[1O3-GS-11] AI and Society:

[1O3-GS-11-05] Materiality Analysis Based on Macro-Environmental Monitoring for Sustainability Transformation

〇Takashi Yamamoto¹, Lena Shinozaki², Hiroshi Shinohara¹, Ken Mohri¹ (1. Deloitte Analytics, Deloitte Tohmatsu Risk Advisory LLC, 2. General Advisory Unit, Deloitte Tohmatsu Risk Advisory LLC)

Keywords:Large language Models, Stacking, Sustainability transformation, Macro-environmental monitoring, Text categorization

Sustainability Transformation (SX) refers to a company's shift toward sustainability-oriented management, emphasizing the imperative of conducting collaborative business with stakeholders, from forecasting to macro-environmental monitoring. In this study, we attempted to classify "sentences" mentioning technological innovation, patents, investment trends, etc. found in SX-related news articles (termed technological factor sentences), examine trends by time-series changes and topic classification, and identify prominent issues (materiality analysis).The task of solving classification problems that cannot be clearly defined, such as technological factor sentences, is a common issue in monitoring not only SX but also geopolitical risk and supply chain risk. Therefore, in this study, we reduced annotation fluctuations by stacking large language models and constructed a highly accurate classifier. The results indicate that decarbonization technologies in SX are positioned in the "installation period" where innovative technologies progress from research and development to market implementation. Notably, energy industries emerge as leaders in this transformation. In addition, it was suggested that there is an increasing social interest ahead of the future creation of a hydrogen economy market.