Machine learning and compliance: A consumer-led approach

Abstract

For compliance professionals, addressing the risks and opportunities of technologies such as machine learning is a huge challenge. This paper examines the issues involved with machine learning and insurance, by combining known regulatory concerns from international and UK supervisors with the results of consumer research, to identify key risks and how they can be mitigated. The paper finds that blending consumer research and wider risk management analysis can lead to a holistic compliance approach to machine learning. This builds from the individual, through job role design, individual accountability and training, to organisational systems and controls including governance, pricing policies and data management, through to sector-wide systems and initiatives including management of third parties and consistent standards for model transparency. In particular, it can help focus financial services firms on elements that consumers consider to be highly important, such as pricing according to risk rather than other factors (for example, price elasticity of demand). Conversely, it can help firms take a more proportionate approach to societal factors, such as exclusion of high-risk groups, that insurers alone cannot resolve without partnership with civic authorities.