What do academics say about artificial intelligence ethics? An overview of the scholarship

Abstract

This paper presents an overview of the academic scholarship in artificial intelligence (AI) ethics. The goal is to assess whether the academic scholarship on AI ethics constitutes a coherent field, with shared concepts and meanings, philosophical underpinnings, and citations. The data for this paper consist of the content of 221 peer-reviewed AI ethics articles published in the fields of medicine, law, science and engineering, and business and marketing. The bulk of the analysis consists of quantitative descriptions of the terms mentioned in each article. In addition, each term’s associations are analyzed to understand the specific meaning attached to each term. The analysis of the content is complemented by a social network analysis of cited authors. The findings suggest that some concepts, problem definitions and suggested solutions in the literature converge, but their content and meaning drive considerable variation across disciplines. Thus, there is limited support for the notion that shared concepts and meanings exist in the AI ethics literature. The field appears united in what it excludes: labor exploitation, poverty, global inequality, and gender inequality are not prominently mentioned as problems. The findings also show that the philosophical underpinnings of this academic field should be rethought: only a small number of texts mentions any major philosophical tradition or concept. Moreover, the field has very few shared citations. Most of the scholarship has been developed in relative isolation from others conducting similar research. Thus, it may be premature to talk about an AI ethics canon or a coherent field.