Automating materiality assessment with a data-driven document-based approach

Materiality assessment is a critical process for companies to understand the interest perceived by its stakeholders towards topics related to environmental, social, and governance issues. Materiality assessment helps companies define their growth and communicative strategies; recently, it has become crucial within sustainability reporting, i.e., the practice of annually declaring the activities conducted to pursue economic growth in a sustainable way for society. In this paper, we propose a data-driven and automated approach to carry out materiality assessment. Stakeholders’ perception of important topics is obtained by analyzing relevant textual documents (e.g., company reports, press releases, social media posts), identifying mentions of potentially interesting topics, and converting them to scores that produce materiality rankings or matrices. An iterative methodology is proposed to incrementally carry out materiality assessment by progressively building the domain knowledge required to automate the process. Efficiency and effectiveness evaluations are carried out in a real-world scenario. • Textual documents are analyzed to detect material topics and quantify their importance. • Material topics are hierarchically modeled in a domain knowledge sup- porting multiple standards. • An iterative methodological framework allows to progressively converge to a reliable assessment. • AMA significantly reduces the time spent to manually read the documents. • AMA shows higher accuracy than domain experts in identifying and quantifying materiality.