FairMind: Automated Causal Fairness Analysis via LLM Reports

A new software prototype called FairMind automates fairness analysis at the dataset level using causal inference and large language models. Developed by researchers referencing the standard fairness model of Plečko and Bareinboim, FairMind computes counterfactual causal effects to evaluate bias involving protected features, confounders, and mediators. After preprocessing, it generates closed-form effect calculations and uses LLMs to produce readable fairness reports. The tool addresses a gap in most AutoML frameworks, which typically ignore fairness in training data and predictions. The work is described in arXiv preprint 2604.27011.