Towards Explainable Artificial Intelligence in Banking and Financial Services

Explainability is critical for trust in financial AI: The paper identifies understanding and appropriately trusting AI algorithm results as "the main challenge for human users," particularly in banking where decisions directly impact customers' financial lives.

A balance between performance and transparency is achievable: The authors present a novel XAI process that "facilitates producing explainable models while maintaining a high level of learning performance," suggesting that organizations don't have to sacrifice accuracy to gain interpretability.

Interactive tools help humans understand AI decisions: The paper demonstrates that a "digital dashboard to facilitate interacting with the algorithm results" can significantly improve how data scientists and presumably other stakeholders comprehend AI outputs, making complex algorithms more accessible to non-technical users.

Real-world application in transaction monitoring: The research applies XAI methods to a practical banking scenario analyzing customer transactions, showing that explainability techniques have concrete, implementable value in actual financial services operations rather than remaining theoretical.

Evidence-based explanations build confidence: The paper proposes an "interactive evidence-based approach" to help users comprehend AI results, suggesting that grounding explanations in concrete evidence—rather than opaque algorithmic outputs—is essential for responsible deployment of AI in high-stakes financial decisions.