Fair Senior Capstone Project Teaming Based on Skills, Preferences, and Friend Groups

Capstone design courses are critical pedagogical components of any engineering curriculum as they allow students to complete open ended projects in a team setting, often while interacting with industry customers. Equitably teaming students for these courses can be a difficult challenge. Each team must have the technical and leadership skills necessary to complete the project, and industry sponsors prefer to have high performing students for recruiting purposes. Moreover, students often have strong preferences based on which technical challenges excite them, and often desire to work with the friends that they have developed close relationships with in other courses. Large universities often have hundreds of students enrolled in these courses, making hand-design of teams intractable. In past years, the Aerospace Engineering Department at our university has allowed students to choose their own teams during the second meeting of the class. While this allows a great deal of freedom to students, many end up assigned to projects that they find undesirable. Further, this methodology inadvertently resulted in the formation of student cliques and peer to peer conflict, which then may have damaged team dynamics and performance throughout the year. In fall 2020, the COVID-19 pandemic precluded the usual teaming method, and the project advisory board adopted an optimization-based fair rank-maximal allocation approach. There have been many other similar optimization approaches proposed in the literature with objectives ranging from balancing team grade-point averages to distributing Meyers-Briggs personality types, and solution approaches ranging from evolutionary algorithms to mixed integer programming. The optimization objective for our approach has 3 components. First, each project has minimum requirements for the number of students with electronics, software, hardware, and project management skills. Second, to allow students to work with friends and avoid isolating minoritized students in uncomfortable situations, they were allowed to designate one or two other students with whom they would be assigned. Finally we attempted to find a fair rank-maximal allocation based on the students' ranked preferences for projects. This optimization problem was formulated as a mixed integer linear program, expressed using the Julia programming language's JuMP optimization framework and solved with the Gurobi optimizer. The resulting team assignments were very successful, as each student was assigned to one of their top three project choices. In addition to discussing the mathematical details and limitations of the approach, we plan to analyze two research questions: 1) Did teams this year perform better than teams in the past in terms of grades in the preliminary and comprehensive design reviews? 2) Was there a notable difference in the team dynamics this year compared to past years? To examine team dynamics, peer evaluations from this year (including comments) will be compared to those from the past two academic years. Limitations of these methods will be acknowledged, particularly as COVID-19 could severely impact how students are interacting in teams. Finally, the code will be made available online as an MIT-licensed open source package.