Artificial intelligence (AI) is rapidly transforming climate science by helping researchers forecast extreme weather, analyze satellite data, and model future climate risks. But a new report from the University of Virginia argues that this technological revolution must also be a social one.

Titled “Broadening Access to Climate AI Innovation,” the report urges scientists, policymakers, and communities to rethink how AI processes for climate research are designed, governed, and shared. The authors, including Mona Sloane (Data Science), Antonios Mamalakis (Data Science and Environmental Sciences), Charity Nyelele (Environmental Sciences), and Ava Birdwell (Environmental Sciences), emphasize that progress in AI, particularly as it relates to climate predictions, will only be effective if it’s inclusive, transparent, and responsive to the needs of diverse stakeholders and not just those with access to powerful computational tools.

The authors not only drew on their interdisciplinary research but also hosted multi-stakeholder workshops, with over 25 people from approximately 20 organizations in attendance.

“We’ve learned that while artificial intelligence is redefining what’s possible in climate research, technological promise alone won’t solve the climate crisis,” said Mona Sloane, co-author of the publication. “Our study demonstrates that the future of AI that models change must be built on shared infrastructure, interdisciplinary collaboration, and expanded participation where scientists and communities co-design the tools that guide discovery, policy, and resilience. When we widen who shapes AI, we deepen the trust, relevance, and impact of the science itself.”

The report identifies five key takeaways:

1. Redefining Extremes:
   Extreme weather events should be understood through their real-world impacts on people and ecosystems, not just through statistical anomalies.
2. Embracing Uncertainty:
   Instead of hiding AI’s uncertainty, researchers can use it to invite broader forms of knowledge and collaboration.
3. Prioritizing Explainability:
   AI models must be interpretable and communicable to multiple audiences (from data scientists to community leaders).
4. Reconfiguring Expertise:
   Climate-impacted communities should help design AI tools, ensuring they reflect local realities and priorities.
5. Building Inclusive Infrastructure:
   Long-term investment in shared data resources, interdisciplinary partnerships, and fair governance is essential for sustained innovation.

“As AI becomes central to climate modeling and decision-making, questions of who participates, whose knowledge counts, and how results are communicated are increasingly urgent,” said Ava Bridwell, graduate student and co-author of the publication.

There are immediate opportunities for change that could guarantee AI is more useful when it comes to climate modeling. Community participation should be embedded in the earliest stages of climate AI development; access should be expanded to give underrepresented researchers and institutions access to data resources; a transparent governing framework around AI-driven climate research should be prioritized; and interdisciplinary collaborations that bridge data science, social science, and environmental studies should continue the work to assess these topics.

“Overall, this report calls for a new kind of climate innovation ecosystem,” stated Sloane, “one that values collaboration over competition, and transparency over opacity.”

About the Project

This report was produced through a collaboration between UVA’s School of Data Science, Department of Media Studies, and Department of Environmental Sciences, with support from the UVA Environmental Institute.