Data for Causal Mediation Analysis

This dataset was generated to support causal and mediation-based fairness analysis in hiring decisions, inspired by the framework presented in Binkytė et al., 2022, arXiv:2207.04053. It simulates a binary hiring outcome Y based on a sensitive attribute A (Political Belief), two mediators M1 , M2 (Community Service, Address), and a confounder C (Socio-Economic Status). The data is designed to allow for decomposing direct and indirect discrimination pathways. The accompanying GitHub Repository contains the full code for dataset generation and detailed analysis, including: Construction of a causal graph Evaluation of statistical parity and causal fairness metrics Mediation analysis via counterfactual modeling Use Cases: This dataset can be used to evaluate fairness-aware algorithms, study mediation effects, or demonstrate causal inference techniques in a controlled, interpretable environment. Keywords: fairness, causality, mediation analysis, synthetic data, AI ethics