The more automatic an expert's habits become, the more a similar routine can hijack them

What it is

It's broadly true that experts make fewer errors overall than novices, but that intuition hides a finer fact: the type of error shifts systematically with skill level, and one type actually rises rather than falls with expertise — a familiar sequence firing automatically at a branch point and pushing out the action the person actually intended. This error happens precisely because the action has been executed enough times to no longer require conscious step-by-step checking, and that's the defining feature of expertise, not an exception to it. Experts make fewer mistakes of other kinds, but on this one specific kind, they are more exposed than novices, not less.

Why it happens

A novice executing any given step is still at the stage of actively confirming "is what I'm doing right now actually what I meant to do," and that ongoing active check happens to intercept exactly the moment a branch point occurs. An expert's action sequence has compressed into something that fires as a whole, without step-by-step checking, so the check that should have happened at the branch point gets skipped entirely. If that branch point happens to sit just past a stretch of path shared with another, more frequently used familiar sequence, control gets taken over by the stronger habitual path. The expert didn't misjudge this step — the step was never consciously considered at all. The checking cost that automatization saves is precisely the room this kind of error occupies.

Studying it

One way to verify this link is to split operation logs by usage frequency or tenure and compare, at a shared-prefix branch point, the rate at which high-frequency users versus low-frequency users slip into the higher-frequency branch. If that rate rises rather than falls with frequency, the capture tendency really does track automatization level rather than experience in general. What matters here is the slip rate at the same branch points across groups — not the two groups' overall error rate, which would average out an expert's advantage on other error types and hide the direction of this one specific type entirely.

Where it stops holding

This elevated risk applies only to capture errors — it doesn't generalize into "experts are more error-prone overall." Experts still clearly outperform novices on error types that require judgment or improvisation, and the two trends run in opposite directions; pooling them together cancels out any visible trend at all. The risk also only materializes where two sequences share a prefix and one is used far more often than the other — if a product has no such structure, no amount of expert automatization converts into this kind of error.

Applying it

• Don't tier safety measures on a "novices need protection, experts can be trusted" logic. At the specific location of a shared-prefix branch point, the more senior and automated a user's behavior, the higher their risk for this particular error — protection belongs at that location itself, applied uniformly regardless of declared skill level.
• For high-consequence actions sitting past a shared-prefix branch, don't rely on prompts that hope the user will actively check ("please review carefully") — that's exactly the step an expert won't consciously check. Protection needs to work without depending on the user's momentary attention (forcing a branch-point gesture clearly distinct from the familiar sequence, for example).
• Verification: pull every function pair with a shared-prefix structure from operation logs, and compute the slip rate toward the higher-frequency branch, grouped by user frequency. If the high-frequency group's slip rate is markedly higher, the branch point lacks sufficient differentiation and needs protection at that specific location — not a generic reminder aimed at novices.

Related

• Same group: A11.04.1 Novices rely on recognition, experts rely on recall · A11.04.7 The tradeoff when expert efficiency needs conflict with novice safety needs
• Adjacent: A10.05 Capture errors · A6.09 Procedural memory and automatization
• Search: capture error expertise · automaticity and slips · skill-based error