The Biology of Economic Systems
Biology
Biology is so amazing. But in all of its amazingness, there's this daunting feeling that this is way too complex of a system for us to be able to do anything with. That's why biology has historically been very... "qualitative" is how I've heard it described. This is made possible by the fact that it's possible to perform experiments with statistically justified conclusions. One has many samples on which to perform experiments: in clinical trials, we can take real copies of the system that we're trying to understand (humans) and test a proposed intervention.
MIT has this very cool group working on "The Physics of Living Systems." Trying to apply tools from non-equilibrium statistical mechanics to model biological systems. Actually, it almost seems like non-equilibrium statistical mechanics is synonymous with biophysics type things nowadays.
I'm a big fan of this direction. It'd be beautiful if we could start to say some more principled, theoretically rigorous things about the emergence of such phenomena.
Economics
Contrast this with economics. Although it studies systems whose complexity rivals that of biological systems (more complex in its constituents, although perhaps less complex, at least as of now, in its organizational structure; although the "meta"ness and adaptiveness of it all I think makes it an even more difficult problem than biology) has historically used a much more mathematical approach. Studying very simplified models of behavior, and deriving mathematical theorems.
I believe one reason for this is just out of necessity. (1) There are not that many economies for one to observe, and (2) there's no real possibility of conducting a controlled study.
It's like nutritional science, but even worse.
Making Economics more Biological
However, I believe that the future of economics, as with biology, lies somewhere closer to where biology is now.
The article Why Agent-Based Modeling Never Happened in Economics – Economist Writing Every Day examines several reasons why agent based modeling didn't take off in economics.
I think another reason is perhaps just the fact that the real people, and their real interactions, are too complex for any simplified rules based model to accurately capture. Similarly to how in protein folding, rule based simulation mechanisms failed.
However, I think the example of protein folding motivates a broader revolution in the sciences: one in which we replace rule based models with learned models.
In the case of economics, perhaps LLMs are a decent starting point.
There's evidence to suggest that LLMs do accurately represent many of the intricacies of human behavior, which have thus far been the boon of behavioral economists and the bane of other economists.
For instance, in Using Large Language Models to Simulate Multiple Humans and Replicate Human Subject Studies, the authors replicate the empirical distributions of several psychological studies, such as the ultimatum game and milgram shock experiments.
Maybe, economics has been too "physicized" out of necessity. And the solution to the oft repeated complaint that economic models are both wrong and useless is to take a more biological approach.
Summary
- Biology and economics have the same goal: design interventions to cure diseases
- Biology has traditionally been a more experiment driven and qualitative field. This is made possible by the fact that, well, we can conduct experiments on biological systems. (1) There are many of them, and (2) we can control them.
- On the other hand, economics has traditionally been a more formalized field. This is not necessarily because economics is any more formalizable than biology, but I believe partially just out of necessity: it's near impossible to actually conduct controlled experiments for economic interventions.
- I predict more of a convergence between these two fields in terms of techniques:
- The physics of living systems approach to biology will lead to discovery of more fundamental laws that can lead to mathematically rigorous theories and predictions.
- And I propose a biology of economics systems approach to economics, whereby we develop agent based models with learned agents rather than rule based agents as their fundamental building blocks. This is motivated by the fact that people and interactions are way too complex to be captured by any simple homo economicus type rules. And behavioral economics can only do so much to patch things: there's infinitely many ways that humans deviate from perfectly rational, that my guess is it's pretty much impossible to capture in a rule based system.