The first question to ask of any model is not “is it simple?” or “is it elegant?” but what kind of work is this model trying to do? A model built to help you intervene in the world should compress aggressively. A model built to help you understand a world you do not yet know how to act on must often remain messy.
Simplicity is a virtue only after purpose has been named.
Simple Picture
You can model an airplane three ways. If you want to estimate range, treat it as a point mass with drag. If you want to know whether the wings will tear off, model the aeroelastic structure. If you want to paint it, model the surface area and livery.
None of these is “the” model of the airplane. Each is a model of the airplane for a purpose. The same object demands different simplifications depending on what you intend to do with it.
The trap is importing the engineering instinct for simple manipulative models into situations where you do not yet know what manipulation is appropriate.
The Two Kinds of Model
Venkatesh Rao’s compressed test is:
how you model something depends on what you want to do with the model
Manipulative models are for doing something specific. They exist to support intervention: estimate, steer, optimize, troubleshoot, persuade, navigate, repair. Their quality is measured by whether they are useful for the intended action. Extra realism is usually a bug because it adds degrees of freedom that do not improve the move.
Appreciative models are for making sense of something on its own terms. They exist before the action is clear. Their quality is measured by whether they create a satisfying, fertile image of the world: a frame that lets you notice more, ask better questions, and later derive more precise manipulative models.
Ode to a Flower is the miniature version: the scientific model is not valuable because it lets you control the flower, but because it lets the flower become more visible as itself.
This is the distinction behind instrumental vs appreciative understanding. Instrumental understanding lets you function inside a system. Appreciative understanding lets you grasp what the system is. The former lets you board the plane. The latter lets you feel what flight means.
Fertile vs Seductive
The danger with appreciative models is aesthetic capture. A model can be elegant, intuitive, and beautiful while going nowhere. It can give the feeling of insight without opening new territory.
A good appreciative model is fertile. It suggests lines of attack. It exposes new distinctions. It makes adjacent questions visible. It does not have to be simple. It does not even have to be pretty. It has to make reality more available to thought.
A bad appreciative model is seductive. It feels profound because it compresses awkwardness away. It is easy to remember, easy to repeat, and sterile in use. This is the danger named in beauty as truth-signal: coherence can be manufactured by deletion.
Elegance is a bonus when it happens. Fertility is the real test.
Why Simplicity Advice Fails
“Make the model as simple as possible” is good advice for manipulative work. Once the purpose is fixed, every irrelevant detail is drag. The model should preserve only the variables that matter for the move.
But in the appreciative phase, the purpose is not fixed. The work is not yet intervention; it is perception. Premature simplification becomes premature deletion. You remove precisely the strange, ugly, inconvenient details that might have revealed what the situation is.
This is why exploration feels inefficient. Exploration increases the entropy of the mental model before it earns the right to simplify. The person who demands elegance too early is often demanding emotional relief, not understanding.
The same move appears in meta-rationality: before solving the problem, change the description of the problem. Appreciative modeling lives upstream of problem-solving. It decides what kind of world the later manipulative model will be allowed to see.
The OODA Loop Problem
Model form can mislead. A diagram, loop, taxonomy, or matrix can look like manipulative knowledge because it resembles a tool. But the test is:
what is this model for?
If the answer is not obvious, the model is probably appreciative. Boyd’s OODA loop is drawn like a process diagram, but its real value is not that it tells you step-by-step what to do. Its value is that it gives you an appreciative frame for decision-making under uncertainty: observation, orientation, decision, and action are not a checklist but a way of seeing tempo, surprise, and adaptive advantage.
This matters because bad users convert appreciative models into fake manipulative recipes. They ask the model to generate procedure before it has generated perception. The result is cargo-cult strategy: fluent diagrams, no contact with reality.
Transformation
Appreciative models can become manipulative models by pruning scope and adding practical details. A rich model of an airplane becomes a flight-range calculator once you throw away everything except the variables relevant to range. The manipulative model may include arbitrary constants that add no appreciation but make action possible.
The reverse is harder. A manipulative model becomes appreciative only when someone extracts insight from it, removes arbitrary implementation details, and generalizes inductively. Pythagorean triples were useful in practical geometry before the Pythagorean theorem made the pattern appreciable. The theorem was less immediately worldly than the builder’s trick, but far more generative.
This is why naming the structure of a problem matters. A taxonomy can be manipulative when it tells you which intervention to try. It becomes appreciative when it helps you see problem-space itself differently.
Dimwit / Midwit / Better Take
The dimwit take is “simple models are better.”
The midwit take is “complex reality requires complex models.”
The better take is that model quality is purpose-relative. A manipulative model should be no more complex than the intended move requires. An appreciative model should be as complex as the situation requires to become visible. The real failure is not simplicity or complexity. The real failure is using the wrong criterion because you have not admitted what kind of model you are building.
Main Payoff
Before criticizing a model, locate it on the appreciative-manipulative spectrum.
If it is manipulative, ask: does this help with the move it claims to support? If not, simplify, delete, parameterize, operationalize.
If it is appreciative, ask: does this make the world more fertile to think with? If not, elegance is irrelevant. A beautiful model that opens no doors is decoration.
The successful thinker keeps both modes alive. Appreciation without manipulation becomes private aestheticism. Manipulation without appreciation becomes optimization inside a shrunken world. The art is sequencing: appreciate before you manipulate, then manipulate without forgetting what your simplifications cost.
References:
- Venkatesh Rao, Appreciative versus Manipulative Mental Models, Ribbonfarm
- John Friedmann, Planning in the Public Domain