we call this

How to spot it

The shape is a short compound, and it arrives already dressed as a term. An ordinary word, usually whatever the paragraph happens to be about, gets welded to a noun such as paradox, trap or divide that sounds like it came out of a journal. Sometimes a writer announces the move out loud, with a flat "we call this" running straight into the new name. More often the compound simply turns up behind a definite article, which does the same job more quietly, because "the" is a promise that writer and reader already know what is being referred to.

Nothing in the grammar tells you whether that promise is good. A genuine coinage and a reflex one are built identically, so reading the sentence harder will not settle it.

The test is in what comes next. Watch for the second case. Does the label return, land on a different example, and pick up a stated limit on how far it reaches? A name that comes back has been worked on, and the return trip is the work. A name that appears once and is never mentioned again was assembled for the length of a single sentence and abandoned there.

Two other things are worth a look. The article is one. "A scheduling trap" is a writer proposing something, while "the scheduling trap" is a writer reporting a consensus, and the distance between those two is an entire argument the sentence skipped. The company the phrase keeps is the other. Real naming is expensive, because it has to be paid for with at least a second case. Reflex naming costs one line, and the paragraph moves on as though the bill had been paid.

Why models overdo it

Producing "the onboarding trap" costs a model nothing extra over producing any other noun phrase. It is a short, fill-in-the-blank frame, and a system trained on a library of real coinages has the shape memorized cold. Checking whether the name deserves to exist is a different kind of work entirely, and nothing about finishing a sentence requires having done that work first.

The deeper reason is arithmetic rather than intent. Each word is picked for how well it follows what came before, and a compound that reads like a term everybody already knows follows beautifully. That is precisely the property that makes it useless as evidence. The more settled a phrase sounds, the likelier it is to be produced, and the likelier it is to be produced, the less its sounding settled can tell you about whether anyone settled it.

These labels read as confident rather than as wrong for that reason. The sentence is fine. The definite article comes free, but the second case has to be earned, and it rarely shows up.

When it's fine

Coining is sound practice whenever the name carries weight the sentence could not carry without it, which happens more often than a suspicious reader expects. Naming a thing makes it arguable, and a writer building a real case wants exactly that.

Susan Barnes brought the privacy paradox to social networking in 2006 and kept returning to it for the rest of the article. She set adults alarmed about collection against teenagers who post freely, then asked whether any of the proposed solutions would resolve the paradox. Graham Allison took up Thucydides's Trap, a phrase Herman Wouk had used in 1980, and backed it with a Harvard study of 16 cases of a rising power meeting a ruling one, 12 of which ended in war. Masahiro Mori's uncanny valley arrived attached to a dip he had already drawn on a graph, which is about as literal as earning a name gets.

None of the three used the label as a flourish. Each came back to the name after introducing it, rather than letting the phrase stand on its own. Give your own coinage the same test: if it never comes back for a second case, a plain description will do its job and usually read better anyway.

Where it comes from

The move has an older name than any of its instances. Philosophers call it reification, or hypostatization, which means treating an abstraction as if it were a concrete thing you could point to or measure. A coined label pulls a small version of that trick when it talks about an idea as though someone had already gone out and measured it. Rhetoric has no single label for the two-word compound itself, but linguistics has one for the wider habit it belongs to. A snowclone is a fixed phrase template with a slot left open for whatever a writer wants to drop into it, and the economist Glen Whitman proposed that name in January 2004, with the linguist Geoffrey Pullum endorsing it on Language Log the following day. A coined label is a snowclone narrowed to its shortest useful form.

George Orwell was complaining about the underlying shortcut in 1946, decades before linguists named the template, when he described a writer gumming together long strips of words "which have already been set in order by someone else" rather than working the claim out from scratch. What keeps the shortcut attractive is that it sometimes works. The digital divide and the tourist trap both began as somebody's fresh phrasing and walked into ordinary English one confirmed case at a time, and so did the three labels above. Every reflex coinage is betting that a reader will assume the same trip has already been made.