Algorithmic decision-making in financial services: economic and normative outcomes in consumer credit

Abstract Consider how much data is created and used based on our online behaviours and choices. Converging foundational technologies now enable analytics of the vast data required for machine learning. As a result, businesses now use algorithmic technologies to inform their processes, pricing and decisions. This article examines the implications of algorithmic decision-making in consumer credit markets from economic and normative perspectives. This article fills a gap in the literature to explore a multi-disciplinary approach to framing economic and normative issues for algorithmic decision-making in the private sector. This article identifies optimal and suboptimal outcomes in the relationships between companies and consumers. The economic approach of this article demonstrates that more data allows for more information which may result in better contracting outcomes. However, it also identifies potential risks of inaccuracy, bias and discrimination, and ‘gaming’ of algorithmic systems for personal benefit. Then, this article argues that these economic costs have normative implications. Connecting economic outcomes to a normative analysis contextualises the challenges in designing and regulating ML fairly. In particular, it identifies the normative implications of the process, as much as the outcome, concerning trust, privacy and autonomy and potential bias and discrimination in ML systems. Credit scoring, as a case study, elucidates the issues relating to private companies. Legal norms tend to mirror economic theory. Therefore, this article frames the critical economic and normative issues required for further regulatory work.