the operators only know that they’ve tuned up their algorithm to get the answers right; they don’t necessarily know the precise details of how it gets there.
On what basis then, one can improve the precision (specificity, positive predictive value) of an algorithm?
