Kolloquium: An LLM-Assisted Qualitative Content Analysis: the Assessment of Biodiversity Reporting

Abstract:
Artificial Intelligence (AI) is rapidly transforming scholarly research, with recent advances in large language models (LLMs) enabling new forms of knowledge discovery, information extraction, text interpretation, and scientific writing. This research examines the use of an LLM, GPT-4o, to augment qualitative content analysis of biodiversity reporting, emphasizing information extraction, text generation, and interpretation. We define the scope of this study as the use of LLMs as interpretive tools for deductive qualitative analysis of textual data, rather than computational approaches for exploratory text analysis and predictive modeling. We employ the 2022 sustainability (ESG) reports of 10 coal mining companies for textual analysis. The analysis is guided by a biodiversity reporting assessment framework developed from the literature and sustainability reporting standards. We establish content validity through iterative prompt refinement with human oversight. Our findings show that most companies consider biodiversity to be material, have biodiversity strategies, and provide detailed information on biodiversity restoration. However, there is a lack of concrete targets and reported target achievements for biodiversity conservation and restoration, and few companies adopt systematic approaches or procedures to mitigate biodiversity loss. Companies also rarely recognize biodiversity loss as a business risk. Notably, biodiversity reporting practices are largely similar across the companies, with minimal regional differences. This research makes a methodological contribution to the IS field by highlighting the potential of LLMs as innovative and rigorous research assistants for qualitative content analysis. It also adds new knowledge to the biodiversity management literature by developing a novel framework for biodiversity reporting assessment.