A special case of machine learning in which a learning algorithm interacts with people (called an 'oracle') to make judgements about labeling data. Especially useful in situations where unlabeled data is abundant and manual labeling is expensive. [1] [2]