Under the guidance of Remy Spliet, Associate Professor at Erasmus School of Economics, a group of students developed a set of algorithms to automatically create exam schedules for all faculties. Around 2,400 exams are held at campus Woudestein every academic year. They are scheduled manually, with the initial work being done by each faculty individually. A case study from the Master Econometrics and Management Science shows that it could be done differently.
How did you come up with the idea of developing a case study for exam schedules?
'We noticed that students increasingly feel the need to work on social issues rather than a case study where the aim is to maximise a company’s profits. Until recently, we mainly worked with large companies such as PostNL, KLM or NS. Case studies for those companies would involve scheduling trains, for example, or cabin crew for a flight. This year, for the first time, we did a case study for Erasmus MC focusing on scheduling population screening programmes in third-world countries. That case study was very popular among the students. Creating exam schedules is also something that resonates with them. What’s more, this is a planning issue that has been around for a long time at EUR, and it’s very time consuming. Fortunately, Stella Verbrugge from Education and Student Affairs (E&S) and the faculties’ schedulers were immediately enthusiastic and eager to participate.'
How are exam schedules currently created?
'The schedulers create an ideal schedule for each faculty based on their knowledge and experience. They use computer programs where you can display the schedule and look for conflicts, but there is no algorithm that creates the schedule. E&S then coordinates the schedules from all the faculties to create a single combined schedule. They hold meetings where conflicts are identified and resolved. That takes a lot of time.'
What did the students do?
'Their task was to create a single, integrated schedule for all faculties for the 2023–2024 academic year. The first obstacle was that EUR does not have a single data system showing which exams need to be administered. In the end, we were able to derive a dataset from a university website. It had obvious shortcomings, but it was good enough for the case study. Next, four groups of four students went and talked to all the schedulers to find out each faculty’s preferences. For instance, most faculties would rather not hold exams on Saturdays or in the evenings, but some actually prefer those times because that’s when their part-time students are not at work. Some programmes prefer not to have exams for second and third-year Bachelor courses on the same day, because some students will have had to redo second-year courses. Based on this set of constraints, the students set to work. I thought it was really cool to see that, within two weeks, all of the groups had some code ready to go with which they could instantly create a schedule that met all the strict requirements and made a certain trade-off between different preferences. Each group then started working on specific ‘what-if’ questions, like ‘what would happen if there were 50% more students’, or ‘what if the default was for exams to be two hours long instead of three?’'
What results did the case study produce?
'The students were able to provide quantitative insights for the first time, showing for example that current room capacity is sufficient. Last year, the Sports Building and the Van der Goot Building were used to provide emergency capacity during the busiest weeks. The case study showed that this would not be necessary if all lecture halls on campus were used for exams. But it seems to me that creating such a schedule manually would be almost impossible, and besides, not all lecture halls have a layout that would work for exams. The students also discovered that it wouldn’t help much if the exams were shortened to two hours, so that there could be four exams a day instead of three. The same turned out to be true for student numbers. The real bottleneck was the diverse range of preferences. If you want to hold the exams for second and third-year Bachelor courses on different days, but not in the evenings or on Saturdays, you will eventually reach a limit.'
Was it an eyeopener for schedulers and E&S?
'Yes, particularly one specific moment during the presentation. One group had a slide in their PowerPoint showing a schedule for the exam week before Christmas, the busiest week of the year. It showed that the extra capacity would not have been necessary if all the lecture halls on campus were set up for exams. When a scheduler noticed that exams had been scheduled on the Saturday before Christmas when they were not supposed to be, the students flipped open their laptops, locked out the Saturday, and were able to show the new schedule at the click of a button. Mouths fell open. I thought that was a really cool moment. The students enjoyed it too, because they heard for the first time that what they do really makes an impact. I was really proud.'
So it was a great success... will it be introduced?
'That’s trickier than you think. The algorithm has to be converted into scheduling software, and for that you need a good, uniform data system. Next, preferences need to be better identified and then prioritised. After that, someone can build the software, either internally or externally. So there’s a lot involved. But it’s great that everyone is enthusiastic and can see the possibilities. As a decision-support system, it would be an enormous help to schedulers, allowing them to spend more time working out how best to serve student interests with scheduling. After all, let’s face it, you don’t make yourself popular as a programme if your exam schedule isn’t up to scratch. You want to avoid that.'
- Associate professor
- More information
This article is derived from an interview of the Erasmus University Rotterdam, that can be found here.
For more information, please contact Ronald de Groot, Media and Public Relations Officer at Erasmus School of Economics, rdegroot@ese.eur.nl, or +31 6 53 641 846.
