Giving absence a word

The machine's confident register is not a personality; it is an average. The example answers its shaping rewarded almost never said "I don't know" — so a blank in the pattern gets filled, fluently, by whatever completes it best. Ask for owners and deadlines, and a task that has neither gets both.

The countermeasure is almost embarrassingly small: give absence its own word. An instruction like where the window does not contain it, write UNKNOWN turns not-knowing into a completable pattern — an honest continuation that can win the draw. Without the word, honesty has no shape to complete; with it, the odds hold a real alternative to invention.

The convention across this site:

• UNKNOWN is the default absence word — for facts, sources, and anything else the window cannot support.
• UNASSIGNED is the same discipline fitted to a column: it marks an owner nobody was given. You met it in the first Lab 00 lesson.
• "no date" likewise, for deadlines nobody said.

The instruction tilts; it does not command. In the first lesson's x-ray you can watch the honest branch sit in the odds and lose even with the instruction present. The word does not guarantee honesty — it makes honesty possible, and over many runs that is the difference between a report you can audit and one you have to re-derive.

Say it back in one line: why does a model invent an owner when no owner is in the window — and what does giving it the word UNKNOWN change about the odds?