A sufficiently complete mental model lets a user derive operations they were never taught

What it is

When a user's model is complete enough, they can work out an entirely new combination of operations without anyone teaching it to them — nobody ever said "you can drag multiple selected files at once," but because the user understands the mechanics of "selecting" and "dragging" separately, they combine the two on their own. This capacity to spontaneously discover untaught operations is a direct sign of how complete a model is, not a lucky guess.

Why it happens

A complete model doesn't memorize every individual feature separately; it captures the deeper rules that generate those features — which kinds of objects can be selected, and how a selected object behaves as a unit in the next action. Once those rules are grasped, a new combination is simply what falls out of arranging the rules together — the user doesn't need to be told about every possible combination, only to apply rules they already have to a situation they haven't encountered before.

Studying it

One diagnostic is to design a test item around a combined operation that was never taught, then observe whether users can discover or guess it with no hints at all. If most users independently arrive at the same untaught operation, that indicates the system's underlying rules are consistent enough to be widely internalized — which tests both how complete users' models are and, indirectly, how clear the system's own rule design is.

Where it stops holding

This kind of derivation requires the underlying rules to actually be consistent, with no hidden exceptions — if the same operation follows different rules in different contexts (selection behaves as a unit here but item-by-item there), users who derive a combined operation often derive it wrong, producing a confident but incorrect prediction. This kind of "looks complete but is actually undermined" model is harder to spot than an obviously incomplete one, because when users hit a failure they tend to blame their own actions rather than suspect that the rule itself is inconsistent across contexts.

Applying it

• When designing interaction rules, keep the same class of objects following the same underlying rule across contexts, and avoid carving out one-off exceptions for local convenience — this is the precondition for a user's model to be genuinely generative.
• Before shipping, sample a few combined operations that never appeared in documentation or onboarding, hand them to new users with no hints, and record whether they're discovered spontaneously; this tests both whether users hold a generative model and, in reverse, whether the system's own rules hide an inconsistency.
• How to check: give the same batch of untrained new users a combined scenario that was never demonstrated, and measure the share who produce a correct prediction on their own; a low rate despite otherwise smooth prior performance usually points to an unnoticed exception in the underlying rule, not a shortfall in user ability.

Related

• Same group: A7.01.1 A mental model is a user's internal explanation of how a system works · A7.01.2 A model can be incorrect and still support successful operation · A7.01.3 The model determines a user's expectations and response when something goes wrong · A7.01.4 The core function of a mental model is predicting system behavior for a given action, not memorizing steps · A7.01.6 A model's internal consistency and its accuracy are two separate things — a consistent but wrong model still works · A7.01.7 Users lack insight into the accuracy of their own model and rarely question it
• Nearby: A7.15 Model layers: functional, structural, and implementation · A7.13 Cross-product transfer of mental models
• Search terms: generative mental model · rule consistency · discoverability