This paper presents FACEGroup, a graph-based framework for generating group counterfactual explanations for group fairness audits. FACEGroup models real-world feasibility constraints, identifies subgroups with similar counterfactual explanations, and captures key trade-offs in generating counterfactual explanations. Distinct from existing methods, FACEGroup introduces novel metrics for group- and subgroup-level analysis to assess fairness. Experiments on benchmark datasets demonstrate that FACEGroup effectively generates feasible group counterfactual explanations while accounting for trade-offs, and that the proposed metrics capture and quantify fairness imbalances.
