Lucas Holter

Recent positive feedback at Harvard University regarding the CS50 “regret clause” — which allows students to retroactively self-report academic dishonesty to skirt large-scale consequences — prompted similar debates on Yale’s campus surrounding the policy’s merits.

In December, three Harvard affiliates published the seven-page report, which spoke to the effectiveness of the honor code employed in their department’s flagship course. Among many points, the report found that CS50’s regret clause is the course’s “most meaningful, ongoing intervention” against plagiarism. Now, the authors of the report recommend other Crimson courses adopt the policy.

Like Harvard’s CS50, Yale’s version of the course — CPSC100, commonly also referred to as CS50 — also employs the regret clause. But unlike its Cambridge counterparts, the University’s CS50 team has not yet publicly assessed the effectiveness of the policy or its broader applications. Still, computer science lecturer Benedict Brown, who teaches CPSC 100, spoke highly of the policy’s effects.

“There’s no way to go through an ExComm hearing and call it a positive experience,” he said, because its high stakes make the process inherently stressful. “The regret clause is a much more positive way of doing things.”

If reported to Yale’s Executive Committee, those found guilty of academic dishonesty could face suspension, probation or other reprimands. But thanks to the regret clause, CS50 instructors on both campuses pledge not to bring such cases to the Executive Committee for students who admit to potential academic dishonesty within 72 hours of the submission deadline. Instead of traditional disciplinary measures, the report states, the team may give a zero for the problem set and connect students with mental or academic support structures across campus. According to Harvard co-author of the report and course instructor David Malan, academic dishonesty can come when students copy code from colleagues or the internet and pass it off as their own in problem sets — an act of plagiarism.

At Harvard, 89 students have invoked the clause since 2014; Brown said he could “count the number on one hand” of such students at Yale since he joined Malan in 2017.

For Brown, he says he has always instituted the regret clause in his courses. After all, he said, admitting to potential dishonesty “should not result in a worse outcome.” Brown added that stress — and rarely pure maliciousness — is a primary motivator for plagiarism among his students. Some students are also unclear on the course policies surrounding collaboration, so he said that invoking the clause could lead to a discussion that demystifies what is and is not appropriate.

“It’s an opportunity to have a constructive conversation,” he explained. “I’d like to see more students use the regret clause.”

But not all faculty members agree with the policy. Computer Science professor David Gelernter ’76 wrote in an email to the News that a regret clause is “contemptible” and a “tragedy.” To Gerlenter, students need to be more accountable than the regret clause permits.

In response, Malan told the News that the clause is not without consequence, but is rather without escalation.

“Insofar as the course’s ultimate mission is to teach, not only computer science but ethical application thereof, we’re thrilled that so many students have taken ownership of situations that might have otherwise been entirely punitive instead of teachable moments,” he wrote.

Even still, at the conclusion of his report, Malan and the two Harvard co-authors stood by the regret clause — and even recommended that other courses adopt it. But when it comes to other fields, Malan wrote in an email to the News that he would leave that decision up to faculty.

Malan’s report, first covered in the Harvard Crimson, was written — and meant — for Harvard University. But CS50’s Yale counterpart, CPSC100, is, in many cases, the same course. Both courses employ sophisticated software to detect plagiarism in problem sets, that not only searches the web for potential similarities but also pours through their dense archive of past submissions — including those from Harvard, Yale and their online edition of the course — to find matches. Students looking to fool the checker by merely changing variable names would be mistaken, as the program is “fairly robust,” Brown said. He added the amount of effort needed to make copied text look like their own is fairly similar to the effort it takes to learn the material.

But only a fraction of flagged problem sets show evidence of plagiarism. Before action is taken, CS50 instructors review the cases by hand and check for false-positives. Those that are left — if the regret clause is not invoked — are referred to the respective university’s honor committee. That partially explains why the number of cases brought to Harvard’s honor council has increased since the clause was added. According to the report, this uptick stems from an increased comfort “referring cases after students have had an opportunity to take ownership themselves but have chosen not to do so.”

Yale College Dean Marvin Chun said that he is not aware of conversations at the University regarding the inclusion of a regret clause in the Undergraduate Regulations. He wrote that he would ask the new Dean of Student Conduct and Community Standards Rachel Russell to see if she wants to discuss the topic.

Director of Writing for the Poorvu Center for Teaching and Learning Alfred Guy, Jr., who is also the Assistant Dean of Academic Affairs for Yale College, wrote in a statement that there has not yet been sufficient research on the value of a regret clause. But Guy added he “would expect” the policy to have some of the same limitations as integrity pledges — declarations a student would make on an assignment promising that they would not cheat. He added that this retroactive reporting “is less effective than measures taken before or during the time of completing the work.”

According to Guy, regret clauses are not yet common enough to support scientific research on the subject. But for Malan, the policy is here to stay.

CS50 first came to Yale in Fall 2015.

Matt Kristoffersen | matthew.kristoffersen@yale.edu

  • Nancy Morris

    “ At Harvard, 89 students have invoked the clause since 2014; Brown said he could ‘count the number on one hand’ of such students at Yale since he joined Malan in 2017.”

    Last time I looked a hand had five fingers. That’s a pretty big disparity between CS50 and CPSC100 if one assumes a steady annual rate of cheating since 2014.

    What, if anything, does that disparity suggest about the relative rates of cheating at Harvard and Yale?

    Both CS50 and CPSC100 employ sophisticated software to detect plagiarism in problem sets. It would be interesting to know what the rate of “flagging” is for each of the two schools. Does Harvard software flag at an elevated rate compared to Yale comparable to the disparity in the “regret clause” rate, for example?

    Sophisticated software could be used to detect plagiarism in all essays, not just computer science coding. There is software that not only searches the web and libraries for potential similarities, but also could pour through the students archive of past submissions — including application essays.

    Some software flags radical changes of style within a single essay, suggesting that another author has inserted a paragraph or extended passage into the essay. Similar comparisons can be made between a given essay and a body of writing purporting to be by the same author. Such “internal” software detects plagiarism by scanning for dissimilarities in compared passages, in contrast to the “external” software that scans for similarities between passages in the essay with passages from external sources.

    Yale should employ such flagging software of both “internal” and “external” types generally, not just in computer science courses. It is a fact that some students – especially those from wealthy families – sometimes hire third parties to write their essays. Indeed, some such students deliberately avoid taking courses with exams, but instead require essays, specifically to have paid third parties write those essays. Thorough use of plagiarism detection software should significantly suppress the incentives for such pernicious behavior.

    And there is obviously the question of why “regret clauses” should be limited to computer science classes, regardless of how extensive the use of plagiarism detection software. Why should a CPSC100 Student have this “regret” option but not someone who incorporated, say, a block of Wikipedia text into a history essay?