I'm pushing back a little on this "classifier that avoids false positives" description because that's what every classifier is in some sense intended to do
Well presumably there's a trade-off between avoiding false positives and avoiding false negatives. And you want a classifier that tries really hard to avoid false positives, as I understand.