We use a variety of text classification algorithms on a large corpus of college admissions essays to model the extent and ways that AI might inform human evaluation.