The cognitive and performance benefits of group work in undergraduate courses depend on understanding how to structure groups to promote communication and comfort while also promoting diversity and reducing conflict. The current study utilized social network analysis combined with self-reported survey data from 555 students in 155 groups to understand how students identified group members whom they wished to work with. Students' willingness to work with their peers was positively associated with behavioral traits pertaining to attention, participation, and preparedness in class. We tested whether preventing students from choosing their group members until completing a multiweek period of random assignment to different groups each week influenced group selection criteria, and we found little effect. Students continued to depend on demographic similarities such as gender and ethnicity when selecting groupmates and enforcing random interactions before the group formation did not influence group satisfaction and/or grades. Random interactions before group formation did influence the willingness of students to continue working with peers who were persistently poorly rated based on behavioral attributes and contribution to the group work. Thus, the effort of random assignment could be beneficial to identify struggling students and improve collaboration.