Part of an occasional series: ideas from outside the usual solarpunk reading list.
There’s a word that gets used almost interchangeably for two genuinely different things, and the confusion isn’t harmless — it leads people to use the wrong tools, misapply research from one field to another, and design interventions that backfire. The word is „complex,” and a 2026 paper by Shutters and Dhein, published in Frontiers in Complex Systems, spends a short, useful article just pulling the two meanings apart.
The distinction, stated plainly
A complex system is one with many interacting parts, nonlinear behavior, and outcomes that emerge from those interactions rather than being programmed in from the start. The classic example is weather: a thunderstorm or a jet stream emerges from the physics of fluids and heat, produces genuinely surprising, hard-to-predict patterns — and none of the air molecules involved are „trying” to do anything or „learning” from what happened yesterday.
A complex adaptive system is a subset of the above, with one extra ingredient: the components, or the system as a whole, learn and change their behavior based on experience. A city is a complex adaptive system. So is an economy. So is your immune system. In each case, the agents involved — commuters, firms, antibodies — actively adjust their rules of behavior in response to what’s happening around them and to each other.
Every complex adaptive system is a complex system. Not every complex system is adaptive. That’s the whole distinction — but it turns out to matter a great deal in practice.
Why the mix-up isn’t just semantic
Three concrete places the distinction bites:
Choosing the right model. If you’re modeling a system where nothing adapts — the spread of heat through a metal plate, say — a system-dynamics model (a set of equations describing flows and stocks) works well. If agents in the system are adapting, you generally need something closer to agent-based modeling, where you actually simulate individual decision-makers changing their behavior. Use the wrong tool and you’ll get numbers that look precise and mean very little.
Reading across fields. Techniques and findings from physics — where „complex” almost always means non-adaptive — get imported into economics or epidemiology, where the systems in question are adaptive through and through. That import can be useful, but treating findings from a non-adaptive domain as if they transfer cleanly to an adaptive one is a quiet, easy-to-miss error, and it happens often enough that the authors think it’s worth naming.
Designing interventions. This is the sharpest practical stake. A complex system that isn’t adaptive doesn’t fight back. A complex adaptive system does. Introduce an antibiotic, and pathogens evolve resistance. Introduce a new financial regulation, and market actors find the loophole. In a CAS, an intervention doesn’t just have an effect — it changes the rules of the game going forward, because the agents inside the system are watching, learning, and responding strategically to the very thing meant to redirect them. Plan an intervention as if you were dealing with weather, and you’ll be blindsided by a system that adapts around your fix.
What it doesn’t tell you
It’s worth being precise about the limits here, because it’s tempting to reach for this distinction and make it do more work than it can. Shutters and Dhein are explicit that CS-versus-CAS is a practical, methodological distinction — a tool for picking the right modeling approach and avoiding a specific, common category error. It is not a claim about consciousness, agency in some deeper philosophical sense, or moral status. A thermostat that „adapts” to room temperature and a city that „adapts” to a housing shortage are both, in this narrow technical sense, exhibiting adaptive behavior — but nobody should conclude from that alone that the thermostat and the city therefore have comparable inner lives. The distinction earns its keep as a modeling choice, not as evidence about anything deeper.
Why it’s worth having in your pocket
Once you have the CS/CAS distinction cleanly separated, you start noticing it everywhere, and noticing where its absence causes trouble. A lot of public arguments about how to „fix” some large, adaptive social system — a housing market, a platform’s content-moderation problem, an epidemic — implicitly treat the system as if it were inert, as if the intervention were the last word rather than the opening move in an ongoing back-and-forth. The people, institutions, and markets on the receiving end of the intervention are rarely inert. They notice. They adjust. Sometimes they adjust in ways that undo the intervention’s whole point.
None of this requires new data or new technology to apply — it’s a question you can ask about any proposed fix to any complex problem: is the thing I’m trying to change going to notice what I did, and change its behavior in response? If yes, you’re in CAS territory, and the fix needs to be designed with that strategic pushback in mind from the start, not discovered after the fact when the pushback shows up.
Did we get something wrong?
This piece is drawn from our own reading notes, not a translation of the paper. If something above misrepresents the argument, or just sounds off, tell us — publicly or privately, whichever you’d rather.
Want the fuller picture?
This is a survey, not the whole paper. The deep-dive version, with more of the argument and the sourcing behind it, is here: deep-dive PDF.
What should we read next?
This is one piece in a loose series drawn from our own research reading. If there’s a thinker, a paper, or a tradition you think belongs in a series like this, we want to hear it.
