Artificial intelligence (AI) and machine learning (ML) are becoming increasingly important tools in volcano science, with applications spanning geophysical monitoring, remote sensing, numerical modelling, volcano-tectonics, igneous petrology, geochemical data analysis, and volcanic hazard assessment. These approaches offer new opportunities to automate tasks, enhance modelling strategies, analyse increasingly complex datasets, and accelerate knowledge discovery. At the same time, their integration into scientific research introduces important methodological and epistemological challenges, including understanding what AI models actually learn, assessing the physical and geological meaning of identified patterns, and ensuring transparency, reproducibility, robustness, and interpretability.
Beyond their predictive capabilities, AI and ML also raise broader questions about how data-driven methods contribute to scientific understanding. In volcano science, where observations are often incomplete and direct constraints on subsurface processes are limited, identifying correlations is not necessarily equivalent to understanding the mechanisms that generate them. Integrating domain knowledge, physical constraints, and causal reasoning may therefore help develop approaches that are both effective and scientifically meaningful. These issues become especially important in high-risk contexts such as volcanic hazard assessment, risk mitigation, crisis management, and risk communication.
This session welcomes contributions across volcano science, including geophysics, remote sensing, experimental and numerical modelling, volcano-tectonics, igneous petrology, geochemistry, monitoring, and hazard assessment. We particularly encourage work on data-driven discovery, physics-informed ML, automated data analysis, model interpretability and explainability, uncertainty assessment, reproducibility, and the integration of AI-derived results with geological and physical understanding. We also welcome emerging approaches such as digital twins and agentic AI, as well as epistemological perspectives on how AI may influence scientific inference, explanation, hypothesis generation, and scientific understanding. Finally, the session welcomes studies on the responsible use of AI in volcanic-hazard applications and its interaction with scientists, volcano observatories, stakeholders, and decision-makers.