A hypothesis-driven responsible AI framework for interpretable ESG forecasting with RuleFit

Muhammad T., Hafeez R., Khidmat W. B., Afzal M., Abid F., Elkiran H., ...More

Scientific Reports, vol.16, no.1, 2026 (SCI-Expanded, Scopus)

  • Publication Type: Article / Article
  • Volume: 16 Issue: 1
  • Publication Date: 2026
  • Doi Number: 10.1038/s41598-026-47870-3
  • Journal Name: Scientific Reports
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, BIOSIS, Chemical Abstracts Core, EMBASE, MEDLINE, Directory of Open Access Journals, Zoological Record, Academic Search Ultimate (EBSCO), Natural Science Collection (ProQuest), Biological Science Database (ProQuest), Biomedical Reference Collection: Corporate Edition (EBSCO), Health Research Premium Collection (ProQuest)
  • Keywords: Corporate governance, Ensemble learning, Responsible AI, RuleFit
  • Open Archive Collection: AVESIS Open Access Collection
  • Yıldız Technical University Affiliated: Yes

Abstract

The systematic integration of artificial intelligence (AI) within business practices and corporate governance has made Responsible AI (RAI) an absolute subset of corporate social responsibility. This research formulates a framework that integrates Responsible AI principles with hypothesis-driven rule extraction to predict Environmental, Social, and Governance (ESG) scores. Employing the ensemble-based RuleFit algorithm, we construct decision rules that algorithmically associate non-linearly with firm-level attributes such as size, leverage, digitization intensity, and managerial ownership. These rules are then transformed into domain-specific hypotheses and subsequently subjected to statistical validation using nonparametric tests to ensure robust results under non-normal outcome distributions. Performance predictive models are augmented to preserve equity by simultaneously assessing subgroup fairness through the distribution of residuals and biased proportions of firm characteristics. In summary, the proposed methodology illustrates how rule-based learning can yield interpretable, statistically sound, and ethically aligned forecasts of ESG outcomes, thereby facilitating reliable decision-making in sustainable finance and corporate governance.