Good method includes knowing when to stop a model from talking. If the assumption set no longer matches the situation, the model may still produce a number but not a trustworthy inference.

Decision drill: choose the label that best fits each case.

A frictionless ramp is used to isolate how the angle changes the component of gravity along the slope.

An ideal gas fit that worked near room temperature is assumed to stay accurate near condensation without checking whether the assumptions still hold.

A planet is inferred from periodic dimming and star wobble rather than by grabbing the planet and testing it directly.

The discipline here is not avoiding models. It is admitting where the model ends and where a new one or new evidence has to take over.

Scientific method