Dataset for: A fairness-enhanced human-in-the-loop framework based on Günther Anders‘ technical critical theory

Dataset: A fairness-enhanced human-in-the-loop framework based on Günther Anders' technical critical theory OverviewThis dataset supports the research paper "A fairness-enhanced human-in-the-loop framework based on Günther Anders' technical critical theory" (Tian & Wang, under review). It contains a simulated agricultural resource allocation dataset of 10,000 farmer records, designed to evaluate algorithmic fairness interventions in high-risk AI decision-making systems.Research Context and PurposeThe dataset was created to simulate a common "poverty trap" phenomenon in agricultural subsidy allocation, where historical bias causes underdeveloped regions to receive systematically fewer resources. The purpose is to provide a controlled benchmark for testing fairness-enhancing mechanisms in human-in-the-loop (HITL) systems, specifically the proposed FE-HITL framework that addresses three theoretical diagnoses from Günther Anders' technical critical theory: human obsolescence, Promethean shame, and technological imperialism. Data Collection and Generation MethodologyThe data were artificially generated using Python (v3.9) with the following methodology: Base generation: Farmer attributes were randomly sampled from realistic distributions based on agricultural statistics from developing regions.Bias injection: To simulate historical structural bias, farmers from Region A (the underdeveloped region) receive 30% less resource allocation than objectively equivalent farmers from Regions B and C.Feedback loop simulation: Historical yield records for Region A farmers were correspondingly reduced proportional to the allocation deficit, creating a self-reinforcing disadvantage cycle.Normalization: All numerical features were normalized to zero mean and unit variance for machine learning compatibility. Generation date: January–February 2026Generating software: Python 3.9 with numpy, pandas, scikit-learnCode availability: The generation script is available in the accompanying GitHub repository: https://github.com/YingWang08/anders-fairness-hitl Data Structure and Variables File formatFilename: `agricultural_dataset.csv`Format: UTF-8 encoded CSV, comma-delimitedRows: 10,000 data rows + 1 header rowColumns: 8 Variable descriptions | Variable Name | Type | Description | Values / Units ||---------------|------|-------------|-----------------|| `land_area` | float | Arable land area owned by the farmer | Hectares, range 0.5–10.0 || `labor_force` | int | Number of workers in the household | Integer, 1–6 || `farming_years` | int | Years of farming experience | Integer, 1–40 || `avg_yield` | float | Average crop yield over past three years | Tons/hectare, range 0.5–8.0 || `irrigation` | float | Irrigation resource index | 0–100 scale (higher = better irrigation) || `fertilizer_subsidy` | float | Fertilizer subsidy received in past year | Monetary units, continuous || `region` | string | Geographical region | A (underdeveloped), B, C || `allocation` | float | Resource allocation amount (target variable) | Monetary units, continuous | Missing data handlingThe dataset contains no missing values. All fields are complete. Data SplitTo ensure reproducibility of experimental results, the dataset is pre-split with a fixed random seed (42): Training set: 7,000 samples (70%)Validation set: 1,500 samples (15%)Test set: 1,500 samples (15%) The split indices are consistent with the random seed used in the accompanying code repository. Quality Control- Validated that bias injection mechanism correctly produced the intended disparate impact (DI ≈ 0.58 in baseline models)- Verified no impossible values (e.g., negative land area, yield outside plausible range)- Cross-checked summary statistics against real agricultural data from FAO reports for ecological validity Ethical ConsiderationsThis dataset is entirely simulated and contains no real human subject data. No personal or sensitive information is included. The simulation intentionally models structural bias to enable research on fairness interventions; it does not represent any specific real-world population or region. LicenseThis dataset is released under the **MIT License**. You are free to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the data, provided that the original authors are credited. A copy of the license is included in the repository. Related MaterialsGitHub repository (source code): [https://github.com/YingWang08/anders-fairness-hitl](https://github.com/YingWang08/anders-fairness-hitl)Research paper: Tian, W. & Wang, Y. (2026). A fairness-enhanced human-in-the-loop framework based on Günther Anders' technical critical theory. *PLOS ONE* (under review) CitationIf you use this dataset in your research, please cite:Tian, Wenjun; Wang, Ying (2026). Dataset: A fairness-enhanced human-in-the-loop framework based on Günther Anders' technical critical theory. Figshare.https://doi.org/10.6084/m9.figshare.31672465 ContactFor questions, issues, or collaboration inquiries, please contact: Wenjun Tian (corresponding author)School of Marxism, Northeastern UniversityEmail: Wenjun.Tian@xxx.edu Ying WangSchool of Computer Science and Engineering, Beijing University of AgricultureEmail: [wangying-314@cpic.com.cn]