Discussion: View Thread

[Learning Action] - Use of Analytics to Provide Optimal Social Distancing in Classroom Assignments

  • 1.  [Learning Action] - Use of Analytics to Provide Optimal Social Distancing in Classroom Assignments

    Posted 08-04-2020 06:22:00 PM
    Greetings-


    The campus wide leadership at Oklahoma State University has been proactive about sending a clear message (with a detailed plan) from the start of the summer that face to face classes would be held on campus as much as feasible in the Fall 2020 semester. 

    As policies were developed over the summer, a variety of efforts were undertaken to accommodate safe face to face classes, while allowing students and faculty to opt for a 'virtual' experience should they desire (or need it for health reasons).

    Social distancing guidelines were used to determine a new social distance capacity (SDC) for each classroom.  Obviously, the SDC tended to be significantly smaller that the usual class capacity.  Unfortunately, enrollment in Fall classes had commenced in March with the old class size limits – creating a dilemma on how to best operationalize face to face classes given the new social distancing capacities.

    A rotational strategy was thus proposed from OSU's leadership whereby up to three (3) rotations of a class would be allowed so that everyone could experience the class 'face-to-face' at least some of the time. Thus, students would rotate between attending class and receiving course content virtually (either synchronously or asynchronously).  In essence, this allows a class enrolment to be up to 3 times as large as the SDC and qualify as 'face-to-face'.

    The process of adapting the Fall schedule to this new social distance/rotational schedule was decentralized – left up to the individual colleges.  The Spears School of Business leadership was quick to act and developed a strategy based upon data analytics and multi-objective optimization that has allowed us to have a great transition plan to this new normal for Fall 2020. 

    Spears Business students were surveyed in late June and asked to indicate whether they preferred a face-to-face experience or were going to opt to take the class virtually (they have the choice, and in fact, can change their mind at any time during the Fall semester). An 80% response rate resulted in our ability to update class enrollments with a new accurate measure of expected face to face enrollments (FFE) for each class.

    The optimization process then went as follows: Classes were kept in their original time slot, so the model was run for each class time slot (MWF 9:30-10:15,  TTh  9:00-10:15, etc.).   

    The core model was a basic assignment model, optimally assigning classes (with attribute of FFE, expected face to face enrollment) to classrooms (with attribute of SDC, capacity). 

    If  FFE (enrollment) was greater than 3*SDC (capacity), that assignment was not allowed (more than 3 rotations would be needed).

    There were multiple sequential objectives that drove the solution process.  The first objective was to maximize the feasible assignments.  At this stage, there were a few decisions that had to made to remove a large class from future consideration because of its size, and it was moved 'on-line'.

    The second objective was to find assignments that minimized classroom changes.  It was felt by the modeling team that the goal of minimizing disruption (given all the other disruption going on!) was paramount to having faculty view the process more favorably.

    The third objective was to then minimize the number of rotations needed to implement the schedule while keeping the first and second objectives achieved at the level previously reached. 

    The modeling team did vary the second and third objectives in 'what-if' analysis to consider different similar alternatives in implementing the changes.

    The model was implemented using easily obtained enrollment and SDC data from the student information system, as well as using OpenSolver within EXCEL to implement the multi-objective assignment model.  Being able to use a spreadsheet as the underlying tool made the model very easy to use for scenario analysis and easy to interpret for implementation.

    The results of the multiple models led to classroom swaps that were implemented in mid-July, giving faculty (and the Registrar's office!) plenty of time to adapt to the new schedule.

    Now, two weeks prior to the start of the Fall semester,  new data is being collected (new enrollments, students changing their minds, etc.) regarding the FFE information.  The Dean's Office team will be reviewing all the assignments one last time prior to the start of the semester to double check any possible anomalies that may have occurred due to changing enrollments over the last month.  

    This is a great example of Analytics in Action during a very difficult 'crisis' situation.  The teamwork of the Dean's office and the use of a reasonably sophisticated yet easy-to-implement prescriptive model will hopefully help maintain some sense of 'normal' for the Fall semester classes.

    I am glad to provide additional details about the model that we developed – please do not hesitate to ask me off-line.

    Thanks

    Rick 



    ------------------------------
    Rick L. Wilson
    Professor and Head-Department of Management Science and Information Systems
    W. Paul Miller Professor of Business Administration
    Dan and Tiffany Howard MBA Faculty Fellow
    Spears School of Business
    Oklahoma State University
    Stillwater OK 74078
    405 744 5084
    ------------------------------