In this article I’ll be looking at the shape of this question regarding complexity that comes from Plato. In the Euthyphro dialogue, Socrates asks whether a thing is pious because the gods love it, or the gods love it because it is pious. If piety is defined by the gods’ love, it has no ground of its own; it’s only whatever the gods happen to prefer. If the gods love something because it’s pious, they’re tracking something that already existed before their love touched it. Their love adds nothing to it. Socrates never gets Euthyphro to close the gap.

Take a light switch. It has two states, on and off. Almost anyone finds it simple, because almost anyone’s needs are already covered by these two states.

A thermostat has more states than a light switch. It includes target temperature, current temperature, schedule, and override. Most people still find a thermostat simple because the manufacturer has already done much of the work of matching those states to what a homeowner is likely to need.

Now take that thermostat as part of a commercial HVAC system. It involves zone overrides, occupancy sensors, and a facilities manager trying to cut energy costs. The thermostat itself doesn’t acquire a new set of capabilities. What changes when a thermostat is part of a larger structure is the structure under consideration. Thermostat alone is one structure. Thermostat plus zones plus occupants plus controls is another. The domain of states relevant to regulation expands when the “system” scope expands. This is a first-order artifact.

W. Ross Ashby’s law of requisite variety is often applied to complexity. It says something narrower than people assume. Ashby’s law is about regulation specifically. A regulator can reduce the variety of a disturbance only to the extent that the regulator itself has sufficient variety in its available responses. Only variety can force down variety. Ashby’s law is a claim about regulators and disturbances. It’s not yet a claim about observers and complexity.

Stafford Beer treats variety, the number of distinguishable states a system can occupy, as a measure of complexity itself. What counts as the system—which states are distinguishable, and which differences matter—depends on how the system gets defined. How a system gets defined depends on who is doing the defining. Ashby supplies the constraint on variety. Beer supplies the complexity reading based on variety. Second-order cybernetics supplies the observer sitting underneath both Ashby and beer.

A thing isn’t complex simply because someone finds it hard to deal with. A person who doesn’t understand basic arithmetic can find a simple equation bewildering. A person’s difficulty doesn’t add variety to the equation. What’s missing in cases of subjective difficulty isn’t in the object, but in the distinctions that person currently has available for acting on it. The complexity that matters for regulation isn’t subjective difficulty. It’s the variety that must be distinguished and responded to for a given purpose, measured against the variety currently available to whoever is doing the distinguishing.

The homeowner’s simple picture of the thermostat can be perfectly adequate. The facilities manager’s elaborate picture can also be adequate. Neither the homeowner nor the facilities manager has discovered the true complexity of the thermostat lying underneath the other’s account. Both accounts are descriptions made viable by different purposes and different capacities for action. Viability, rather than correspondence to one fixed description of the device, is what tells you whether a given level of complexity suits the purpose at hand.

Heinz von Foerster’s aesthetic imperative is: If you desire to see, learn how to act. The facilities manager treats zone conflicts as a live problem because years of managing a building provided distinctions to act on. Someone without that background might see the same readings and alarms without seeing them as a zone conflict at all. The difference between seeing a zone conflict and not seeing it isn’t in what reaches the eyes. The difference is in which distinctions have become available for acting on what the eyes already see.

Von Foerster’s ethical imperative is: I shall act always so as to increase the number of choices. A checklist, a training manual, a policy, or an interface doesn’t just tell someone what to do. These tools determine which distinctions are handed to someone as distinctions worth acting on. Handing someone three permitted responses limits what they can do. In fact, handing someone three permitted responses might make everything outside those three effectively invisible as a problem requiring a response.

It’s tempting to picture the regulator as a single person, the facilities manager’s own head holding whatever variety they can muster. Beer never meant the regulator to be just a single person. The variety available to a regulator can live in training, in procedures, in instruments, in software, in other people, and in the whole organizational structure around the person. A novice with a good diagnostic procedure can regulate something an experienced person without one can’t. The term “observer” in this article should be read as shorthand for the observer and whatever variety they have been handed.

Some variety arrives as vocabulary. It’s tempting to say the facilities manager has language where the novice doesn’t, with terms like “zone conflict” sitting ready on a shelf. The picture of language as a structure existing on its own, independent of anyone using it, reintroduces the very thing this article argues against. Humberto Maturana’s term for the alternative to static language is “languaging.” Languaging is a recurrent, consensual coordination of behavior among people. Languaging is something done together rather than consulted alone. There’s no language sitting apart from the languaging that keeps regenerating it. The facilities manager and their colleagues aren’t retrieving “zone conflict” from a shared dictionary each time they use it. They are reenacting the distinction in every handoff and every report, keeping it alive and slightly recalibrated. The term is actionable for the same reason the thermostat’s complexity is never fixed. Both language and system complexity are constituted, not just applied, each time the coupling is tested again.

Ashby’s law of requisite variety is that only variety can destroy variety. Complexity was never simply a label sitting on the thermostat or the light switch, waiting there to be found. Not every difficulty is an invention of the observer. A novice can make a genuinely simple system look difficult because the distinctions needed aren’t yet theirs to use. A facilities manager can meet a real limit even with years behind them because the available responses still fall short of what the disturbance demands. Neither case is settled by declaring the complexity belongs to one side or the other.

None of this gets settled by argument. It gets settled the way pragmatism settles most things: in action, by what happens when the world pushes back. The novice finds out how much of the trouble was theirs the first time a guess fails and the thermostat keeps behaving exactly as it would have anyway. The facilities manager finds out how much belongs to the structure itself the first time a genuinely new configuration exposes a limit no accumulated experience prepared them for. What the resistance reveals isn’t some complexity sitting behind the scenes, waiting to be exposed once and for all. What the resistance reveals is where the current coupling between observer, description, and system has stopped being adequate and needs to change.

If you liked what you’ve read, please consider my book Second Order Cybernetics, available in hard copy and e-book formats. https://www.cyb3rsyn.com/products/soc-book

Published Oct. 5, 2026, in Harish Jose’s blog.