Mapping sustainability in higher education: a natural language processing analysis of student theses

Artikel i vetenskaplig tidskrift, 2026

This paper examines the integration of sustainability in higher education, focusing on how business, economics, and finance students relate their work to the Sustainable Development Goals (SDGs). We develop SustainSight, an inference engine using natural language processing to assess sustainability content in written text. Using SustainSight, we analyze over 23,000 bachelor’s and master’s theses from the five largest Swedish universities offering programs in business, economics, and finance spanning a 15-year period. This unique dataset allows us to provide insights into how sustainability is reflected in students’ theses, with a focus on variation by gender, education level (bachelor/master), discipline, and university. For the subset of theses, for which we have the authors educational program and data on curriculum reforms, a staggered difference-in-differences model is estimated. We find that integration of sustainability into degree requirement substantially increases the likelihood that students include sustainability perspectives into their theses. Finally, we conduct a topic analysis to explore in greater depth what students write about when addressing sustainability-related issues.