Evidence and Learning in the Context of Climate Change: Invitation to Action

In the Climate sector, rapid learning and adaptive management are critical. Monitoring, evaluation, research and learning (MERL) professionals have a critical role to play in expanding and leveraging the evidence base and advocating for the most effective and equitable solutions for addressing the climate crisis. Business as usual is not enough in the current times – MERL needs to evolve to meet the moment in light of the urgency of the climate crisis.

The MERL Tech Initiative joins some 20 global organizations who are launching “Evaluation and Learning in the Context of Climate Change: An Invitation to Take Action,” an initiative that aims to support climate action through evidence and learning.

This effort was spearheaded by 22 leaders working at the nexus of climate change, evaluation, and evidence generation and use. Representing philanthropy, climate funds, government, civil society, private sector, and multilateral organizations, we met for three days at the Rockefeller’s Bellagio Center in September 2024 to discuss ways to foster collaboration opportunities across sectors, geographies, and actors. The core question we explored during the Bellagio convening was:

“How can we amplify the value and impact of evidence and learning in addressing the urgent climate crisis?”

I was invited to join the meeting to bring in perspectives on how artificial intelligence (AI) could support climate evidence and learning. It was important for me to also raise the downsides of AI in terms of environmental impact, global warming and its extractive nature. If we’re going to use AI for climate-related MERL, how do we manage the tensions and tradeoffs? How do we ensure it’s doing more good than harm? 

AI in Climate Learning and Evaluation

AI can serve as a tool to improve climate evidence and learning – for example through deepened understanding of the effectiveness of climate policy and efforts to improve collective impact of the ecosystem of climate funders. At the same time, concerns are growing about the amount of water and energy required to fuel more and more data centers to support AI. 

As we highlighted in November 2024, the climate costs of AI are high in terms of global emissions, energy use, and water use, and there is little accountability required from Big Tech. The incoming US administration is paving the way for unfettered growth of AI, which means negative impacts will only worsen. The question then is how can we use AI to advance climate-focused MERL while ensuring that the harm doesn’t outweigh the benefits? 

AI is an extractive economy, built in the same way as other similar extractive industries. In 2023, MTI warned of this and called on evaluators to support AI as a regenerative economy.

At the Bellagio convening, we learned about emerging initiatives using AI to speed up and improve evidence and insights that can then lead to more effective actions. We also discussed the challenges of evidence uptake in the current global context and the need to find better ways to use evidence and narrative to influence decision-making and action at various levels – from the public to climate funders to elected officials.

Invitation to Take Action

The Invitation to Action proposes five shifts in climate evidence and learning, and as MTI we endorse these. They include 1) shifting the focus from projects to systems transformation; 2) championing evidence-informed decision making; 3) deploying evaluative practice early and often; 4) embracing multiple ways of knowing; and 5) learning collectively to scale impact. We believe that AI can play a role in supporting these, as long as the downsides of AI are acknowledged and addressed.

If we are going to use AI in the fight for the planet, MTI also calls on Climate Funders to invest in green, sustainable, local-first AI. 

This means funding the development of small, climate conscious AI that is:

• Federated and decentralized, and that moves away from centralized, resource-intensive Big Tech
• Low-resource and able to function using minimal data and computing power
• Transparent, so that users can understand the ecological impact of the models they are using
• Energy efficient and running on renewable energy sources
• Contextual and respectful of culture diversity, local relevance, and indigenous data principles
• Intentional and limited to actual need – where is AI appropriate and where is it overkill?

Joana Varon, Sasha Costanza-Chock, and Timnit Gebru have outlined a series of recommendations for shifting AI development away from big tech. The local first movement also offers inspiration on how to do this. 

The signatories of the Invitation to Action call for individuals and organizations that generate, fund and use evidence and evaluations to join us in evolving our practices. No one organization can accomplish this and we need collective action.

From our side, MTI’s Natural Language Processing Community of Practice has recently launched a Climate, MERL and AI Working group aiming to explore these ideas further in order to offer space for learning, sharing, and defining actionable steps for AI and climate-related MERL. The Working Group space is also aimed at those who are using AI for MERL in other areas and have questions about the potential benefits and environmentally-related downsides of AI in MERL. 

We join the signatories of the Invitation to Action in hoping that by articulating and building consensus around ways of working, we can begin to change practice within our own organizations and across climate evidence and learning more broadly. In doing so, we aim to amplify the relevance and impact of evidence in addressing the climate crisis.

Learn more about the Evidence and Learning in the Context of Climate Change Initiative here and join the NLP-CoP (select the Climate, MERL and AI Working Group when you sign up) to contribute to discussions (and ideally actions!) in this area. We’d love to talk with funders about exploring this, too!