Seminar Recap: Climate Change and the Convergence of Complexities

By: Tyler Wells Lynch

Understanding how complex systems interact with one another is an all-consuming goal of artificial intelligence. More accurate models of how natural, engineered, and social systems converge are not only useful but necessary if we’re to make better decisions about how to deal with anthropogenic climate change.

The problem, according to Auroop Ganguly, professor of civil and environmental engineering at Northeastern University and core faculty member at the Institute for Experiential AI (EAI), is that through mutual support these systems inherit mutual vulnerabilities. Climate change is a threat precisely because of those vulnerabilities: the first domino in a very long line of system collapse. The good news is we have increasingly powerful tools for understanding the nature of these cascading systems and how they interact with each other. On this charge, artificial intelligence is leading the way.

As part of EAI’s Expeditions in Experiential AI Seminar Series, Ganguly identified some of these tools and made the case for advances in climate resilience that go beyond mere models and simulations.

Modeling the Planet

At the top of the pecking order of complex systems are earth systems. These can be thought of as high-level composites of smaller systems, which include climate, weather, industrial activity, reforestation, population dynamics, and even geological processes—a vastly complicated representation of all the activity on the planet. So complex are these systems that only the biggest and smartest supercomputers in the world can properly model them. 

“There's a need to be more aware of all the correlations that can occur here, the kind of long range dependence that we see over time and space,” Ganguly said. “You have to be aware of the physics, the different variables guiding, let's say, rainfall extremes and so on … And you have to be aware of uncertainties.”

That’s a tall order for any computer, which is why climate modeling tends to require high-performance computing.

Modeling the Anthroposphere

Further down the system pecking order, we have engineered systems. These include power grids, communications systems, transportation systems, water distribution systems, and other kinds of infrastructure. Some of these systems are tightly coupled, either because humans built them that way or because they self-organized due to efficiency advantages. But here’s where the vulnerabilities begin to show. Ganguly: “Given that we are seeing more and more extremes from natural systems—heat waves, cold snaps, floods, droughts, hurricanes—if any of these extremes cause damage at one single point, that damage might percolate across the system.”

In this realm, artificial intelligence is enormously helpful for modeling, predicting, and responding to vulnerabilities and system shocks. Specifically, scientists rely on probabilistic graphical models, causal flows, and engineering principles to make sense of them.

“So in one case it's spatial-temporal machine learning,” Ganguly said, referring back to earth systems modeling, “and in another case it's graphical models and machine learning.”

Modeling People

Deeper down, the challenge is more human-focused. Social systems, being made up of people, are notoriously difficult to predict. They’re sensitive to squishy, equivocal phenomena like greed, corruption, regulatory inertia, and behavioral problems like the Tragedy of the Commons. But even here, Ganguly said, activity can be predicted using agent-based models and games for multi-stakeholder decisions—all guided by social science theories and policy frameworks.

Summing up, Ganguly said, “In each of these areas—natural systems, engineered systems, social systems—we have different flavors of AI which could be immediately useful.”

Beyond Modeling

As helpful as AI may be, modeling is not enough. Despite their vulnerable position in the pecking order of complex systems, human beings have the ultimate say over the extent of system cascade.

“If you think of the climate problem, it's a combination of things,” Ganguly said. “It's policy myopia, it's financial disincentives that are built-in in terms of climate preparedness, which lead to outdated best practices, engineering stagnation. It's a vicious cycle.”

But, through ongoing research, Ganguly showed how it doesn’t take a whole lot to transform that vicious cycle into a virtuous one. What’s needed is the right set of financial incentives and transformative signs to put the wheel in motion.

“I'm not a pessimist by any means,” Ganguly said, “but we need to do this soon.”

Want to learn more? Watch a replay of Gangulay’s talk or flip through his slides. And don’t forget to register for upcoming Seminar Series events.