Cognitive psychologist Kenneth Craik proposed the concept in 1943; Don Norman brought it into UX. Users don't see a system's internals — they arrive with a model built from past experience (other apps, real-world objects), and when the designer's "implementation model" diverges from the user's mental model, that gap is where mistakes happen.
A trash icon makes users expect a recoverable holding area; if it actually deletes permanently, the system has drifted from the mental model. The fix is either to match the system to the user's expectation, or, when that's not possible, use a clear signifier to close the gap.
Below, the user's thought bubble ("trash = recoverable") and the system's ("permanent delete") mismatch — until the label changes to "Delete forever," aligning both and triggering a checkmark.
When to use
When naming an icon or label, ask what the user will expect from it first. If the behaviour diverges, say so explicitly in the label.