Applying Machine Learning and Artificial Intelligence to CSR-Compliance. A Conceptual Framework with Illustrations.

Abstract

Corporate Social Responsibility (CSR) became an ever more relevant theme for firms in their dealings with investors, customers, and the public at large. CSR is also a domain, where problems of misconduct and non-compliance may occur. Investors and Activists are interested in the CSR-related compliance for practical reasons, researchers are interested in whether compliance in the domain of CSR can be predicted or detected. Structurally, this problem is similar to detecting non-compliance in the domain of financial regulation (fraud detection), a standard application of methods of artificial intelligence and machine learning, which are already applied successfully in the domain of legal /financial compliance. A yet unanswered question is, whether such methods can be applied in the domain of CSR, too. This conceptual paper outlines possible strategies for applying methods of artificial intelligence and machine learning to the domain of CSR. The paper starts out from elaborating differences between compliance in the domain of CSR, compared to legal and financial compliance. A crucial difference is that in the case of legal and financial compliance, non-compliance is defined in legal terms, and can, in principle, be recognized objectively, with official institutions doing the classification on which data sets are based. In the domain of CSR, compliance is more difficult to define and, consequentially, much more difficult to detect. This paper proposes and illustrates options for harnessing the potential of methods of machine learning and artificial intelligence for issues of CSR-related non-compliance.