Arguably, assigning a particular floating point number between 0.0 and 1.0 to represent subjective degrees of belief is a specialized skill and it could take years of practice in order to become fluent in numerical-probability-speak.* Another possibility is that it merely adds a kind of pseudo-precision without any benefit over natural language.

In any case, it seems to be an empirical question and so should be answered with empirical data. I guess we won't really know until we have a good-sized number of people using things such as PredictionBook for extended periods of time. I'll keep you posted.

*There does exist rigorously defined verbal probabilities, but as far as I know they haven't been used much since the Late Middle Ages/Early Modern Period.