RESCIENCEC Replication / ML Reproducibility Challenge 2021 [Re] Replication study of ’Data-Driven Methods for Balancing Fairness and Effciency in Ride-Pooling’ Vera Neplenbroek 1,2, ID , Sabijn Perdijk 1,2, ID , and Victor Prins 1,2, ID 1 University of Amsterdam, Amsterdam, the Netherlands – 2 Equal contributions Edited by Koustuv Sinha, Sharath Chandra Raparthy Reviewed by Anonymous Reviewers Received 04 February 2022 Published 23 May 2022 DOI 10.5281/zenodo.6574683 Reproducibility Summary Scope of Reproducibility We evaluate the following claims related to fairness‐based objective functions presented in [1]: (1) For the four objective functions, the success rate in the worst‐served neighbor‐ hood increases monotonically with respect to the overall success rate. (2) The proposed objective functions do not lead to a higher income for the lowest‐earning drivers, nor a higher total income, compared to a request‐maximizing objective function. (3) The driver‐side fairness objective can outperform a request‐maximizing objective in terms of overall success rate and success rate in the worst‐served neighborhood. This means that this objective, whilst reducing the spread of income, also positively impacts rider fairness and proftability. Methodology The code provided by [1] was used as a base for our re‐implementation in PyTorch. We evaluate the claims by the original authors by (a) replicating their experiments, (b) test‐ ing for sensitivity to a diferent value estimator, (c) examining sensitivity to changes in the preprocessing method, and (d) testing for generalizability by applying their method to a diferent dataset. Results We reproduced the frst claim since we observed the same monotonic increase of the success rate in the worst‐served neighborhood with respect to the overall success rate. The second claim we did not reproduce, since we found that the driver‐side fairness ob‐ jective function obtains a higher income for the lowest‐earning drivers than the request‐ maximizing objective function. We reproduced the third claim, since the driver‐side objective function performs best in terms of overall success rate and success rate in the worst‐served neighborhood, and also reduces the spread of income. Changes of the value estimator, preprocessing method and even dataset all led to consistent results re‐ garding these claims. Copyright © 2022 V. Neplenbroek, S. Perdijk and V. Prins, released under a Creative Commons Attribution 4.0 International license. Correspondence should be addressed to Vera Neplenbroek (vera.neplenbroek@student.auc.nl) The authors have declared that no competing interests exist. Code is available at https://github.com/Veranep/rideshare-replication – DOI 10.5281/zenodo.6501799. – SWH swh:1:dir:f5439c1a7a15c4eb709da6f32eb252679a1d44bd. Data is available at https://www1.nyc.gov/site/tlc/about/tlc-trip-record-data.page. Open peer review is available at https://openreview.net/forum?id=BEhgn2zm3CK. ReScience C 8.2 (#29) – Neplenbroek, Perdijk and Prins 2022 1