Code, raw results, and statistical tables for: Translating Günther Anders’ Critical Theory into Algorithmic Fairness: A Simulation-Based Proof of Concept for Fairness-Enhanced Decision Support with Simulated Human Review

This item contains the source code, the simulated-data generator, the raw per-seed results, and the regenerated statistical tables underlying the article "Translating Günther Anders’ Critical Theory into Algorithmic Fairness: A Simulation-Based Proof of Concept for Fairness-Enhanced Decision Support with Simulated Human Review" (Tian W, Wang Y; submitted to PLOS One). It is a snapshot of the GitHub repository https://github.com/YingWang08/anders-fairness-hitl, the version cited in the article. About the studyThe study is a simulation-based proof of concept. All "human" decisions are generated by fixed rule-based proxies; no human participants were involved. A fairness-enhanced, HITL-inspired post-processing framework (FE-HITL) is evaluated on a simulated agricultural resource-allocation task (30 independently generated datasets of 10,000 farmers) and on the UCI German Credit benchmark (50 resampled splits and 5-fold cross-validation). Contents- experiments/run_sensitivity_standalone.py: the eight agricultural analysis suites. The simulated data are generated in memory by generate_agricultural_data_biased(), generate_agricultural_data_biased_v2(), and generate_full_dataset(); no data file is needed.- experiments/run_credit_clean.py: the German Credit experiments (data loaded from OpenML, data ID 31; a copy of the UCI data file is in data/).- experiments/run_r3_additions.py and src/fe_hitl_r3.py: read-only analyses added in the third revision (Algorithm 1 as specified, single-candidate controls, feedback-and-update diagnostics).- experiments/revision_results/ and experiments/credit_revision_results/: raw per-seed results, i.e., the data points behind all reported means, standard deviations, and tests.- regen_stats.py, inference_fixed.py, fairness_cost_joint.py: regenerate every statistical table of the article and its S1 Appendix from the raw results into r3_tables/. r3_tables/MANIFEST.json records input hashes, library versions, and statistical conventions.- requirements.txt and requirements-lock.txt: the software environment (results generated with Python 3.14.3). Reproducing the statisticspip install -r requirements.txtpython regen_stats.pypython inference_fixed.pypython fairness_cost_joint.pyThe regenerated files in r3_tables/ match the deposited ones to numerical precision. See README.md for the full run order and seed sets. Notes- The fixed-ratio post-processing baseline called FRP in the article is labelled "Debiased-HITL" in the code and data files; it is an author-designed heuristic.- Files listed as superseded in README.md, as well as data/generate_agricultural_data.py and appendix/S1_Appendix.md from the first submission, are not used by any reported result.- German Credit data: Hofmann H. Statlog (German Credit Data). UCI Machine Learning Repository; This version replaces earlier versions, which corresponded to earlier revisions of the manuscript. The code is released under the MIT License.