The ultimate hidden truth of the world is that it is something that we make, and could just as easily make differently.
— David Graeber
Systems Realism is a philosophical and analytical frame for thinking about human systems, which has two key precepts:
The "rules" of any system will irrevocably determine what effects it has.
If the system was designed, then the intentions of the designer are irrelevant.
This is just a restatement of the systems thinking adage that the purpose of a system is what it does.
Some conclusions follow from this precept:
- It is impossible to durably solve any problem if some structural change doesn't happen. The structural change could be small, but it is unavoidable. Otherwise the "gravity" of the system will take over and the problem will reemerge.
- Some problems aren't worth the difficulty and risk necessary to durably fix them.
- Designing systems is an inherently technical and creative activity. There is such thing as "good" vs "bad" systems designs, where good ones create outcomes that accord with our values and bad ones don't. Like any discipline of science, our ability to design systems can improve through empirical and theoretical iteration.
- Designing good systems can be difficult and time-consuming, and is worth investing in as a society.
- The system design that would be "best" for a situation might be completely different than what is already in place. Historical precedent is meaningless and path-dependent.
Human systems always boil down to promises.
An individual human can only effect the world through their own physical actions, which are extremely limited. In order to make large effects on the world, humans must work together to coordinate their actions.
The only way a human can influence the choices of another human is to make promises.
Promises are what underly laws (if you take an object from someone without their agreement we promise we'll punish you), marriages (if this person tells us they don't like you anymore we promise to help them take some of your shared property), borders (if your people cross these lines on the map we promise we'll arrest them), and every other human system.
"Rules" are just promises to take some action in particular circumstances.
Taken together with what we have empirically discovered about human nature, some more conclusions follow about how to design successful systems:
- A system must seem credible to its participants. If people don't believe some promises will be upheld they'll ignore them.
- A system must actually be credible. Perception isn't always reality, and eventually someone will call your bluff.
- A system must contend both with participants' self-interest and their seeming altruism. It has to simultaneously promise to benefit them without upsetting their moral and social convictions.
- A system must contend with participants' tendency to seek social status, regardless of other benefits.
- A system must contend with participants' bounded rationality, their penchant to use heuristics and shortcuts.
- A system must be understandable and simple enough for participants to actually follow it given their limited time and energy.
Systems don't have to suck
As difficult and annoying as designing and implementing systems of promises can be, doing so is often the only way to solve our most important problems.
It is arrogant and foolish to assume we can design perfect systems, ones that will always do the right thing or automatically fully correct themselves. But this doesn't mean we should give up.
Systems design is just like any branch of science: theory informs experimental design, and experimental result informs new theory. This virtuous cycle can dramatically improve the world we live in if we pursue goals that actually align with our values.
Systems Realism allows the possibility things could get better, whereas fatalism guarantees failure.
Aphorisms
These are "catchphrases" meant to help people gain intuition for Systems Realism, remember it in moments where it is useful, and persuade others of its truth.
- Human systems always boil down to promises.
- Systems don't care about your intentions.
- Reality doesn't care about your intentions.
- Systems don't have to suck.
- Fatalism guarantees failure.
- It's possible to beat prisoner's dilemmas.
- Structure equals function.
- Every human system was designed by someone.
- Human systems can be designed and changed.
- A better world is possible.
- If they don't believe it they won't do it.
- There's no such thing as human error.
- Just 'cause that's what we've always done doesn't mean that's what we should do.
- Hope isn't a strategy.
- Diligence isn't a plan.
Heuristics for good systems design
Systems should be self-correcting
This is most often achieved with balancing feedback loops.
Self-correction is why Democracy is in general such a good system concept. A group of people having control over the system itself allows them to correct it when it fails them.
Reinforcing feedback loops are powerful but dangerous
Reinforcing feedback loops can sometimes be desirable, such as if they're generating some genuinely desirable quantity or extinguishing some undesirable one.
But any runaway feedback loop tends to become pathological at the extremes. If we don't intentionally design reasonable "ceilings" or "floors" ourselves, some higher-level system will eventually violently intervene. If something cannot go on forever, it will stop.
See also "Success to the Successful"
Use rather than fight human nature
You have to work with human nature, not try to avoid or overpower it. Systems that tend to acknowledge and dance with our flaws will tend to fair better than those that don't.
The best systems work across social contexts
Human societies can have extremely stark differences that make the same system design behave very differently.
In general it is better if a system is fluid and adaptable enough that it can work in many different cultural environments. Even more so if it can operate across contexts simultaneously, helping to bridge their differences in real time.
Continuous is better than discontinuous
Any time a system has "discontinuities" (e.g. discrete tax brackets) distortions in behavior will tend to appear at these discontinuities.
Often discontinuities are unavoidable, since they relate to some real constraint of the world (you can't assign 5.63 jury seats, or give someone two tenths of a college slot). But in general if it's possible to design a system in terms of smooth continuous functions, that will tend to yield smoother real world outcomes.
Arbitrary things are arbitrary
Arbitrary numbers or rules (e.g. a $1000 damage boundary between misdemeanor or felony) tend to both undermine the legitimacy of a system and reveal it doesn't have a coherent design.
In general the design of the system should be structurally implied by its intended goals. This can be very difficult to achieve!
Holism is better than fragmentation
If a system has slowly "accreted", and is now an incoherent jumble of unrelated ideas and patches on patches, then the contradictions between these concepts will cause drag, waste, and failures.
It is difficult to redesign systems to incorporate all the ideas together and smooth out the system, but it is possible.
Iterative change is easier than disruptive change
It isn't impossible to change a lot very quickly, but it is very hard. Sometimes it is only doable in moments of genuine crisis.
Small systems are easier than big systems
It's usually easier to get a small thing working and scale it up than to get a big thing working from the beginning.
Systems must have social legitimacy
A system is by definition not good if people don't like it, even if they aren't being rational. The environment of social preferences and biases is part of reality: feelings are facts.
You either have to change the social preferences and biases through some higher-level system, or you have to design the system to accommodate them.