Models help us understand how to achieve a desired outcome given certain inputs. What we choose to include and what we choose to exclude from a model dictates what gets consideration and what gets forgotten. Everyday millions of models come into the world. Each modeler striving for elegant simplicity. Make the fewest assumptions possible and then at the very bottom of the spreadsheet add a footnote in the smallest font listing everything that isn't included. The climate crisis was born from these footnotes of exclusion. Interactions that were too complex or consequences that were too far away were simply left out. All stemming from a lack of understanding paired with a lack of incentive. The sinister thing about a model is that once we start using one, we anchor to it. We stop seeing reality and only perceive reality through what we’ve constructed. Our modern world is built on incomplete models. And incomplete models are worse than nonexistent ones. Mitigating climate change dictates we do the hard work of building complete models across all of our human activities.
Incomplete models lull us into a false sense of confidence. The respect for actual observation is lost as reliance on a model grows. If we have no information about how something works then we’re forced to start from a more holistic place where everything merits consideration. If we have a model which indicates that more of input A leads to more of the desired output B, then there’s little reason to think about output C. To ground this in example, we can look to the differences in agricultural practices between North American indigenous tribes in the 15th century and North American industrial farmers in the 20th century.
When indigenous tribes lived off the land, there was no knowledge of nitrogen, oxygen, or carbon dioxide. No scientific understanding for why plants grow. Indigenous tribes had their own ways of knowing but the Nitrogen-Phosphorous- Potassium model of agriculture we’re familiar with today was nonexistent. Despite lacking an understanding of underlying mechanisms, indigenous tribes practiced some of the most sustainable agriculture that’s ever taken place on the continent. Without a streamlined model to use, their ways of knowing relied on generational observations that considered all aspects of the land.
Whatever knowledge indigenous tribes had accumulated was wiped out by the genocide of American expansion in the intervening period between the 15th and 20th centuries. In its place industrial agriculture took root along with new methods of thinking. Armed with scientific and financial knowledge, the model for agricultural production could be vastly simplified and consequently scale across the continent. Once the importance of nitrogen to plant growth and the means for manufacturing synthetic nitrogen fertilizer were established, the agricultural model became a single line: Add more nitrogen as input to get more crop as output. Most agriculture today still rests on this model: Fertilize fertilize fertilize! All the other elements of the farm ecosystem were lumped into the footnotes and promptly forgotten. Or at least forgotten until the past few decades when the effects of such an incomplete model became too problematic to ignore. From anoxic dead zones in the ocean to degraded soils across the globe, the one line model has shown it’s vast limitations.
This is the story of climate change. Overtly simplistic models scaled to the point where all that was ignored comes leaping out of the footnotes to wreak existential havoc. The reasons for relying on these incomplete models have mostly to do with our choice of capitalism as our dominant economic system and the value or lack thereof it places on anything other than profits. Kate Raworth’s Doughnut Economics provides the most cogent analysis on how the flawed models of free market capitalism have led us to this point. However, outside of our choice of economic system, incomplete models are also the result of a historical lack of capability to accurately simulate how the world works in a way that can scale. As we move further into the 21st century this part of the incomplete model problem can more readily be addressed.
We’ve past the point of being able to return to indigenous “no model” ways of knowing due to genocide and current size of the human population on Earth. We also can no longer rely on the incomplete models that have brought us to this climate precipice. What’s left is to build complete models that account for all the interaction effects in a system and that work on multigenerational time scales. The only advantage that simple models and dirt cheap fossil energy have left us is computing technology. That’s everything from pea sized smart sensors to farm sized data centers. Prehistoric humans learning to cook food resulted in less energy going to the stomach for digestion and more energy going to the brain for thinking. Similarly, we can view reliance on fossil energy for everything from farming to air conditioning resulting in less effort needed for survival and more effort available for creating computing technology. Despite cornering ourselves into a climate crisis, our advances in computation are already serving as the building blocks for new models that require no footnotes. Indigenous tribes considered all aspects of nature because no models were available, we will now consider all aspects of nature because any model can be created.
The models to come must center on how to live within the thresholds of Earth’s natural systems. This means creating the tools to understand ecosystem level interactions as well as molecular level interactions. Much of this work has begun but an overarching vision has yet to tie it all together. We must make sense of the world anew, in a way that doesn’t ignore what we find too complex or too inconvenient. And to do so we must leverage technologies ranging from cameras counting fish in the water to computer simulations of bacteria. Industrial agriculture must yield to precision and regenerative agriculture. The future involves applying research from soil microbiology and mycology instead of applying more nitrogen onto our infertile lands. We have to remember what was once known before: Everything is connected to everything else. It’s all one system and we should see ourselves as the orchestrators not the dominators. In practice this is revenue forecasts accounting for river run-off or demand planning accounting for desertification. That’s what it means to make models without footnotes. And while humans may not be able to comprehend such levels of complexity, machine learning algorithms may be well suited to do just that. Given there’s no going back, the only remaining path is to go deeper down the rabbit hole. To construct models that reflect reality so closely that we return to observing the world for what it is, not as a loose interpretation locked in a spreadsheet.
There’s a popular piece of business advice that to get started in a new industry one only needs to know enough to be a "little dangerous". In reality knowing enough to be a little dangerous is extremely dangerous. We’ve seen this play out over the last 200 years with our planet as the test subject. Not knowing anything or knowing everything forces us to respect the entirety of the planet. The middle ground of incomplete information is the worst possible scenario and one that we've been stuck in for far too long. When we look back we should be outraged that we ever had footnotes listing out all that we let ourselves ignore.