Explainable Artificial Intelligence For ESG-Integrated Portfolio Management: Toward A Conceptual Decision-Making Framework

The integration of Environmental, Social, and Governance (ESG) criteria into portfolio management has become both a necessity and a challenge, introducing complexities such as heterogeneous data sources, divergent rating methodologies, and the risk of greenwashing that traditional portfolio models were not designed to address. Artificial intelligence emerges as a natural response to these limitations, offering the capacity to process complex ESG signals, detect inconsistencies in corporate disclosures, and support adaptive allocation strategies. However, this solution introduces a new contradiction: the more sophisticated the model, the more difficult it becomes to understand and evaluate its decision-making process. This opacity raises important concerns regarding transparency, accountability, and regulatory compliance within sustainable finance. Using a narrative literature review across sustainable finance, portfolio theory, and explainable artificial intelligence research, this study examines the structural challenges underlying ESG-informed decision-making and proposes a novel four-layer conceptual framework, bridging ESG data, algorithmic processing, explainability, and institutional governance, that positions explainability as a structural governance mechanism rather than a purely technical feature. The framework offers a foundation for evaluating how AI-driven investment systems can remain interpretable, contestable, and institutionally accountable.