Recall is the share of real positives that a system manages to catch. If there are 100 actual violations and the system finds 80, recall is 80%. In medicine it's usually called sensitivity, and you may also see it called the true positive rate.

‍

‍Why it matters

Recall asks the opposite question from precision: of everything that should have been caught, how much was? It matters most when a miss is costly, like a policy violation staying visible or a real problem going unflagged. A system can reach near-perfect recall by flagging almost everything, which is why it's read alongside precision.

‍

‍How it's calculated and traded off

Recall is true positives divided by the sum of true positives and false negatives, so correct catches divided by everything that should have been caught. Most systems produce a confidence score, and the threshold for flagging is the main lever. Lowering it usually raises recall and lowers precision, and raising it usually does the opposite, so teams set it based on which mistake costs more.

‍

‍When recall matters most  

Medical screening leans on recall, since missing a real case is usually worse than sending a healthy person for a follow-up check. Content moderation puts the same weight on recall for the most severe harm categories, where leaving a violation up is the bigger risk.