Down To Earth: A podcast for Geoscientists by Geoscientist S06E08 Down to Earth: Help or Hindrance? The Ethics of AI-Driven Geoscience
Feb 6, 2025
Dr. Caroline Gevaert, an Associate Professor in Geoinformatics at the University of Twente and a consultant for the World Bank, dives into the ethical landscape surrounding AI in geoscience. She discusses the potential biases in AI technology and the critical need for explainability to support sustainable development, particularly in low-income countries. The conversation highlights the importance of local knowledge, community involvement, and ethical practices while addressing the complex relationship between machine learning and human decision-making.
AI Snips
Chapters
Transcript
Episode notes
Core Ethical Challenges in GeoAI
- Ethics in GeoAI involves cultural values and fairness to reduce biases in algorithms.
- Explainability and accountability are key components to ensure responsible AI use in geoscience.
Ethical Issues Using Drones
- Drone imagery for flood response helps estimate aid needs but risks biased outcomes harming vulnerable groups.
- Group privacy concerns arise as combined geospatial data may expose sensitive group characteristics without personal identifiers.
Value of Local Knowledge
- Local knowledge reveals details in geospatial data invisible to outsiders, improving research interpretation.
- Even with 90% model accuracy, errors reduce end user trust and must be critically evaluated with users.