In a comment on an earlier post, a professor suggested as a topic for discussion, “Academic integrity violations becoming commonplace (analog, not just AI!) and what to do about it.”
I’ve certainly heard from many professors about rampant AI-enabled cheating in their courses. Here’s one horror story from an instructor of an online course:
24 students started the course.
7 dropped on the first day immediately after being informed there would be in-person [live] Zoom oral exams. (You can draw your own conclusions about why….)
Of the 17 left, 4 handed in either nothing at all during the course, or at most one assignment (out of five). So, they failed.
That leaves 13 ‘active’ students.
Of these 13, I had to turn in 10 for cheating using AI – and of those 10, 3 continued to use AI to cheat after I had already turned them in for this (and informed them that I had done so). So, they got turned in again. And again. And again. And, in one case, again.
So, of the original 24 students, only 3 did the work required in the course, and did so honestly.
That may be an extreme example, but even some professors teaching in-person courses, including those who have taken steps to reduce the incentives students have to use AI for their coursework, have reported a crushing amount of AI-enabled cheating.
I’ve heard less about an increase in “analog” violations of academic integrity—reports on that welcome— but perhaps an environment in which AI cheating is common diminishes people’s inhibitions against other kinds of cheating.
We’ve discussed in the past ways of preventing AI cheating or trying to make cheating “not worth it.”
That’s all well and good, but one question I’d like to ask about is: What should one do about all the cheating that remains?
Part of the answer to this may come from official university policies. But most of us have the discretion to decide whether to go the “official” route. Do you?
Should we?
How much of a burden will it be to officially report the cheating, given how much there is? Any one case can be time consuming: forms must be completed, extra meetings must be held, hearings may sometimes be required. What if one has 10 or 20 or 50 instances of cheating at the same time?
What is the university’s likely response to the kind of cheating you’re reporting? Are you likely to be supported by the administration? Are the disciplinary measures for those found to have violated the honor code trivial?
What effects will officially reporting the cheating have? Will the student caught be less likely to cheat? Will case confidentiality cut too much against the deterrent effect that reporting one student might have on others?
Is this a collective action problem? One might hope for university-wide measures to curb cheating if enough of it is reported. But such measures may be unlikely to arise if only a few professors here and there report cheating. Any one of us may be skeptical that our own reporting of cheating is going to make enough of a difference, and so be discouraged from reporting, and so no university-wide measures arise. Would it make a difference if faculty members at a university formed a pact to each more carefully attend to cheating and officially report it when they find it?
Discussion of these and related questions welcome.
This is why our department now requires all exams to be in person, even for online courses, and for 60%+ of assessments to be conducted in person. Otherwise it’s unmanageable.
One thing that has happened a few times here is when large numbers of students are failed for AI work but not reported, they form a syndicate and challenge the grade on the grounds that the instructor did not follow the official procedure.
Of course, if you follow the official procedure and report dozens a semester, you get called in for a (friendly but…) meeting with the dean.
Our department is strongly leaning towards requiring all online courses to have in person exams. Where do you teach? It would be good to know of precedents when we make the case to our administrators.
When I moved to Australia, the university I moved to (Deakin, in Melbourne) had a large amount of on-line students. They all had to take exams in-person, in supervised test centers. (They were also done by hand, but that was because the univesity was 15-20 years behind the times, and not using exam software that had been widely available since the early 2000s, not because of AI, which hand’t been invented yet.) So, it’s certainly possible to do that, but your university will have to care about more than whether the student’s check clears or not.
From Michigan State University (https://online.msu.edu/online-students):
“Online and hybrid courses may require regularly scheduled classroom time or face-to-face contact (for requirements such as exams, labs, etc.). Such requirements will be noted in the Schedule of Courses.”
I did not do enough digging to see which online courses currently have in-person exams. However, I know that in the early days of online courses at MSU (around 25 years ago), it was standard protocol for online courses to have in-person proctored exams with ID checks, etc.
We have also pivoted to requiring in-person exams for online courses. For in-person courses, we do paper-and-pencil exams to assess the philosophical skills we’re learning. This cuts down on a lot of the most pernicious cheating, and it seems to make sure that the grades reflect actual learning.
The problem is what we’re forced to leave out. I still do iterative writing assignments (developing an argument paper over time), but students’ learning-avoidance AI use makes these less successful. I’m trying some collaborative papers written in class over several class periods, but that takes up a lot of class time. The point is: students just don’t get the chance to struggle with writing in the way I found valuable (back when we walked both ways up the hill to get to school). So, there are things that work, but the shameless marketing push of AI companies is depriving students of valuable learning experiences.
I work in the math department of a large state university (read: party school), and have been proctoring in person exams for the calculus classes. We have been largely reticent to get involved in “prosecuting” cheaters due to scale issues: providing solid evidence to a dean or honor committee for 20 unrelated cases, three times a semester, in each of calc 1 through calc 3, has been too much work. It puts TAs above contract hours and requires everyone to focus on procedural meetings instead of grading exams.
One other effect is the idea of cheating as a “trump card”: if a student struggles and needs to cheat, or is incentivized to otherwise, they typically do not continue on well. I have also seen cheating cases in a previous institution where intro polisci students “needed the A” for grad school admissions, but did not seem to consider that they need writing skills. When we can see the value of the material for further development (especially some idea of job training), but they cannot, it puts us in creates a principle agent problem, and at the risk of sounding extremely conservative, the best we can do is point out their future failure, and give them the dignity to choose between long term success and short term cheating.
I teach online philosophy classes at a community college. I began suspecting that people were using AI to cheat on the multiple choice exams about a year and a half ago. So, I made multiple choice exams in a way that baited AI to choose a particular wrong answer choice. (If you studied the material, it is obvious that the bait answer choice is wrong). Through this method, I have found that a staggering number of students very likely cheat using AI: around 90%, quite possibly more. For instance, in a class of 40, 38 were confirmed to be very probably using AI on all exam questions. (I made it so that AI chooses a particular pattern of wrong answers, and that exact pattern showed up on 38 exams). The remaining two appears to have used AI only for the ones they weren’t sure about, though this I cannot confirm with high probability.
I no longer report these cases because the sheer number would require a lot of mental energy policing this and increasing administrative work that I don’t have time or psychological energy for (answering angry emails, collecting and collating evidence, etc.). However, maybe I should – I don’t know what my duty is in this regard. I’m at a loss as to how all of this is to be handled, except to move all examinations to in-person in some manner. It’s quite overwhelming.
Cheating is absolutely rampant in my online classes. I redesigned them early on in the AI crisis to take away as much of the incentives to cheat as I could. I give a lot of points for just doing your assignments, so much so that you could blatantly bullshit every single assignment and do no reading and still get a solid C. A C would literally take you about four or five hours of work for the whole class I think. I also allowed rewrites of assignments, so that anyone who was willing to do a bit– not a lot mind you a bit– of work could get an A. But the cheating was overwhelming. So I started coming at the problem from the other direction and I’ve gotten more and more hardassed about it. Right now if I have definitive proof a student used AI they get dropped if we’re before the deadline or an F if after. (I won’t go into my procedure but suffice it to say it’s got a couple failsafes in case of a false positive.) It’s a bloodbath. In one class of 30 I’m down to 5 and in another 9. My best online class is at 14/30. I’m not exactly conflict averse and admin supports me, but it does wear on me. Getting baldfaced lie after baldfaced lie and stupid threat after stupid threat isn’t exactly rewarding as a teacher. (The worst part is when they get AI to write their threatening emails. It’s so wordy and such a bullshitter; I’ve lost track of how many times I’ve been told I’m violating federal laws that just don’t exist.) On the other hand I don’t think it’s a bad thing at all if I’m the reason someone whose response to even the most minor challenge is to cheat and lie doesn’t get to be a nurse or HVAC repairman. Maybe I’d feel differently if I our students went into bullshit jobs like consulting or law. In person classes aren’t nearly so bad, but even there I’m moving to mostly in person assessments. I’m not sure what to do about online asynchronous since that’s not an option. In a better world we’d just stop offering them or at least require in person assessments but given how much money they bring in that’s not happening. If anyone has a good solution for those classes I’d love to hear it. It can’t go on like this, but I don’t know how it stops.
I’m also a CC professor and echo the two above. A few things to add:
This is my fourth semester of giving (almost–see below) no take-home assignments, either papers or tests. It’s either bluebook tests or oral tests for pretty much everything.
The main exception is short weekly assignments, graded pass/fail, counting 1% of the course for each one, adding up to 10% of the course grade total. Some may well cheat on these, but the value of doing so is marginal.
Meanwhile, the tests are designed to distinguish who has done the readings from who has not, which incentivizes them to just do the readings (and then why bother cheating on the little weekly assignments?).
My original reason for eliminating papers was as a kind of deal: you guys do the readings, and I won’t make you write papers. Only later did I realize how awesome it is not to worry about AI cheaters.
I dunno, they manged to do good philosophy in the 13th century, and I’m pretty sure there were no term papers.
One thing I like to do is make a few assessment questions class specific. I give open note, open book assessments and ask things like explain the utility principle. elaborate on one of the examples used in lecture. Or, pick a comment /question one of your peers made in class today and … respond , make sure you use the term ‘categorical imperative’.
I’ve done much the same thing as Prof. Gorman the past couple years: switched all grading in all my classes to 100% handwritten in-class proctored exams. I do traditional exams (combo of multiple choice, T/F, short answer etc.) plus essay exams. The latter are open-book (hardcopy textbooks and lecture notes allowed but no electronics).
So far it seems to be working well. Haven’t had any student complaints, and the grades have been OK (usual bell curve). I’ve seen improvements in writing over the duration of a year.
I think this will have to become the standard in philosophy for the foreseeable future, until we have really reliable AI detection software.
I don’t do any online teaching — not sure how I’d manage things in that context….
That sounds very problematic to me. If we have to have handwritten examinations until AI can be reliably detected, are we examining the things valuable to philosophy, or are we examining things AI cannot replicate easily?
And what then for students with disability, or those utilizing remote learning? Handwritten in person exams starts to exclude these kinds of students who benefit the most from these kinds of technological advancements.
We do not typically “DO” philosophy in invigilated exams, we practice philosophy in tutorial, in forums, in zoom meetings. by reading in our own time and crafting arguments and reflecting and replying to others in real time.
The goal of philosophy should be to evaluate philosophical ability, not create conditions where AI is useless. And then what is being tested? Memory and recall, writing speed, hand writing legibility.
What would happen if latter years Steven Hawking was an undergrad today? Would we punish a great mind to archaic technology that probably made his life difficult, or would we allow him to use the latest AI that made his life less painful?
I am disabled, I cannot type, I am speaking this into the computer using an AI, and have it clean up the spoken mess into something presentable. I am a Phd candidate researching disability and political philosophy. I remember my undergrad in the 90’s vividly and just how draining and excruciating it was to have to apply semester after semester for alternative examinations because I could not hold a pencil to write with, or asking others to press record on my dictation recorder because I could not take notes, or having other students hate me because my high distinctions were from sympathy examinations.
From my perspective, proposals to return to handwritten examinations as the default feel less like a defense of academic integrity and more like a willingness to sacrifice accessibility in order to solve an assessment problem. Part of the problem is that lecturers simply cannot meaningfully interact with hundreds of students in a cohort, and universities often do not want to invest the time, staffing, and resources required to assess students in ways that more accurately measure their individual abilities.
But disabled students can be (and are, because of the ADA) given accommodations for exams like this. My students who can’t write use computers in a designated area. Surely this is a false binary, isn’t it? You don’t need to have the entire class using AI-assisted technology to accommodate the students who need it, do you?
Is it a binary? Or is it a matter of accommodation verses inclusion? They are very different things and from my perspective, its like the disabled have to carry the burden for an examination system that is incapable of dealing with AI use in examinations.
Accommodation says we will treat you differently, inclusion asks whether the examination system itself creates unfair barriers for some. The fact that disabled students can be accommodated does not by itself show that returning to handwritten examinations is the best or fairest way to assess philosophical ability.
There are ways to design inclusive examination systems that are resistant to AI use, but they staff cost time and university money. At my institution, viva voce, zoom meetings and active tutorial participation are replacing some written examinations. Easier to do in 2nd and 3rd year classes that are smaller and cumbersome in first year units with 1000 students. But it costs and it takes effort, but, it is vastly more inclusive than in class written examinations, something I do not think is a valuable way to test someone’s philosophical ability.
Reminder: The comments policy says of LLMs, “please do not submit comments written in whole or in part by these tools.”
I removed all possibilities for cheating in my courses, started a website (https://certifiedaifreeskillsandknowledge.org/) to help students in these no-AI conditions brag and distinguish themselves from AI cheaters, and have restored my love of teaching.
How did you remove all possibilities of cheating in your courses?
I roll a die for each student: 1 is an “A”; 2 is a “B” etc. That determines their grade, completely. I tell them it is a metaphor for the human condition. There is no appeal, and no possibility of cheating.
Students sneak into your office and weight the dice. What now?!
I don’t teach online, thank goodness. I’m tremendously sorry for those of us who are having to do that now. I think I’d be sticking forks in my eyes.
I recently wrote a post for the APA Blog about how I’m approaching this problem in my classes. I’ve decided I have no interest in trying to police students’ AI use. The main reason for this is one that several commenters have mentioned already: it’s just too much damn work. And by “work” I include both the mental energy one has to expend on the task, and the administrative or bureaucratic rigmarole that will often be required, especially if a student appeals.
So I now tell students they can use AI on out-of-class work if they want. But I’ve set up a double-barrelled assignment system in which they have to show the ability to think and write for themselves in the classroom (with no devices allowed). If they can’t, then their grade for the in-class assignments will determine their overall course grade, even if they have juiced their out-of-class grade with AI.
I also do some interviewing of students when it comes to essays. But my purpose is not to uncover AI use. Rather, it is to see if the student can talk to me about their essay. If they can’t, they lose credit.
Can you explain how you set up your grading system so that “If they can’t, then their grade for the in-class assignments will determine their overall course grade”? I’m just wondering how one does this in a fashion that doesn’t lead to endless arguments and appeals.
I know it sounds like a recipe for a total student revolt. But I’ve actually been surprised at how few students have complained. Some do, to be sure, But they’re very much in a minority, and n ever-decreasing minority now that I’m getting better at explaining why I set the class up the way I do.
Students know very well how rampant cheating with AI is. Most of them are not happy about it, I think, even when they themselves have done it. (Like Thomas Nadelhoffer below, I’ve polled students on this a little.) So when I explain that my grading structure is designed to prevent students from benefiting from having AI do the work for them, they get it.
I get this, and I find it quite plausible. Students still want to think.
But can you explain the technical side? Does your syllabus simply say, “If you appear totally clueless in your in-person meetings, then I will ignore all your out-of-class assignments”? Or is there some more technical way that you do it? E.g., “If your grades on your in-class assignments are more than so-and-so many points lower than your grades on your out-of-class assignments, then….” In other words, I’m wondering how you set this up technically/mathematically (if you do).
Michael, at this point I’m going to suggest you read my blog post. Here’s the part that addresses your current question:
In short, this is not just a matter of a few in-person meetings. The entire class is structured, from the ground up, to address the problem. As I explain in the post, I completely overhauled the way that I teach.
Gary, I’m so sorry. I somehow missed your reference to your blog post, even though it’s right there!! SMH.
Gotcha. That’s a pretty hardcore way of doing it, but anything less hardcore probably wouldn’t work at all: the way you do it makes sense, and it’s easy for students to understand, which is crucial.
I have thought of an objection and a reply. The objection is that you don’t know whether you are actually holding them to the same standard on the IC stream and the OC stream The reply, I imagine, is that this is an imaginary worry and that it’s not a problem at all. (And in truth, it’s only in the course of writing this reply that it even occurred to me to wonder whether there is something uneven about how I grade oral and written exams! So yeah, I don’t think this is a serious problem.)
Thanks!
I admit I was a bit puzzled about why you were continuing to pepper me with questions! No worries.
Yes, this is pretty hardcore. It’s been a huge amount of work to implement, but something like this seemed (and still seems) to me to be the only feasible route that still retains some out-of-class work. The work has gotten more manageable now that I’ve got the system up and running and have had a few iterations of it.
FWIW, in most cases the two stream grades come out pretty comparable. It’s rare to see a student with, say, a B in one stream and a D in the other.
I have been teaching remotely full-time since 2021 at the College of Charleston (where I am a full professor).
I began administering an anonymous survey about AI at the start of each term, beginning in AY 2023, in the wake of ChatGPT 3.5’s release in 2022.
One of the questions is the following:
“Have you ever knowingly cheated on written work using AI (either in high school or in college)?”
Roughly 30% admitted that they had knowingly cheated back in 2023.
In Spring 2026, that number had jumped to nearly 60% for both of my courses. I usually have between 100 and 120 students, for reference.
Student responses to other items on the survey have remained fairly constant during the sme time period when their self-reported rates of cheating had skyrocketed. For example, more than 75% students have consistently reported that it is very easy to cheat using AI. They have similarly reported that it is very common for students to use AI to cheat on their assignments.
Also, more than 75% have reported that their professors are not adequately helping them learn to use AI, and that they are not being adequately prepared for the job market.
Importantly, I allow students to use AI in prescribed ways and spend time helping them learn to use it ethically, while enabling them to develop their writing and critical-thinking skills. But that is a pedagogical story for another day.
For now, I just wanted to share the alarming (and depressing) results from my own anonymous student surveys. If anyone wants the full results, they can contact me at [email protected]
I am currently compiling all the results into a single document to share with my department chair and dean. So, I will have something to share soon.
In Fall 2025, I TA’ed for an instructor who assigned a typical essay to be done outside of class due at the beginning of the week, but additionally required that students reproduce the same essay in-class at the end of the week. I thought this was strange at first and there are still aspects to it that are odd in my opinion (what option isn’t these days?!). However, having lived through it, I found that the preparation students had to make to reproduce the essay led to them being remarkably more adept at recognizing strengths and weaknesses to their papers. The conversations that week were some of the best I’ve had with undergraduates in the age of AI. It’s not a perfect system by any stretch and students could still use AI for the initial paper. But, it required that they knew the scope and sequence of the paper as well so if they cheated then they couldn’t do it mindlessly.
Has anyone thought about (or tried to tabulate) a history of cheats? Were there rich, famous and influential cheaters who are beloved historical figures? There are many who cheat or would if it were a safer bet…so to speak.
JFK?
His younger brother, Ted Kennedy, was expelled from Harvard for two years for cheating on an exam in 1951.
Susan D. Blum, an Anthropology professor at Notre Dame, published some of the history of academic cheating in My Word! Plagiarism and College Culture (2009). She reports that Yale had a cheating problem in the 1860s in which, allegedly, half the students were cheating. I think she relies (at least partly) on evidence from diaries.
In Cormac McCarthy’s Stella Maris, the protagonist claims that Carl Jung cheated on the mathematics portion of his university entry exam. I’m not sure whether that’s historically true or not, but it would be an interesting example of this.
I would read that book. And buy a copy for my office, to display it where my students could see it!
Question for those who know some recent history: in the 1940-1960s or earlier, wouldn’t a student’s graded work for a course often be just one or maybe two tests, or perhaps short papers? So, no graded homework for instance; and no long essays. The student had one or maybe two shots at doing well in the assessment, and too bad for them if they screwed it up. I no longer teach, but I think I’d be tempted to give a blue-book midterm and final but nothing else. I’d also point out to them that there are loads of times in life when one has only one short chance to either make it or not, so this is good prep.
The two exam format was not uncommon when I was an undergraduate in the US in the late 90s/early 2000s. They were bluebook exams, no study guide or do-overs or anything, and it was typically a 40/60 or 30/70 split.
Students that did not turn up to class or do the reading generally failed.
I remember being taken aback when I started teaching in the US and a student asked me if I could review the draft of her term paper. I replied I would do no such thing. To me back then this would have been like giving her the answers before the exam. Then I slowly learned this was an extremely common thing and started workshopping papers with students, requiring drafts, peer review, and so on. But in my experience as a student, this would have been very unusual. Same with study guides, do-overs, and exquisitely detailed instructions and rubrics. Maybe these are all very good things, and this is closer to how I teach now, but this is all to say we shouldn’t take graded homework and take-home essays as some kind of ancestral tradition of higher education. It’s historically recent and geographically restricted.
This is still what I experienced in college from 2014-2018, even through graduate school beyond that.
Actually, my schools also counted classroom participation very highly, so it would not break down as a 40/60 or 30/70 between an essay and an exam exactly, but the same spirit was present.
This is very common in the UK. One of my classes is graded 100% a single essay. Another 2 semester class is 50% 3hour exam 50% essay. There are almost never more than one or two pieces of work that determine the full grade. (And in oxbridge a few weeks of exams determine your entire final degree classification (~GPA) )
Most of my classes were like that in France. It may still be true of most classes in the humanities.
To Bryan Frances
1) At Cambridge in the late 1970s when I was an undergraduate our entire mark was dependent the end-of-year exams. We were expected write weekly essays for our supervisors but our mark did not depend on them (and indeed they were not formally marked though they were extensively discussed).
2) At Otago, in New Zealand where I have taught since 1988 the common pattern was some internal assessment –typically a couple of take-home essays – but with the bulk of the mark going on an end-of-semester exam. I gradually moved away from this is in the 2010s to a three take home essays system with plus a semi-optional class presentation. I did this partly for pedagogic and partly for selfish reasons. (I think that, absent cheating, take-home essays are a better form of assessment, and I loathed the process of marking students’ exam scrawls.). However, in the last few years I have been forced to revert to the earlier system to deal with the threat of AI-based cheating.
I think that the graduates of universities and colleges who can guarantee that at least half of their assessment is due to cheat-proof procedures will be at a considerable advantage in the job market.
I think the bluebook might be the least bad option, but I do think we should notice what’s being lost here and what the downsides of the bluebook instead of papers are. 1. Having a professor give you comments is not like “getting the answers before an exam.” Instead, it’s an exercise in learning how to take and use criticism, which is an immensely useful skill. I never had a professor I met with as a student tell me “do x, y, z and you’ll get an A” much less “You say x and y but those are wrong the right answers are a and b so change it”. Rather it was always something more like “You claim that x, but that needs a lot more support. Go back and look at __ in the reading and it might help you” or “This all depends on tacit claim y, which you need to be explicit about. And you also need to think about and respond to this counter” or at the most basic “I’ve no idea what you mean on page 3. Can you explain it to me? “________” “Okay that is clearer put that and not this in the paper.” 2. As to the claim that, “I’d also point out to them that there are loads of times in life when one has only one short chance to either make it or not…” That’s not false, but much more common in the world of white color work is where you have a decent, but limited amount of time to get something done and you need to use it wisely. Writing a paper is much better preparation for that than a bluebook. 3. Papers can test depth of knowledge in ways that I don’t think bluebooks can. They also foster learning in ways that cramming for tests don’t. 4. Bluebooks are really problematic for students with any sort of accomodations. I was talking to a friend whose intelligent and capable but mildly autistic son left my college without taking a degree because even with accomodations he had trouble with timed exams. Accessibility was one huge reason I used to try to avoid bluebook exams entirely.
None of this is to say that bluebooks aren’t the least bad option given the current situation. But let’s be honest that a situation where we are accessing with bluebooks is one that’s worse than the one where we were still able to assign paper.
Of all my frustrations with AI cheating, the reason that made me stop assigning at-home essays is that the AI generated essays were unfairly raising my expectations of what students should be able to produce, and that caused me to more harshly grade the honest students.
Ultimately, I want my teaching to focus on serving the students who are doing the work to learn. From this point of view, the biggest problem with cheating isn’t that dishonest students were getting higher grades than they deserved, but that that the honest students were getting lower grades than they deserved. Since I couldn’t shake the worry that reading excellent AI generated papers created a bias too difficult to overcome, the at-home essays had to go.
Gary Bartlett’s essay at the APA blog is worth a read. This in particular resonated with me:
I do a lot of online asynchronous teaching, and AI/LLM use has completely upended the way these courses can be taught. It also threatens my pedagogical relationship to my students, as I’m forced to constantly be suspicious that the work they’ve submitted isn’t their own. I’ve given a lot of thought about what to do about this, and for the last couple of years I’ve started having occasional required meetings with individual students, and I stipulate in the syllabus (in bold and italics) that they may need to meet with me to discuss the course material and ensure they are mastering the content. So if AI use begins to be an issue or suspected issue with a student, I schedule a meeting to go over what they’ve submitted. It’s led to some of the most pedagogically rewarding experiences I’ve had with online instruction, because the conversations typically go one way or the other very fast: either we have a pleasant discussion about the course content, or it’s an opportunity to speak with a student who is clearly not approaching higher education in the right way. I’ll mention two other things I’ve done recently that others might find worth trying.
This summer I taught an online course where I used the Group function at Canvas to put students into groups of 3-5. Over the term, they completed work together as a group, and each group met with me twice for oral evaluations over the material. It went very well, in part because the students seem to realize they were letting their peers down if they didn’t do the work themselves, and I look forward to trying variations on that approach in the future.
In one of my courses this term, I’ve instituted the following policy: students cannot receive higher than a C+ unless they meet with me one-on-one at least once in the term, for half an hour, to discuss the course content; and they cannot receive higher than a B+ unless they meet with me twice. I had the first oral evaluation meetings three weeks ago, and they went very well. Of the 29 students in the course, 17 chose to meet with me, and while that week was somewhat busy with meetings it was manageable (to scale across a larger courseload the meetings would have to be in small groups, and it probably wouldn’t work in large courses). These were pass/fail evaluations and most of the students did very well, although there were a few who evidently treated the evaluation as if no preparation was needed. The second evaluation, required for receiving an A, will be more strict.
In preparation for the evaluation, I produced a handout for students to study with, and a question bank I could use to guide the evaluations with individual students. Both of these were produced with the help of AI, and I told my students that they are welcome to use AI in the study phase of preparation, but that the evaluation itself would be without notes and without AI assistance (the prohibition against submitting AI-generated work remains in place).
My experience the last few years has been that AI/LLM use is rampant among students, particularly in online courses, and outside a handful of specialized topics I can’t see how asynchronous online education is going to survive without more in-person evaluation. It’s also a much better (and more rewarding) way to evaluate students over the material (I wrote up some remarks at Daily Nous in favor of this sort of thing during the pandemic quarantine: https://dailynous.com/2020/04/06/now-may-time-transition-tutorial-instruction-guest-post/).
This is intelligent and generous.
I can’t help noticing that when you say, “I can’t see how asynchronous online education is going to survive without more in-person evaluation,” the upshot is that asynchronous education can’t survive without becoming at least somewhat synchronous!
Agreed. The asynchronous mode was never a great way of teaching, and with new media technologies being so advanced and ubiquitous today we should be looking for ways to integrate some degree of face-to-face interaction in every course.
The teaching profession is obviously not good at keeping up with new technology. Instead of fighting AI, teachers need to teach students how to use AI effectively. In the future, businesses who use AI will completely dominate the businesses of luddites who try to do things the old way. The best way to prepare for the modern world is to learn how to use AI. Online courses that require essays or that give multiple choice questions are obsolete. Deal with it. You can mourn the loss of the old way of doing things, but you won’t be able to change it.
Unless you are teaching an innovative new course on the philosophy of how to use AI effectively, why is it our job to teach philosophy students how to use AI effectively? It certainly isn’t my job to teach them effective ways to “dominate” others in business when they graduate. It sounds like you are suggesting that philosophy itself is obsolete.
Anyone can “use AI” in a relatively thoughtless way. To use it really effectively requires someone whose cognitive powers are already well-developed. And acquiring well-developed cognitive powers comes from thinking on your own, not from using AI to think for you.
Whether or not our students will, at some point, need to use AI in their jobs, their task when in school is mostly not to use AI but instead to think on their own, thereby becoming smart and disciplined.
There’s a flip response to this and a more serious one. The flip response, which is probably all it merits since I strongly suspect trolling, is “Just how long do you think it takes to teach kids to copy paste? I think we can get it down in a morning even if we have to cover Macs and PCs!” (Also, before you go bashing the Luddites, who are awesome, you need to read Mechant’s “Blood in the Machine.”)
The more serious one is that the absolute bare minimum of what employers expect from college graduates is that they can do their work, in a timely fashion without constant babysitting. A student who gets through a class– even an online one– in the usual way does a lot to show that those claims are true. That even goes for those who halfass it. It’s not true of kids who cheat using AI. On top of that, I doubt that a lot of the jobs that my students will take will require them to use standard issue LLMs at all. Now an X-Ray tech or nurse may well need to use specialized programs that might be best described as AI, but God help us all if the powers that be expect philosophers to teach them how to use those. What we can do is try to teach soft skills like basic adulting and, ideally, critical thinking and research. What’s going to happen if we abandon all that in the name of some ill defined “teaching AI skills.” Well how long do you expect employers to pay a premium for us handing out degrees to illiterates who respond to criticism with pompous, prolix, bot-written nastygrams?
Imagine if physical trainers said “instead of fighting heavy machinery, we should understand that the best way to prepare gymgoers for the modern world is to learn how to use heavy machinery”.
It certainly is important for some people to learn how to use heavy machinery in some ways – and perhaps nearly everyone in some ways. (Very few modern adults never learn to operate a car, elevator, or escalator, though it’s true that some never learn to operate a bicycle, and most never learn to operate a forklift.) But the boom in gymgoing over the past several decades resulted because people realized that it *wasn’t* enough to learn how to use heavy machinery – developing your own body’s abilities is also very useful, whether because it is an important adjunct to machines, or just because it’s hard to be a healthy human being without using your own body in some significant ways.
We need to get ahead of the same issue for the use of mental machinery.
There is a broader problem with education on this topic, which is credentialing. The best thought experiment I’ve read recently asked why, even though machines that lift weights have existed for centuries, we have no problem with people bringing machines to gyms to cheat at weight training.
The reason is that the strength and fitness are the goal, not the weight being lifted. The problem with much formal education, is that it’s viewed as worthless even by its students. The credential qualifying them for a job is the goal and they expect to learn nothing of real use. This isn’t new, look at how students celebrate snow days, while gym goers lament missing a day. The only solution I see is formal education where the point the knowledge taught is the point, which will be painfully smaller and divorced from employment credentials.
“…we have no problem with people bringing machines to gyms to cheat at weight training.”
Did you mean “…we have a problem…”?
That seems a better fit for the rest of your comment.
No, they’re saying that people don’t bring such machines to the gym; we have no problem with that happening, whereas we have a problem with people using AI to cheat. There is not a problem whereby the bringing of machines to the gym keeps happening.
I stumbled on it at first too.
Ah, I see.
“we have no problem with” means simply, “this is not a problem that arises,” rather than “we do not object to.”
Yes, you’re probably right.
I might have seen the solution myself, except I was too busy aligning my fork-lift with the Smith machine.
We also wouldn’t care if someone did do it so I think the comment can be interpreted either way.
It would be interesting to study the (small) populations that do need a formal credential from a gym (perhaps this is a certain class of military recruits?) Do we see significant problems with attempted cheating in that population?
And do they cheat by using a forklift to lift the weights?
I teach at a community college, and over half my load is asynchronous. We use Respondus Monitor to proctor online exams. Last year, my sense was this was largely useless, in that students could easily do things like hide their cell phones out of sight of the web camera.
This year, though, we’ve had the option of requiring a “dual camera” set up (with webcam plus a mobile phone monitoring the keyboard), and my overall sense is that it has cut down on cheating quite a bit. Specifically, when I look at essays that show signs of AI use these are (almost without exception) students who turned off their second camera (or pointed it at the ceiling, etc.). This makes it a lot easier to deal with since I can just say “Well, a requirement was that you used a second camera. You didn’t do this.” There’s no need to have debates about trusting AI checkers or the like.
I’ve also noticed that essays and exam quality has returned to (something like) their pre-AI norms.
This isn’t perfect, and I anticipate (some) students will find work arounds. And, of course, this doesn’t work for all sorts of assessments.
Thanks for the update on Respondus. I think this what I’m going to have to do too. That or just refuse to teach online.
The only mass cheating I can remember in the news involved our military’s academies. Noplace else. I don’t doubt groups of cheaters get tossed out now and then, but it’s obviously not as newsworthy.