ANNOUNCEMENTS
Training Artificial Intelligence (AI) platform for analysing non-financial disclosures of companies for ESG due-diligence needs number of guidelines to be framed to ensure that training data is annotated consistently. The platform, as a product, should be flexible in terms of analysing the non-financial disclosures as different clients (Asset owners / Asset managers) may have varied approach towards Environment Social Governance aspects, may have different understanding of how an aspect must be answered even if they follow same standards, protocols and use similar methodologies to analyse the companies. In order for the platform to be flexible, the underlying AI model must be able to accommodate the considerations of different clients. This involves framing different guidelines (even for same ask) for different clients to guide how the AI platform should analyse the disclosures. The current guidelines framing process is ad-hoc and not structured. This study attempts to generalize the process and bring structured approach by proposing dimensions which needs to considered while framing guidelines. The non-financial disclosures of companies are analysed to capture the variations in disclosures and dimensions associated with it. This process is iterated for different ESG aspects and most common dimensions are identified. As a result of the study, 10 dimensions are proposed which have to be decided based on client preference to train the model. Changes in the response to the dimensions will change the guidelines which in turn changes the platform’s output.
Keywords: Artificial Intelligence, Subjective variations, ESG aspects, Annotations, Dimensions.
