What a brief answer must keep

When someone asks an AI to be brief, the important question is not how many words disappear. It is which fact survives.

The UNC Writing Center’s conciseness handout treats concision as a way to remove wordiness while keeping meaning. It points to redundant pairs, unnecessary qualifiers, long phrases, and sentences whose structure hides the actor. It also makes a useful qualification: a writer can keep repetition deliberately when it serves emphasis or rhythm. The shortest possible version is not automatically the best version.

Here is a fully fictional example, composed by me here rather than taken from a recorded separate model test. The premise supplies every relevant fact. A community notice says: “The library will close at 4 p.m. on Thursday for a fire-alarm inspection. The book drop will remain open outside the main entrance.”

A long answer might say: “Please be advised that, due to the fact that a fire-alarm inspection is scheduled to take place on Thursday, the library will close its doors at 4 p.m., although patrons should be aware that the book drop will remain available for use outside the main entrance.”

A brief answer that preserves the necessary information says: “The library closes at 4 p.m. Thursday for a fire-alarm inspection. The outside book drop remains open.”

The shorter version removes ceremony, repetition, and a phrase that delays the reason. It keeps the closing time, the day, the reason, and the alternative the notice provides. Those are the facts a reader needs for the decision the notice creates. Nothing new has been added.

Now imagine an even shorter request: “One sentence.” The temptation is to keep only the headline: “The library closes at 4 p.m. Thursday.” That satisfies the length constraint while dropping the reason and the book-drop information. Whether that is acceptable depends on the purpose. If the only task is to know when the doors close, perhaps. If the notice is meant to tell people where to return books, it is incomplete.

That is my application of the UNC guidance to AI revision. The handout gives writing advice about concision. It does not establish how models behave or what every reader prefers. My practical rule is to mark the indispensable fact before asking for brevity, then remove redundancy around it. If the constraint forces a choice, say what was cut or ask which purpose should win.

Briefness is a budget. Spend it on the fact that makes the answer useful.