP, FACE, AR) but does not provide specific version numbers for software dependencies or libraries used. |
| Experiment Setup | Yes | We employ a simple tuning procedure to set the λ parameters (details in Appendix) and the constraint values are assigned as ϵ1 = 20, ϵ2 = 7, and ϵ3 = 10. Support threshold for Apriori rule mining algorithm is set to 1%. |