#1: Why use AI for climate cooperation?
#1: Why use AI for climate cooperation?
Episode 1 of a plain-English guide to using AI for modeling climate cooperation.
Reading time: 5 mins.
Prerequisites: An interest in learning about how we could tackle climate change policymaking with the help of AI and multi-agent modeling.
Climate change affects all
Climate change and its impact are becoming more tangible, for example, extreme weather events are becoming more frequent and more severe across the globe. However, mitigating climate change is not easy: it is a complex and global problem – no single solution is enough on its own. Moreover, cognitive biases and uncertainty can impact how much action we are willing to take.
As such, it is becoming increasingly clear that climate change is not a problem that can be solved by any individual nation but requires a collaborative effort.
Why is global cooperation essential?
Climate change is a complex issue that affects everyone in different ways. Our climate is a shared resource and a common environment: each nation’s actions can affect the whole planet.
Hence, mitigating climate change is not only the responsibility of one nation or one community. It involves the collective responsibility of the whole world.
This is why effective climate diplomacy is critical.
Why is cooperation not easy?
First, climate change poses a tragedy of the commons
; we’re all living under one shared roof: the climate is a public good. However, individual nations may not meet sustainability limits on their own and emit too much. Hence, the climate will not be sustainable if all nations act out of pure self-interest and without cooperation.Second, climate policies face the tragedy of the horizon
; Our actions today not only affect us but also future generations who are not aware of our actions. How can we take their concerns into account? How does that affect our sense of urgency to act now?
Third, the behavior of economies is complex and the stakes are different across nations. Many factors play a role: changes in technology, domestic and international politics, and different beliefs and traditions. Climate innovation is expensive and can be slow, especially for developing nations. Moreover, the impacts of climate change are not equally distributed and nonlinear.
Fourth, there is uncertainty about the future. How good are we at foreseeing how the climate will evolve? How should we invest our resources and balance economic versus climate goals, if we are not sure when climate damages will occur? How can we cooperate when our perception of risk and level of risk aversion is different?
Lastly, decisions have to be coordinated amongst all nations, but there is no central authority that can enforce agreements. How can we ensure that global cooperation on climate change is sustained over the long term even in the absence of a central authority?
How can AI help solve this problem?
Our goal is to understand the incentives and factors that influence cooperation on climate change. and how we can design negotiation protocols and agreements that maximally align the incentives of nations and global regions (also called agents).
A useful framework to do this is multi-agent reinforcement learning. Why?
We can use AI agents to build more realistic models of how agents make decisions.
We can use simulations to compare millions of different outcomes, including the effects of different decisions that agents might make.
Reinforcement learning gives us great flexibility in what goals and constraints agents optimize for and match real-world policy objectives.
We need experts from many disciplines to collaborate to design these simulations, define the goals and constraints on how nations make decisions, etc. This is crucial to make sure that the outcomes are relevant and usable in the real world.
Food for thought
In this guide, we will explain key technical concepts in plain English to help everyone get a basic understanding of this framework and suggest ways in which you can help shape the research. To this end, here are some open-ended questions.
Does climate change pose a collective action problem or a distributive conflict?
Does one nation’s climate decisions depend upon the decisions of its neighbors?
Which other agents play a crucial part in climate negotiations other than individual nations?
What makes a nation adopt climate initiatives even without climate negotiations? What makes nations honor agreements?
How do cultural and social factors affect climate negotiations and agreements?
In the next post, we will explain how the AI framework works, and how these questions might be addressed within the framework.
Authors and Collaborators
Niranjana Ragavan, Lu Li, Andrew Williams, Stephan Zheng.
Check out www.ai4climatecoop.org to learn more about AI4GCC.
Follow us on Twitter and WeChat.
Fagan, Moira, and Christine Huang. “A Look at How People around the World View Climate Change.” Pew Research Center, 18 Apr. 2019, https://pewrsr.ch/2UpGcq7.
Hardin, Garrett. “The Tragedy of the Commons.” Science, vol. 162, no. 3859, American Association for the Advancement of Science (AAAS), Dec. 1968, pp. 1243–48. Crossref, https://doi.org/10.1126/science.162.3859.1243.
Carney, Mark. "Breaking the tragedy of the horizon–climate change and financial stability." Speech given at Lloyd’s of London 29, 29 September 2015, https://www.bis.org/review/r151009a.pdf.
Aklin, Michaël, and Matto Mildenberger. “Prisoners of the Wrong Dilemma: Why Distributive Conflict, Not Collective Action, Characterizes the Politics of Climate Change.” Global Environmental Politics, vol. 20, no. 4, MIT Press, Nov. 2020, pp. 4–27. Crossref, https://doi.org/10.1162/glep_a_00578.
Some resources from where you could learn more
Climate statistics: https://climate.nasa.gov/
UNFCC’s COP27 conference: https://unfccc.int/cop27