Learning from data without answers: the algorithm finds structure, such as groups, and people decide what the groups mean.

ExampleClustering photos of two cats may group them by sunny versus cloudy, not by cat.