Are judgments a form of data clustering? Reexamining contrast effects with the K-Means Algorithm