Utilizing large language models for identifying future research opportunities in environmental science

Facing pressing global challenges such as climate change, biodiversity loss, resource scarcity, and environmental pollution, the field of environmental science urgently needs innovative research methods. However, identifying meaningful and cutting-edge research topics is a significant challenge, as it requires a thorough understanding of existing literature and the ability to discern knowledge gaps. Traditional bibliometrics often fall short of capturing nascent interdisciplinary fields. Recent advancements in artificial intelligence (AI) offer potential solutions to this challenge. This study explores the capabilities of large language models (LLMs) in identifying and analyzing emerging research opportunities in environmental science. We employ a text retrieval method based on word embeddings, utilizing the emergent reasoning abilities of LLMs combined with embedded search techniques to dynamically integrate the latest literature. By comparing the GPT-3.5 API with supplementary literature, ChatGPT, and GPT-4, we find that the GPT-3.5 API provides a more comprehensive, detailed, and current analysis of cutting-edge environmental science, emphasizing the importance of understanding the dynamics and timeliness of the field. Our findings underscore the critical role of interdisciplinary research, AI, and big data in addressing urgent environmental challenges. LLMs can serve as valuable tools for researchers, offering guidance and inspiration for future directions in environmental science research. [Display omitted] •Explores how LLMs identify emerging research directions in environmental science.•Illustrates the superiority of LLMs over traditional bibliometrics in identifying interdisciplinary fields.•Highlights how LLMs extrapolate from existing literature patterns to discover new trajectories.•Emphasizes the role of machine learning in diverse environmental research domains.