Handling Late Homework

Kristin Stephens-Martinez
7 min readAug 5, 2022


I’m going to take a break from my hybrid conference series and write a post about my latest attempt at handling late work. So far, this system is my favorite, and I plan to reuse it almost unchanged this coming semester. I created it to balance the needs of my students, my teaching staff, and myself.

The single-line explanation for my late system is I used tokens to empower my students to decide when they’ll turn in late work. But the devil is in the details, and I put a lot of thought into how and why I set things up the way I did. This post will outline the system itself, what the students thought of it, what I thought of it, and my plans on what to do next time.

[This article is also posted on my older personal blog.]

The Late Submission Token System

This policy applied to the homeworks that were part of the course’s ten modules. They were released as soon as we had them ready and a minimum of one week before they were due. Each homework had a one-week late window. We aimed to grade and return the homework within eight days of the due date (basically one day after the late window). This goal was to ensure we didn’t return graded work before everyone had submitted it and to accommodate the exam schedule, so students got feedback on a homework before a test on that content.

Each student got 9 late tokens. Each day a homework was late cost 1 token with no grade penalty. I don’t apply late penalties to anything after reading Grading for Equity. I chose 9 because it was the biggest number before 10. I was trying to tap into the left-digit bias (using digits instead of spelling them out to make this point). I wanted students to feel like they had enough tokens so they didn’t hoard them. While at the same time, I wanted to limit them so students did not use them with such abandon that we couldn’t keep the promise that we’d return homework right after the late window.

If a student needed more tokens, the syllabus said they needed to reach out to me. This request could be in an email or an office hour visit. I always granted the request and gave them another 9 tokens. The reasons ranged from “I’ll be honest, it’s senioritis” to “I’m having a crisis.” The former was refreshing and at least showed the student owned up to their situation. The latter was of greater concern. I built in this requirement to mainly catch the latter students. It served as a touch-point conversation to check in with the student to see how things were going. To me, needing more than 9 tokens could signal that the student needs extra help. Usually, the student didn’t need anything except those 9 extra tokens. But sometimes, the student had something going on that was majorly affecting not just my class but all of their classes. And I was the first “adult” to hear of this and could connect them to the needed resources.

I purposely added a “hurdle” to using a token. I required students to fill out a form telling us which homework and how many tokens. They had to fill out this form within 24 hours of the due date. The form did not ask why. That was a deliberate choice to message that they can use their tokens as they please. It was their responsibility to get their homework in on time, and it was none of my business why they were late unless it affected their overall learning. If a student used up more/fewer tokens than they said on the form, we deducted based on how many days they were late. I framed the form to the students as you fill out this form to tell us, “heads up, I’m going to be late, but don’t worry about me.” It was a way for us to know who we needed to check in on and be worried about since there were more than 200 of them. Another reason I required the form was to make students consciously choose to use a token rather than “slide” into it by just submitting late, which “looks” the same as submitting on time. I was tapping into the endowment effect and forcing students to explicitly acknowledge they were spending something that was theirs in hopes it would make them think about how they could not spend their tokens next time (and therefore not be late).

To enforce the form and to have another check-in point with students, I had two UTAs conduct an audit 24 hours after the due date. They checked who had and had not submitted to the autograder and form and when they submitted. There were three cases where we would send the student an email:

  1. If they finished the homework and submitted it 30 minutes into the late period, they got an email saying that we would not deduct a token regardless of whether they submitted the form.
  2. If they finished the homework, submitted it more than 30 minutes late, but did not submit the form, they got an email reminding them to fill out the form next time.
  3. If they did not finish the homework and had not filled out the form, they got an email reminding them about the policy, about help resources, and to fill out the form.

We did another audit after the late due date to see who did not submit the homework at all. These students got an email asking how things were going. The email encouraged them to at least try the homework for the sake of learning, even if they could no longer submit it, and to reach out if they needed more support. In reality, all of these emails emphasized the importance of homework for their learning and made it clear we were reaching out because we wanted to ensure everything was okay.

To help students know how many tokens they had used up, I used a column in our learning management system’s gradebook with 0% weight. We updated it once a week after each late period was over.

How Students Used the Tokens

Here’s a quick rundown:

  • 19% — Used no tokens
  • 65% — 1–8 tokens — 4% to 9% for each possible count
  • 11% — 9 tokens
  • 11% — Used 10–18 tokens
  • 5% — Used more than 18 tokens

So about a fifth of the students never used a token. A majority only used the ones they were initially given. Of those that needed more tokens, they only needed one extra round, and half the time, the student only really needed one more token. The 5% that needed more than 18 tokens feels about normal to me. Those students are that 5% in the very long tail distribution where something unique is happening to them, and the only thing not unique about it is that it’s affecting their learning in the course and usually also across all of their courses.

How Students Felt about the Tokens

I ran a mid-semester survey and included a question about the token system. The students loved it. It was a Likert scale question of strongly agree to strongly disagree about the statement “I like the slip day policy.” 87% strongly agreed, none disagreed, and only 4% were neutral. Some students even mention how much they appreciated the policy or the auditing emails directly to me, in other emails, and in evaluations.

How I Felt about the System and What I’ll do Differently

I love this system. I feel like I’ve finally found a system that works with my current philosophy and has a good framing for the students. In the future, I hope to use these late tokens as part of an early warning system. For example, besides the emails I mentioned earlier, I can imagine reaching out to a student if they plan to (or do) use an unusual number of tokens, like 3. Just to check in, since most students only use one token for a homework.

In addition, I want to be more systematic in tallying how often a student gets each kind of email with some kind of triggers that might warrant a full audit to check whether this is a consistent pattern or just an anomaly. I don’t yet have a good sense of what signals matter, and I suspect that my class size is too small to have super consistent patterns. 5% of my students are about ten students, after all. So this kind of system will likely be a lot of me checking and then reacting to what I see rather than more automating it away.


So as I said at the beginning of this post, I love this late token system. I give it 5/5 stars, would do again. And, in fact, I am. One of the UTAs in charge of the audits is continuing on them with me next semester. I plan to have her work on writing code to help make the system more automatic and less human error-prone. Of course, there will be a teaching staff in the loop deciding what email goes to each student, but we’ll probably do a mail merge after that. I’m of two minds on who should send the email. I had my UTAs do it last semester mainly so I wouldn’t be a bottleneck. But this coming semester, if it’s a mail merge, maybe I can handle it? And if I send it and CC the UTA in charge of it, it might send a clear message that I care about my students learning. And the UTA being CCed is an excellent backup if a student asks a simple question they can answer.

As for how many tokens, I think I will keep the 9, but I’m sure it will depend on your context, dear reader. It’s a balancing act that I suspect I’ll never get right for every student every time. It has to be enough so students don’t hoard and use tokens rather than be stressed. While not enough so that there aren’t surges in late submissions and students feel a “sting” when they use one to motivate them to do better next time. (sigh) A philosophical question I will likely never have a perfect answer.



Kristin Stephens-Martinez

Assistant Professor of the Practice in Computer Science at Duke University