Examining Algorithmic Bias in AI-Powered Credit Scoring: Implications for Stakeholders and Public Perception in an EU Country

Abstract

The presence of artificial intelligence (AI) in financial markets is becoming increasingly common, with AI algorithms used in fields as alternative credit scoring to render complex multifactorial decisions. However, concerns have been raised by agencies and researchers regarding the objectivity of AI based algorithms and their potential to perpetuate systematic inequalities among vulnerable populations. Through a simulation exercise using non-financial data, testing 126 application profiles, this research investigates the impact of AI in credit scoring, examining the presence of algorithmic bias and its implications for stakeholders. Additionally, it explores the perceptions of AI systems among the general population in an EU country surveying 144 individuals. Results from the simulation show presence of unequal treatment towards women applicants, making them less likely to get approved for a financial instrument compared to men. Furthermore, this study also reveals scepticism of general population towards automated decision-making systems and highlights their concerns about data privacy when interacting with AI systems.