What separates violation from mistake is whether the deviation was intentional

What it is

Violations and mistakes look alike on the surface — both are "doing something you shouldn't have," and an outside observer will often describe either one as "the user got it wrong." The one criterion that actually separates them is whether the person knew, at the moment of acting, what the correct procedure was. Believing the deviant action is the correct one is a mistake. Knowing what's correct and choosing not to do it is a violation.

Why it happens

This criterion matters because working backward from an outcome to its cause is easy to get wrong: the same "skipped a required step" action might happen because the user genuinely didn't know the step existed or mattered (mistake), or because the user knew about it, had done it correctly before, but deliberately skipped it this time to save time (violation). The action and the result alone can't tell these apart — only reconstructing what the person actually knew about the correct procedure before acting can. This is also why the criterion has to rest on intent rather than outcome or frequency: a deviation that recurs frequently could be the same wrong belief firing repeatedly (meaning the belief was never corrected), or the same violation happening again and again (meaning the motivation never changed). Frequency alone cannot substitute for verifying intent.

Studying it

The most reliable way to verify a person's intentional state is to check whether they had actually encountered and mastered the correct procedure before acting — training records, a history of correct execution, or an interview taken before rather than after the incident. Post-hoc questioning is easily contaminated by self-justification: "I always knew the right way" and "I didn't understand it at the time" can both be statements shaped by the situation the person now finds themselves in. Absent independent evidence that the person knew the correct procedure beforehand, a post-hoc claim alone shouldn't be used to label something a violation.

Where it stops holding

Intent isn't a clean binary — there's a middle ground where a user has only a vague sense of a rule, aware that "there's probably supposed to be a step" but unable to say what it is. This half-knowing state resists being cleanly filed under either mistake or violation, and forcing a classification can do more harm than admitting "insufficient evidence, undetermined for now," because a forced classification steers the wrong intervention downstream.

Applying it

Turn this criterion into a mandatory question at every retrospective: "before acting, did this person clearly know the correct procedure and have the ability to perform it?" A "no" answer routes the case to mistake and the model-correction path; a "yes, but didn't do it" answer routes it to violation and the motivation-and-institution path; an "unclear" answer means the evidence isn't there yet — gather more before classifying, rather than going with first impressions. Verification: sample a batch of records already labeled as violations and check, one by one, whether there's independent evidence the person knew the correct procedure before acting. If a substantial share turn out to rest only on post-hoc attribution with no prior evidence, that reveals a systematic bias toward misclassifying mistakes as violations — a bias that channels intervention toward disciplining people who never knew the correct action, which accomplishes nothing.

Related

• Same group: A10.01.1 Slip — correct intention, deviated execution · A10.01.2 Mistake — the intention rests on a wrong mental model · A10.01.3 Violation — deviating on purpose despite knowing the rule · A10.01.4 Slips are fixed by interface constraints, mistakes by model correction — the remedies are not interchangeable · A10.01.5 Classification must precede any redesign · A10.01.6 Blaming "user carelessness" halts diagnosis
• Nearby: A10.13 Violations and non-compliance · A7.14 Diagnosing and correcting wrong mental models
• Search terms: intent criterion · mistake versus violation · hindsight bias