The encyclopedia · Software & IT · Operational decision · 2024–2025
Cornell replaced its lecture-time exam table with MIP and cut schedule conflicts
Cornell's registrar swapped a fixed lecture-time-to-exam table for a mixed-integer program, ending the 30-meeting-times vs 22-exam-slots conflict mess.
Cornell University
The solution
Final exams at Cornell were scheduled from a static table that tied exam groups to first lecture meeting times. With around 30 meeting times and only 22 exam slots, students routinely ended up with colliding finals, and faculty had to write make-up exams for students with simultaneous commitments.
Faculty and students in ORIE partnered with the University Registrar to build an integer programming framework. It pairs a multi-stage Group-then-Sequence model with an iterative Layer-Cake heuristic, generating many schedule variants under different parameter choices.
The model handles not just direct conflict and back-to-back exams but higher-order conflicts — students facing more than two exams in 24 hours or over four consecutive slots. The registrar chooses among Pareto-style options, balancing schedule comfort against front-loading and slot exclusions.
Why it worked
- The MIP makes the conflict structure explicit — including the higher-order cases spreadsheets and lookup tables simply don't see.
- Every semester thousands of students with hundreds of exams interact; only an optimizer can balance student comfort at that scale, and doing it by hand was never feasible.
- Keeping the registrar in the loop via schedule variants converts optimization from a black box into a decision-support tool the institution trusts.
What can be applied
When a static lookup table no longer fits the data, replace the table with an optimizer — but keep the human choosing among schedules via trade-off comparisons.
Aftermath
Cornell has run the framework for its final exam scheduling for five semesters since 2024, with the paper documenting advantages over the historical lecture-time approach in administration time and schedule comfort. An open-source MIP generator accompanies the work.
Sources
- Cornell University Uses Integer Programming to Optimize Final Exam Scheduling
- Cornell University Uses Integer Programming to Optimize Final Exam Scheduling (HTML full text)
- Cornell Final Exam Scheduling (publication page)
spotted an error? The archive wants to know.