The research aims to analyse whether Artificial Intelligence (AI) and Machine Learning (ML) alone are efficient and self-sustainable in detecting financial fraud, or if they need human intervention and supervision for better fraud detection activity in digital finance.
The Kaggle credit card fraud dataset of 284,807 transactions was used to determine the grey zone of suspicious flagging. To analyse the occurrence of false positives, a linear regression model was applied to the first 1,000 anonymised banking transactions.
Results demonstrated a high degree of false-positive exposure in automated anomaly-detection systems. The regression analysis also yielded a statistically significant model (F = 4.255; p < 0.001); this underscores the complexity and unpredictability of fraud behaviour in online banking contexts.
The current research advances the investigation into the likelihood of false positives detected by AI and whether further human supervision is required to intervene in the detection or validation of suspicious flagging.
The study is limited to the first 1,000 observations from the dataset to ensure manageable statistical interpretation and analytical consistency.
The research offers a transformational roadmap for applying human contextual judgement to banking ecosystems, balancing the predictive accuracy of AI with the moral responsibility of human intelligence to create a resilient, inclusive, and sustainable financial future.
Masanawa, A.U. and Shaheen, A. (2026): The Human-Centric Algorithm: Synergising AI-Driven Fraud Detection with Ethical Human Oversight to Optimise Sustainable Banking Integrity. In Ahmed, A. (Ed.): World Sustainable Development Outlook 2026, Vol. 22, pp. 195-213. WASD: London, United Kingdom.
