Measuring responsible artificial intelligence (RAI) in banking: a valid and reliable instrument

Abstract

Widespread use of artificial intelligence (AI) and machine learning (ML) in the US banking industry raises red flags with regulators and social groups due to potential risk of data-driven algorithmic bias in credit lending decisions. The absence of a valid and reliable measure of responsible AI (RAI) has stunted the growth of organizational research on RAI (i.e., the organizational balancing act to optimize efficiency and equity). To address this void, we develop a novel measurement instrument to assess RAI maturity in firms. A review of the nascent literature reveals that there is a wide distribution of RAI capabilities. The RAI instrument that we advance is based on the exhaustive review of this dispersed literature. Analyses of data from large US banks show strong evidence of validity and reliability of the RAI maturity instrument.