I don't remember the number of questions but I played it for quite a while. I remember being frustrated because I couldn't get the 50% option calibrated, since I didn't want to purposely choose the option that felt less likely. Presumably more practice would have helped distinguish between weaker and stronger impressions of that kind.

I agree that the brain does this kind of thing computationally, and it would be a mistake to suppose that there is some other non-computational way to do it.