Splitting the Deadline to Split the Queue

The CS 124 case at the University of Illinois shows what happens when a deadline problem belongs to a whole course, not only to individual students. The course had long programming-project checkpoints, and office-hour requests accumulated around the same due dates. Its split-deadline policy staggered groups' release and submission dates by one day while aiming to preserve equal working time. The case makes it possible to judge a deadline intervention on three separate dimensions: whether it spreads support demand, whether it preserves academic performance, and whether students regard the arrangement as fair.

The bottleneck was a shared support queue

CS 124 is a large introductory computer science course that serves both computing majors and a much larger population of non-majors. Its asynchronous format uses daily lessons with programming exercises, weekly proctored quizzes, and a longer multi-part Android programming project whose checkpoints are spaced one to two weeks apart. The course offers extensive online office-hour support. Students request help through a queue, wait for a staff member to enter an online room, and may need more than one visit before a problem is resolved.

That support model works well when requests arrive at a manageable rate. It becomes fragile when many students delay work until the same project checkpoint. The paper describes recurring spikes in office-hour demand near deadlines, creating long waits, overcrowding, and a stressful environment for students and staff. The course's smaller daily exercises did not create the same problematic spikes, partly because students had more flexibility in when to complete them. The intervention therefore targeted the longer project checkpoints rather than applying split deadlines indiscriminately to every assignment.

This choice matters. The policy was not introduced as a general lesson that students should be forced to start earlier. It was a response to a capacity problem produced by the interaction of a long task, deadline-oriented work, and limited support at particular times. If the bottleneck had been only participation in a short exercise, splitting the due date would have addressed the wrong problem. The course first had to identify where demand was becoming concentrated.

The paper defines office-hour demand as the number of support requests in a time interval, including requests that were resolved, requests that remained open, and requests for which a student was never visited. That definition captures pressure on the queue rather than only successful service. A student can make more than one request in the same interval, and a count of distinct students would miss that repeated demand. The course was therefore measuring how much work the support system was being asked to absorb at a particular moment.

This is a different objective from reducing the total number of questions. If a project is genuinely difficult, students may still need help after the deadline is split. The policy can be successful if those questions arrive in a pattern that staff can answer well.