Decision Trust Drivers of Explainable AI : A Qualitative Study in Consumer Credit

Sammanfattning

Artificial Intelligence (AI) applications are now widely used to score retail loan applications and handle credit risk portfolios. Nevertheless, the “black box” nature of these traditional systems prevents full adoption. A survey did in 2023 by Deloitte, reports 56% of US finance organizations hesitant to adopt AI due to fairness and accountability concerns. This qualitative study inquires how XAI features can reinforce trust among supervisory authorities and consumers in credit decisions. Therefore, study combined (i) systematic literature review, (ii) two consumer credit use cases, and (iii) semi-structured interviews with banking stakeholders. Thematic analysis carried in the study guided by Deloitte’s Trustworthy AI pillars and identified transparency, model reliability and user-relevant explanations as key drivers. Participants expressed preference for explainable models over complex and high accuracy. But expressed requirement for actionable feedback (how to raise credit score) rather than technical justifications. The study synthesized findings into “Trust in XAI” model framework that aligns explanations with stakeholder’s expectations.