Recall is the fraction of actual positives correctly identified by the model (true positives / all actual positives). High recall means fewer false negatives, which can be critical in contexts like disease detection. It is often analyzed alongside precision for classification tasks.