Playing with the data: what legal scholars should learn about machine learning