Is it better just to choose random students for groups? Or first evaluate each student and then assign students in groups to help with diversity?