Opportunities, epistemological assessment and potential risks of machine learning applications in volcano science

Abstract

This manuscript explores the opportunities and epistemological risks of using machine learning in the Earth sciences with a focus on igneous petrology and volcanology. It begins by highlighting the benefits of machine learning, particularly in automating tasks, enhancing modelling strategies, and accelerating knowledge discovery. However, the integration of machine learning into scientific research also introduces significant challenges. Key concerns include understanding what machine learning models learn, ensuring transparency, reproducibility, and improving model interpretability. These issues become especially critical in high-risk contexts such as volcanic hazard assessment, risk mitigation, and crisis management, where the reliance on machine learning outcomes can have profound consequences for human lives. The manuscript also introduces additional ethical considerations, such as the risk of over-reliance on machine learning models and the broader implications of geopolitical development plans, laws and regulations in the EU, China, and the US.