Where do we get sufficient self-confidence to pull probabilities for ill-defined and under-measured quantities out of our butts so we can use them in The Formula?

Is there any actually interesting intellectual task that rests on nice justifiable grounded probabilities?