Kristin Stephens-Martinez is an Associate Professor of the Practice in Computer Science at Duke University. She studies how students learn and engage in computing courses and uses classroom data to improve teaching, assessment, and student support. Her interest in grading policies comes from balancing two goals: creating systems that help students learn and building processes that instructors can realistically sustain in large courses. She writes about teaching and computing education on her blog and hosts The CS-Ed Podcast, where she explores evidence-based approaches to teaching and learning with educators from across the computing education community.
When I first read Grading for Equity, it inspired me to eliminate penalizing students’ grades for late submissions while still allowing them to submit work late. I preferred for students to submit late and learn from the work than use any means necessary to get it in without learning. However, I also wanted to push them to submit as close to the due date as possible. This way, the grading workload would never get too out of hand, and students would mostly stay on track.
These goals led to my late-token assignment system that does more than manage deadlines. It helps students make intentional decisions, gives teaching staff structured opportunities to check in, and keeps the administrative workload manageable. And I love it. In this post, I will outline my details and design choices and explain why I made them, so you can adapt the ideas to your own context.
This is based on a post I wrote on my blog in 2022, but I’ve rewritten it for a broader audience and updated it with what I do now.
Context
First, I strongly believe context matters. The main course in which I implement my late-token system is a data science elective focused on programming. The course prerequisites expect students to know how to program, and if their skills are not as strong as those of Computer Science majors, they are expected to have some statistics experience, so those parts of the course are easier as they practice more with programming.
My course attracts students from across the disciplines, with only half of them majoring or intending to major in computer science. The other half includes majors like economics, psychology, public policy, engineering, chemistry, and literature. The class enrolls 70 to 120 students and meets on Wednesday and Friday for 75 minutes. My teaching staff includes two graduate teaching assistants (TAs) and about 1 undergraduate TA per 8.5 students, with undergrad TAs working about 5 hours a week.
The course has 10 one-week modules. A module includes prep work due on Mondays, in-class work, and a homework assignment. The prep work and assignments are released as soon as I have them ready, and at least one week before they are due.
Homework assignments are due on Sundays at 11:59 pm. They are a mix of coding and written responses. Some portions can be autograded, and students may submit those as many times as they want. Typically, students submit until they receive full credit on the autograded part. The rest is hand graded, and there is no resubmission opportunity for this part. However, students only need 90% of the possible points to receive full credit on a homework. The split between autograding and hand grading varies widely from assignment to assignment.
The Late-Token System
Students can use “late tokens” on the homework assignments. The late-token system starts with a simple rule: students may submit assignments late with no grade penalty, at the cost of one token per day. After an assignment’s due date, students have a 7-day window to submit it late, except for assignments right before an exam that covers material from that assignment. Those assignments have a 4-day late window (see the next section if you are curious why). Submitting beyond the late window requires an extension from me. Students start with 9 late tokens, and the class has opportunities to earn 2 more tokens. They often earn both, resulting in a total of 11 available tokens by the middle of the semester. To help students keep track of their remaining tokens, after each assignment’s late period closes, my grad TAs or I update a column in our learning management system’s gradebook with how many they have used.
To use tokens, students must fill out a 2-question form indicating which assignment and how many tokens1. They have to fill out this form within 24 hours of the due date. The number of tokens a student uses is based on their submission date, not on what they say in the form. I frame 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, I have a plan to be done in X days.’” As a result, we also then know who we need to check in on, since the class is so big.
If a student needs more tokens, the syllabus states that they should reach out to me. This request can be made during an office hours visit or via email to me or to the class email address. I’ve always granted the request and given 10 more tokens after at least a brief check-in exchange.
To enforce the use of the form and provide another check-in point with students, I have a TA run a script 24 hours after the due date that audits who submitted the assignment and when, versus who filled out the form. My system is more elaborate than strictly necessary, such as including a 30-minute grace period to account for potential connectivity issues. See my later section on ways to get started. Here are the possible scenarios. All the emails sent out also remind students about the late-token policy and to fill out the form.
Done on time – No action taken. Done is defined as getting >90% credit on the autograder tests.
Late form submitted – No action taken, unless they finished during the grace period (see next bullet).
Done during the grace period – Regardless of whether they filled out the form, they receive an email notifying them that we will not deduct late tokens and reminding them about the late policy.
Done after grace period and no late form – Email sent noting they are done, but the form was not submitted.
Not done and no late form – Email noting the <90% on the autograder portion.
No submission and no late form – Email sent asking how things are going and noting that there is no submission.
Why I Made These Choices
Now that I’ve explained the system, here I’ll address some common questions about my design choices.
Why is 1 token equal to 1 day, rather than some other unit? – I’ve heard of having a token equal to an hour. I like 1 day because the odds are good that the student will be able to get to office hours within those 24 hours if they need help. Having an entire day also nudges the student to reassess whether rushing to finish the assignment (at the potential cost of their learning) is worth it now that they’ve crossed the threshold and have an entire day to work on the assignment. What’s important is that you pick a unit long enough for the student to have a self-regulatory reflective moment to assess the best way to approach finishing the assignment for the sake of their learning, not the deadline. And your choice needs to be a whole unit, rather than something like 0.5 or 3 days, so that it’s easy to remember.
Why a 7-day late window? - So students have another weekend to do the work if they need the time. My students are often weekend workers. This many days also helps them find time to work on the assignment again around the rest of their weekly schedule. On the other hand, using 6 or 7 tokens on a single assignment occurs only for 1 or 2 students per assignment.
Why a 4-day late window before exams? – Exams have two parts and take the entire week. Wednesday has a paper part, and Friday is a take-home part. The late window for an assignment closes on the Thursday before, so we can grade the assignment in time for students to use the feedback over part of the weekend to study and ask questions in office hours well before they take their Wednesday exam.
When do we grade? - We start grading two days into the late window with the goal of returning the assignment one or two days after the late window closes. This goal was to ensure we didn’t return graded work before everyone had submitted the assignment and to accommodate the exam schedule.
Why 9 late tokens? - I chose 9 because it is the biggest number before 10. I’m trying to tap into the anchoring bias toward the leftmost digit of a number (using digits instead of spelling them out to make this point). I want students to anchor on the 9, not the 1, so they remember it and feel they have enough tokens and won’t hoard them to the detriment of their learning. At the same time, I wanted to limit them so students didn’t use them with such abandon that we couldn’t keep our promise to return assignments right after the late window.
Why let students get more tokens? - To help the students focus on the learning, not penalties. Moreover, I require them to talk to me to help catch struggling students. This serves as a touchpoint to check in with the student to see how things are going. To me, needing more tokens than the original number can signal that the student needs extra help. Usually, the student doesn’t need anything except those extra tokens. But sometimes, the student has something going on that is majorly affecting not just my class but all of their classes. This means I may be the first “adult” to hear about this and can connect them with the necessary resources.
What do students say when they ask for more tokens? - The reasons range from “I’ll be honest, it’s senioritis” to “I’m having a crisis.” The former is refreshing and at least shows the student owning up to their situation. The latter helps me better support the student.
Why have a form? – It adds some friction and a reflection moment so students consciously choose to use a token rather than “slide” into being late by just submitting late, which looks the same as submitting on time. This taps into the endowment effect by forcing students to explicitly acknowledge that they were spending something that was theirs, in hopes of prompting them to consider how they could avoid spending their tokens next time (and therefore not be late).
Why not ask the students why they are using a token? – I want to lower the barrier to using the tokens and treat them like the adults they are practicing becoming, not require students to “perform their hardship.” I want to signal directly to them that turning the assignment in on time is their responsibility, and that it is none of my business why they are late unless the reason affects their overall learning. And I will know if that reason is affecting their overall learning when they run out or need more.
What if students never submit the assignment? – We audit after the late due date and send them an email asking how things are going. Often, the student responds by explaining what is happening. If they do not respond, I often reach out offering an extension. The script also generates a note about this student for a running spreadsheet of notes on all students, which I collaboratively update with my graduate TAs. These notes help me support students more holistically and provide evidence when I decide to call in extra support, such as the student’s advisor.
Student Reactions
In Spring 2026, I had 118 students. By mid-semester, everyone had had 11 tokens. This is how they used them:
25% - used no tokens
44% - used 1 to 10
14% - used exactly 11
15% - requested an additional 10 and used up to 20
3% - requested and used more than 20 tokens
Four years ago, when I first posted about this system, my Spring 2022 class had a similar distribution. The main difference is that Spring 2022 students had fewer students at the extremes. The 3% of students who used more than 20 were all students whom I was aware of and connected to more resources during the semester to make sure they had the support they needed. I see this as the system working, because in the past, I might not have noticed these students until well past the crisis point.
As for how many emails went out, the percentage of students who received an email ranged from 13.6% to 33.9% per assignment, with a mean of 21.8%. Most of the emails were due to the lack of submission and form. This does not surprise me because there is no penalty for not submitting the form, except for an email that they can easily ignore. In the future, I’m considering adding to my student notes spreadsheet whether a student received an email and why, which will give me more context if I ever reach out to the student or they come talk to me.
Regardless, the students love the flexibility of the late-token system. In my mid-semester survey, I ask students 5-point Likert scale questions about various aspects of the class, and the late-token system is always the most liked, and it’s been climbing. In academic year 24-25, over 80% of the students strongly agreed or agreed with the statement “I like the late-token system.” And in the following year, it was over 90%.
Should you use this?
I will close with how I recommend deciding whether this is worth implementing. If you have a large class and want a way for students to submit late work without deducting points, while still having a forcing function that pushes students to turn things in on time, then yes, this system could help. To get started, I recommend the following steps:
Decide on a simple set of rules to make it easy to know who should receive a message (start small, add things like grace periods later).
Create a form that auto-captures student email addresses (typos happen; avoid them if possible).
Figure out how you will message students (I’ve written a post on my blog if you’d like details on how to do this).
Figure out how students will know how many tokens they have left.
If you have a small class where you can handle one-off cases of late work, I suspect this system is not worth your time. Instead, you could add a paragraph to your syllabus stating that students can submit each assignment late for up to X days, with no explanation required. But you reserve the right to talk to the student about what is going on if it becomes a pattern. I still recommend having a form or providing them with a template email to send you, so they make a conscious choice rather than slide into it. But there’s no need to track how many tokens they use or audit and email them to confirm whether they actually filled out the form or sent the email. Just note how many days late the student is each time, and follow up by talking to them about it when you see a pattern. In my experience, students want to feel seen by their teachers and act more wisely when they know they are. So if you follow through when students overuse the policy, the word will get out that you are both generous and paying attention. The cost of this kind of late system will be the administrative work of tracking, and the one-off discussions when submitting late becomes a pattern.
I do not recommend including an open-text box question asking “Why?” See the next section for my reasoning.
