AI-powered analysis of ESG disclosure: a clustering approach to determinants and motivations

Abstract The main aim of this study is to explore the determinants of corporate ESG reporting, and to highlight the various factors, determinants and motivations likely to explain the adoption of ESG reporting by entities operating in different contexts. Using the systematic review method, we set out to review and synthesize the literature on the impact of company characteristics, sector of activity and institutional context on ESG reporting. Referring to the PRISMA guidelines, we reviewed over 70 articles selected on the basis of inclusion and exclusion criteria covering the main aspects of the research question. The selected articles were processed using an approach integrating artificial intelligence tools, in particular natural language processing (NLP) to ensure thematic and semantic analysis of the data, followed by clustering analysis based on TF-IDF vectorization to analyze the identified determinants. The study demonstrated that ESG communication can be influenced by various factors such as company size, sector of activity, financial performance, governance structure… Algorithmic analysis of these factors led to the identification of six distinct clusters of underlying drivers, providing a two-dimensional analytical framework that illuminates not only the elements influencing ESG communication, but also the strategic motivations behind this approach. This empirical classification contributes, on a theoretical level, to a more detailed understanding of ESG signaling processes, and suggests a model encompassing the determinants and determinants of ESG communication.