Sustainable and Trustworthy AI for Business Intelligence and Financial Decision-Making in the Digital Era

Unofficial AI-generated summary based on the public title and abstract. Not an official translation.

🔬研究者:Provides a novel multi-objective framework for sustainable and trustworthy AI, highlighting trade-offs between accuracy, energy, and fairness.

🏢実務担当者:Offers practical guidance on model selection for green finance applications, showing how to balance performance with sustainability and bias.

🏛政策担当者:Illustrates the need for regulations that consider both AI sustainability and fairness, providing evidence for informed policy-making.

In the modern digital era, Business Intelligence (BI) and financial decision-making depend more on Artificial Intelligence (AI) to provide more efficiency and accurate predictions. Nonetheless, the fast uptake of high-compute models poses serious questions about the bias of the algorithms, their non-transparency, and the sustainability of the environment. This study presents these issues through a solution, the new three-layer unified architecture of Data, Modeling, and Evaluation layers that will allow balancing the performance and the ethical and ecological values. This study breaks the norm of integrative frameworks in which the minimization of prediction loss, energy consumption, and demographic bias are not considered simultaneously, since it proposes a multi-objective optimization approach that considers all three. A case study, Green Finance Credit Audit, was used to validate the framework based on the German Credit and Stock Market datasets. Experimental findings demonstrate that a major trade-off is that the Neural Networks had the best accuracy (89.4%) and AUC-ROC (0.91), but they also have the biggest Demographic Parity Gap, and the largest energy consumption in terms of FLOPs and training time. On the other hand, the most balanced solutions were given by the Random Forest and Logistic Regression models, as they had better explainability (SHAP scores of 0.75 -0.80) and minimal resource usage (reduced by a significant margin, 83.7% -86.1%). The study reaches a conclusion that implementing this sustainable and reliable paradigm is a long-term strategic requirement in terms of corporate value and regulatory discharge in the financial industry. The study offers practical information to practitioners and policymakers who would like to promote responsible use of AI.

• semanticscholar https://doi.org/10.1051/itmconf/20268601005first seen 2026-06-09 04:59:15 · last seen 2026-06-15 05:34:50