Recurring errors of the same kind point to a faulty mental model
Aliases: error diagnosis · mistake identification
What it is
When a user makes the same kind of operational error across different tasks and different sessions — not a random misclick, but the same piece of reasoning applied in the wrong place — this is usually not carelessness but a sign that the underlying mental model is wrong. Human-error research separates this out as a mistake, as opposed to a slip: a slip is a correct plan executed badly (a misclick), while a mistake is an incorrect plan executed exactly as intended — the user did exactly what their understanding told them to do; the understanding itself was wrong.
Why it happens
Once a mental model forms, it gets invoked repeatedly to generate the next move: faced with the same kind of situation, the model produces the same inference every time. If the model is wrong, that inference is wrong in the same way every time — and that is exactly what makes it diagnosable. Stability and repeatability are the signature of a model error, not the severity or raw frequency of any single incident. A one-off slip can come from a lapse in attention, but the same faulty logic recurring in the same kind of situation means a persistent rule is being invoked underneath.
Studying it
- Search logs or session recordings for the same action sequence recurring, especially cases where an action is rejected, errors out, or gets undone — and the user then retries with the same approach. A single occurrence for one user proves nothing; the same user hitting it across contexts, or multiple users hitting it at the same step, is the credible signal.
- The key diagnostic difference: after a slip, users typically notice and self-correct almost immediately ("oh wait, wrong button"). After a mistake, users usually complete the action believing they did it correctly, and only look surprised once they see the outcome — the timing of that surprise is more diagnostic than the error itself.
- After a user hits an obstacle, let them try again and watch whether they switch strategy or repeat the very approach that just failed. Repeating the same failed approach points more strongly to a model error, since a slip is usually avoided on retry while a faulty model gets invoked again and produces the same wrong inference.
Where it stops holding
- A single error is not evidence — anyone can misclick. Only a pattern recurring across contexts and over time points to a model problem.
- Low frequency of a given error does not mean the underlying model is correct; the path that would trigger it may simply be rarely taken, so the model has never actually been tested there.
- High error rates during onboarding are not necessarily model errors — they can just be unlearned steps. In that case users tend to try different approaches on retry rather than repeating the same logic, so this should be cross-checked against how users narrate their own actions.