Session feedback scale: replace Bad / Fine / Good with a four-point satisfaction scale

This has been reported at least twice — #75198 (Fine vs Good ambiguous for non-native speakers) and #68085 ("Good" grammatically mismatched with the question) — and both were closed for inactivity rather than on the merits. Filing it once more as a design proposal, because the two reports share one root cause: the three labels were chosen as loose evaluative words rather than as points on a defined scale.

"How is Claude doing this session?" asks for an adverb — well, badly. "Good" is an adjective, so it does not complete the question. But the deeper problem is that the three options are not on one dimension at all:

• "Bad" and "Good" are quality judgements; "Fine" expresses acceptance. It answers a different question. The set is therefore not monotone, which is why ranking "Fine" against "Good" feels arbitrary in any language, not only to non-native speakers.
• A three-point scale has no neutral exit, so the middle becomes one. It absorbs both "slightly dissatisfied" and "slightly satisfied". Those cancel out, and the aggregate is hard to act on.
• A three-point scale fits the seller, not the user. Small / medium / large works because every option is acceptable to the one offering. A satisfaction scale needs an option that means "no". "Bad" is formally that option, but it reads strongly enough that people avoid it — so in practice the scale collapses to Fine or Good, and "Fine" gets used as politeness rather than as a measurement.

Proposal: change the question and the scale together, so that they agree both grammatically and semantically:

How satisfied are you with this session? Not satisfied · Partly satisfied · Largely satisfied · Fully satisfied

That is monotone, it has no midpoint, it asks how much of the need was met rather than how the user feels — and every option is a grammatically correct answer to the question asked, which resolves #68085 as well.

For context: I have used this four-point form for years working with quality in the automotive industry, where a rating scale that cannot separate "acceptable" from "good" is not usable as an input to corrective action.