The balancing concern is that, in the case of "too small to be worthwhile", we want people to not consider it, and there are strong biases that make people consider it a la but there's still a chance! If anything less than "absolutely zero evidence" lets people consider evidence - and overestimating the evidence is worse than underestimating it - then it is preferred language.

Controlling behavior by removing the ability to express concepts. Lets call this new language 'newspeak'. :)