Reproducibility study of "Learning Decision Trees and Forests with Algorithmic Recourse"
Abstract: Decision trees and random forests are widely recognized machine learning models, particu-
larly for their interpretability. However, ensuring algorithmic recourse—providing individ-
uals with actionable steps to alter model predictions—remains a significant challenge. The
authors of the paper Learning Decision Trees and Forests with Algorithmic Re-
course (Kanamori et al. (2024)) introduce a novel method for training tree-based models
while guaranteeing the existence of recourse actions. In this study, we attempt to repli-
cate the original findings and validate their data using the open-source implementation and
datasets provided in the original paper. While we observe some differences in the per-
formance of sensitivity forests, we confirm that our results closely align with those of the
decision trees presented in the original study.
Submission Length: Regular submission (no more than 12 pages of main content)
Assigned Action Editor: ~Satoshi_Hara1
Submission Number: 4252
Loading