The study introduces an AI-enhanced Environmental, Social, and Governance (ESG) framework that integrates explainable artificial intelligence (XAI) and bias-mitigation techniques to improve transparency and comparability of sustainability assessments across sectors. Addressing the persistent gap in standardized ESG evaluation methods, the framework combines gradient-boosting models (XGBoost) with SHAP-based explainability and a human-in-the-loop (HITL) validation layer. The approach is demonstrated using ESG indicators for 18 firms across three industries from the banking, aviation, and chemical sectors between 2021 and 2023. Results indicate an average 12.4% improvement in ESG-score consistency and a 9% reduction in inter-sector variance relative to baseline traditional ESG evaluations. Fairness metrics (Disparate-Impact Ratio = 0.81–0.86) provide preliminary evidence of improved alignment across sectors. The findings provide preliminary evidence that XAI-driven frameworks can enhance the trustworthiness and regulatory compliance of ESG analytics, particularly under the EU AI Act and Corporate Sustainability Reporting Directive (CSRD). The framework contributes to both research and practice by operationalizing explainability, fairness, and human oversight within ESG analytics, thereby supporting more reliable and comparable sustainability reporting.
Ahmad et al. (Tue,) studied this question.