A useful model leaves things out on purpose. The difficulty is not omission itself but forgetting where the omission occurred. A clean diagram can make its boundary feel like the boundary of the world.

Companion demonstration draft · philosophy and models.

State the small world first

The tile experiment in The Shape of Possibility makes a limited promise: given independent, equally likely choices, it counts the chance of one specified arrangement.[^1] Its precision belongs to that small world. The arithmetic does not establish that every larger world follows the same rules.

One way to test the boundary is to alter a premise. Let a choice remember an earlier outcome, as in When Feedback Changes the Question, and the old calculation needs reconsideration.[^2] This is not a failure of explanation. It is the beginning of finding out what the explanation actually depends on.

Keep two kinds of confidence apart

We can be confident that an argument follows from its premises while remaining uncertain that those premises describe a particular situation. These are separate achievements. A beautifully drawn relationship is not, by itself, evidence of a mechanism.[^3]

The distinction becomes practical when someone must act. A model may help compare alternatives without being adequate to decide between them alone. We can ask what it captures, what it omits, and whose experience would reveal an omission.

That last question leads to The Care of Incomplete Evidence.[^4] A boundary stated plainly gives another person somewhere to begin a correction.