Storytelling with Data: Ethical AI and Machine Learning
NSF Unidata hosted the Unidata Users Workshop, Storytelling with Earth System Science Data: Challenges and Opportunities for Effective, Ethical, and Reproducible Science, on 5-8 June 2023 in Boulder, Colorado. This selection of presentations focuses on the ethical use of AI and machine learning.
Presentations
Artificial Intelligence/Machine Learning: Why and When
Skylar Williams This session introduces key considerations for deciding when and why to apply artificial intelligence (AI) and machine learning (ML) to scientific data. Participants explore the conditions necessary for successful implementation, learn about common misconceptions, and examine real-world examples from both the public and private sectors. The session emphasizes a thoughtful, context-aware approach to AI/ML in data analysis. |
|
AI/ML: Foundations
Thomas Martin This session provides an introduction to machine learning terminology, use cases, and common pitfalls, with a focus on applications in Earth system science. It is designed for participants who are curious about ML but have limited experience. |
|
Machine Learning Data Challenges Within the Earth Sciences
Charlie Becker This session examines the unique data-related challenges faced when applying machine learning in the atmospheric sciences. Topics include data sparsity, quality concerns, computational demands, non-linear relationships, and spatiotemporal dependencies. The presenter shares both high-level insights and case study examples, offering strategies for overcoming these challenges and ideas for future exploration. |
|
Data Ethics and Data Sovereignty
Jeff Weber, Dennis Dye, Peter Romine This session focuses on Indigenous Data Governance and Indigenous Data Sovereignty. Participants learn about the Sovereign Network Project, a collaborative NSF-funded effort between Southwestern Indian Polytechnic Institute, Navajo Technical University, and the Unidata Program Center. The session highlights ethical considerations when working with Indigenous communities and data. |
|
Introduction to scikit-learn
John Schreck, Thomas Martin This session introduces fundamental machine learning concepts using the scikit-learn Python library. Participants engage in hands-on exercises involving data preprocessing, model selection, and model evaluation. By the end of the session, attendees gain a foundational understanding of how to apply machine learning techniques in their own projects. |
|
eXplainable AI Methods
Will Chapman, Kirsten Mayer This breakout session explores explainable artificial intelligence (XAI) methods for interpreting neural network predictions. Participants apply XAI techniques to a climate science case study and examine the benefits and limitations of different interpretability approaches. |
|
Techniques in Education: Geoscience Machine Learning Resources and Training
Nicoleta Cristea This session discusses educational strategies and resources for teaching AI/ML concepts within the geosciences. Participants explore approaches to curriculum design and training that help learners connect machine learning theory with scientific applications. |
|
Details
Learning Series
Storytelling with Data
Intended Audience
Earth Systems scientists and students
Format
Video, Jupyter Notebook