What’s funny is that native linguists don’t seem to agree on this topic.
The source I posted mentions this contradiction from authors mentioned in SE (Tanaka, Noda) and then compared another linguist (Kawaguchi) and stated:
Kawaguchi showed that the shift from masen to nai desu forms occurs more in declarative sentences than in interrogative sentences by using natural speech material. In the case of declarative sentences, the shift to nai desu forms is more likely if the emphasis is on negation. This also seems to be true in interrogative sentences, especially for ‘negative proposition’ type sentences where negation takes emphasis. (p 40)
There were a couple tables that caught my attention that were interesting.
Here is a variety of nai des/masen ratios across several studies:
Here is another on particle ending variations with parts of speech
I didn’t read into all the hypothesis’ details (kinda just skipped around) but several revolved around Kawaguchi’s claim that nai desu is when denoting negation (results agreed), nai desu preferred for yes/no questions or repeated (results agreed with some exception), nai desu preferred for interrogative sentences (agreed, but not significant). There was another one about nai desu being used more in declarative versus interrogative but sort of a mixed result.
Of course not saying one is better that the other (have no idea), but there are clearly differences of opinion. With that, not sure what to take a away as a language learner other than it’s nice to have some forgiveness in grammar choices or least some personal choice in expression.