In reading social media posts by philosophers and speaking with colleagues around the country, there seem to be four faculty responses to the academic dishonesty made possible by large language models (LLMs), like ChatGPT:
- I don’t want to be in a policing relationship with my students; the teacher-student relationship requires trust, and I trust my students.
- The students who are using artificial intelligence to complete their assignments don’t understand the purpose of education; that’s their loss, not my business.
- ChatGPT is the new calculator; we will adjust eventually to how artificial intelligence systems make our work easier.
- ChatGPT, if used to replace challenging intellectual processes, like essay composition, will rob students of valuable educational experiences; as educators, we should do everything we can to clarify and regulate the proper use of LLMs.
The first three responses do not require any action on the part of overworked professors, which is convenient. The last response, 4, demands that we take active steps to uphold whatever virtue remains in the U.S. system of higher education. Collegiate structures are supposed to reward hard work and talent, in contrast, perhaps, to corporate structures, which reward gameplay, appearance, identity, success, and connection. But students are struggling lately to discern the difference.
- Because the number of meritorious candidates for admission at selective colleges fully eclipses the number of open spots, getting into any particular selective college has become a matter of luck, at best, and academically irrelevant factors, like wealth, legacy status, and athletic ability, at worst.
- Once you get into college, students utilize shortcuts, whenever possible, to complete homework assignments. Many students will do anything they can to avoid difficult reading—reading with dense, antiquated, technical, or translated language. Instead of slogging through the text, they will hunt for Wifi videos, Quizlets, or Wikipedia and other internet encyclopedia entries, which have more familiar and accessible formats. About to make matters worse, Google has just introduced Illuminate, https://illuminate.withgoogle.com/, which will generate audio summaries of academic papers.
- Cheating is prevalent, even at schools that tout their culture-defining honor codes. A survey of 2024 graduates of Princeton University by The Daily Princetonian, for example, found that 28.8% of students report cheating “on an assignment or exam in violation of the Honor Code.” Almost half of students there (42%) report having knowledge of a peer violating the Honor Code and choosing not to report the transgression, which itself violates the Code. This is not to pick on Princeton, which, first, should be recognized for the honest and open spirit with which it continues to study the issue. Second, one can imagine that the situation is much worse at institutions without honor codes, given that studies estimate 60-80% of students cheat in high school.
- Beyond the fact that virtue will not get you into college or to the top of an assignment curve, students are wise to the many other arenas in which virtuous behavior goes unrewarded in academia.
All of this has students considering what the collegiate experience is really about. College begins with a tuition payment and ends with a credential. That much is clear; that message is sent. Thus, students are wondering, as long as the terms of this transaction are fulfilled, as long as people are stamped as qualified for graduate schools and employers, does it matter to anyone what happens in between? Is college a sham?
No, college is not a sham. All four of the above responses to ChatGPT seem minimally reasonable. But 4 holds special urgency because it foregrounds our faculty’s interest in teaching and learning.
This past year, I made two changes in my courses with the hope of preventing cheating, reinforcing the importance of virtue, and signaling that the learning of each student matters to me. In philosophy courses in which essays were assigned, I required ChatGPT usage reports. In my logic courses, I conducted in-person final exams broken into stages to improve accessibility while still supporting academic honesty.
When writing essays for my courses, students are required to use a Google doc, save that doc as a .pdf file, and upload the pdf to our learning management system. Students then have to complete the ChatGPT table shown in Appendix 1 below, which indicates the extent to which ChatGPT was used for the assignment. Given my secondary sympathy for response 3 above and my recognition of our new reality, I allow students to use ChatGPT as a tool in their writing. International students, especially, use ChatGPT to streamline and polish their English prose. But all students must document the extent of that usage.
The table in Appendix 1 below is a modified version of one originated by John Bartucz, a Teaching Specialist in the University of Minnesota’s Department of Curriculum and Instruction. At the bottom of the table, students enter the address of their Google Docs. If I suspect more ChatGPT usage than what was reported by the student, I can turn to the version history and last edit features within Google Docs to get a detailed look at the student’s writing and editing process. If I see a large cut-and-paste or very few revisions, I can call the student in to discuss the content of the paper as a further check.
Throughout the year, students seemed to honestly report the extent of their ChatGPT use, in that not all students exclusively circled the left-hand column of the table. I had one underreported use of ChatGPT, and I was able to use Google’s version history feature to confirm that the student did not craft the essay bit by bit. Several students mentioned the system to their other professors and in their student evaluations as an effective way to address LLMs.
In the future, I would like to see the advent of cheap, simple, internet-disabled, laptop-like, word-processing, electric typewriters that can either print or send student files to an instructor. Writing labs with this equipment could be established for philosophy courses. Students would schedule a time to enter the lab and write, print, or submit their work at the end of the time period. If this seems too cruel for paper writing, which benefits from revisitation, I believe it would be ideal, at least, for in-person essay exams.
Turning to logic courses, these more quantitative courses typically involve several in-person exams. Professors used to require students to remain in the testing room throughout an exam. But the current thinking is that people have bodies with different needs and should be able to serve those needs. Unfortunately, some students take advantage of this new openness by using their phones outside the testing room to look up answers or by leaving notes to themselves in restroom stalls.
In order to enable students to take whatever kind of break they need, physical or mental, I divide my logic exams into 30-minute chunks and allow students to leave the room at any time, as long as they turn in to me the portions of the exam that they already have completed. All students begin with the first 30-minute portion of the exam and a checklist enumerating each portion of the exam with a short description, like “test the validity of six categorical syllogisms” or “identify 15 fallacies,” so they know what is coming and can plan accordingly. A sample checklist appears in Appendix 3 below.
After completing the first 30-minute portion of an exam, students are free to come up to the front of the room and grab whatever 30-minute portion of the exam they wish to work on next. (Please see the photo below in Appendix 4.) If they want to leave the room to visit the restroom, get coffee, or just take a mental break, they can, as long as they finally submit to me their completed exam portions. In other words, students who decide to leave the testing room cannot get back the portions of the exam that they have submitted to leave the room. I hold the exam portions at the front of the room until the end of the allotted time. As students finish the exam, they come up to the front with their checklists, reunite their just-finished portions with their earlier submitted portions, review the checklist, to ensure that every portion is complete, and submit the whole stack.
The system involves a bit of chaos, as students wander up to the front at different times to pick up exam parts, stand by my desk assemble their final stack, and complete their checklist. It’s best to set the instructor’s desk as far away as you can from the area in which the students are working and ensure that students with accommodations for silence have those needs met.
There is still room to game this system. Students recently asked me if it would be okay to break after portion 1 to study more for portion 2. I found myself having to say “yes.” It would be possible for two students to meet on a break, with one conveying to the other the content of some earlier portion of the exam. But I believe the risk, along with the added testing room hubbub and instructor effort, is worth it, because honest students appreciate the system and the thought behind the system.
An additional way to promote academic honesty in logic is to give oral exams. I began doing this during the COVID-19 pandemic and have a video presentation on the pedagogy here. Students were required to take me through two proofs, one easy and one difficult, within a 30-minute one-on-one meeting.
I prepare my classes for these unique logic exam and paper submission systems by explaining the procedures and delivering a little speech on the first day of class:
Maintaining an academically honest climate is about turning off intrusive voices that plague the minds of honest strivers. In an environment where cheating on tests and copying papers is widespread, students who resist those temptations and put in the work often find themselves with lower grades than the students who have gained an unfair advantage. Honest students have less time to spend on other courses and cannot submit assignments that are as polished as the machine versions. Are such students fools? Strivers often second-guess themselves and their costly decisions with worries like, ‘Am I a dupe to spend six hours on this paper, when I could have generated something instantly with ChatGPT?’ ‘Am I a chump for not writing the probability formulas under my sleeve?’
I just can’t have good students punishing themselves for having the maturity to think beyond the tuition-diploma transaction—for going after the learning, in addition to the credential. The two small changes discussed here say to the genuinely engaged learners, “You’re not a fool. What you’re doing is virtuous, and I am going to support your efforts.”
Alexandra Bradner
Alexandra Bradner is an adjunct philosopher of explanation and understanding, care, and pedagogy who has taught more than 90 sections of 27 courses at institutions including Northwestern University, University of Michigan, Marshall University, Denison University, University of Kentucky, Bluegrass Community and Technical College, the Fayette County Public Schools (k-12), Eastern Kentucky University, Capital University, and Kenyon College. She served on the APA Board of Officers from 2014-18 as the chair of the APA Committee on the Teaching of Philosophy, and she presently serves as the Executive Director of the American Association of Philosophy Teachers.
We do indeed need something like typewriters in the classroom! Handwriting is another alternative, but many students have trouble writing legibly by hand. Electronic typewriters are unfortunately on the expensive side (around $300). The AlphaSmart is a now-discontinued minimal word processor that could serve as a model for a new device, or one can buy used AlphaSmarts for a song on eBay.