The Human-Centric Algorithm: Synergising AI-Driven Fraud Detection with Ethical Human Oversight to Optimise Sustainable Banking Integrity (World Journal of Entrepreneurship Management and Sustainable Development)

Summary

• In 2025, global digital payments value reached $2.0 quadrillion across 3.6 trillion transactions, of which $1 trillion were fraudulent.p.2
• The study evaluated the Kaggle credit card fraud dataset containing 284,807 anonymised transactions to assess AI fraud detection models.p.6
• Analysis showed that the total transaction amount committed by fraudsters (12,653,519.54) was very similar to that of authentic users (12,509,070.47).p.8
• Transactions flagged as suspicious by the AI system totaled a value of 24,994,248.97, demonstrating significant overlap between genuine and fraudulent behavior.p.9
• Automated anomaly detection exhibited a false-positive rate of 99.43%, incorrectly classifying 12,437,155.08 in legitimate transaction value.p.10

Key insights

1. Over-reliance on purely automated AI fraud detection increases false positive risks and causes unintended financial exclusion for legitimate banking customers.p.2
2. Ethical Human Oversight theory provides necessary accountability by maintaining human authority over AI systems during high-stakes financial evaluations.p.3
3. Transaction monetary value alone is not a sufficient discriminator for identifying fraudulent financial activity.p.8
4. AI systems relying exclusively on hardcoded anomaly logic fail to distinguish complex legitimate user behavior from malicious transactions in grey-zone clusters.p.9
5. High rates of false positives cause severe operational disruption, reduced customer trust, and reputational damage to financial institutions.p.10
6. Integrating Human-in-the-Loop mechanisms with explainable AI enhances banking governance, transactional fairness, and sustainable banking integrity.p.13