Recommended Reading: "To teach in the time of ChatGPT is to know pain" by Scott K. Johnson

Today, I wanted to share some work from a fellow educator who is also struggling with how generative AI has "disrupted" education. It's titled, "To teach in the time of ChatGPT is to know pain." I hope you'll take a moment to read it before you come back here to hear my thoughts. How I Found This Article As you might know, I like to kick off my recommendations by talking about how I found the article. In this case, I found Scott's work while doing some research for one of my own articles, which should be published by the time this piece goes out. In that article, I talk about deception as a fundamental aspect of generative AI. It's why I think people generally don't like it but don't have the words to explain why. It turns out that being deceived doesn't feel good. Anyway, I found Scott's piece while trying to look for some evidence that students are deceiving their instructors by using generative AI (and conversely, that "educators" are doing the same). I didn't necessarily need statistics because the impact of the technology is obvious, but I did want folks to hear from other educators. So, Scott fit the bill. To be honest, it's rare that I read an entire article. Usually, I skim them to get the gist, but Scott's piece is excellent. As a result, I'm going to spend the rest of my recommendation talking about why. Article Summary If you don't have a chance to read Scott's piece, here's a quick summary. Scott starts by giving a bit of background about his work (i.e., he teaches Earth science courses). Like other educators, Scott admits that he got into teaching because it's fulfilling, which he uses to springboard into the meat of the piece: "But thanks to generative AI, [teaching] has become mostly miserable—at least in certain settings." That last note (i.e., "at least in certain settings") is a key piece of his argument because Scott mostly teaches asynchronously. Like Scott, I find these kinds of courses painful, and he does a great job explaining how generative AI has made teaching these kinds of courses much worse. Specifically, it's shifted the role away from teaching and toward discipline. In other words, much of his time is spent figuring out if an assignment was completely honestly. Next, Scott talks about why he thinks students turn to generative AI to complete their work: students think the goal of education is to get the correct answer. I've even complained about this myself, and I would have loved for Scott to dig into this topic deeper (e.g., why are students fixated on the right answer?). Instead, Scott spends much of this section talking about the importance of "friction" and why learning requires work. He uses this argument to then rant about how LLMs can remove all friction by completing any assessments that would normally require deep thought, even the low stakes assessments that are graded for completion. This causes Scott to question whether the effort on the instructor's end is worth it. In the following section, Scott builds on this frustration by talking about solutions. For example, should we only use in-class assessments? Scott demonstrates how this doesn't work in his context (i.e., asynchronous course work), and he would hate to see asynchronous classes disappear as some folks need them. While talking about the variety of solutions that don't really work, Scott shares how frustrating it is to constantly be told that he just needs to make better assessments. He then shares some frustrations he has with the lack of support he and others are getting from administrators while also acknowledging that basically every instructor is screaming from the rooftops about generative AI. Scott finishes out his article by addressing a few of the common talking points directly. For example, he talks about how even math instructors ban the use of calculators as they get in the way of learning. He also talks about how LLMs make horrible tutors. He then talks about his vision for education, one which guides students "up the mountain," but LLMs push students "every direction but up." Finally, he wraps up his piece by arguing that there isn't a single student who thinks they're learning when they use chat bots. They're simply using them to manage their workload. Until the AI bubble pops, this will continue to be a very frustrating reality for educators. Why I Like This Article Aside from hoping that Scott would have dug a bit deeper on this idea of prioritizing correctness/grades, I think Scott did a phenomenal job. In fact, he writes a lot better than me. Each paragraph has a nice flow, and his language is really descriptive and evocative. What I really liked about this piece, however, is that it contained several wonderful nuggets that I would like to dissect in this section. First, the sentence "it's to moonlight as a detective and prosecutor..." jumps out at me. I don't know how Scott handles cheating now, but I can tell you that the academic integrity board at my institution no longer lets us submit AI cheating cases. There's just no way to prove it, and false positives (i.e., "convicting" students that did not cheat) are far worse than false negatives (i.e., missing students that did). Like, an expert's intuition is probably enough evidence, but people are starting to behave like chat bots. So, I'm not sure where to go from here. Of course, Scott acknowledges this problem later in the piece when he says, "it’s all incredibly difficult to prosecute by our traditional standards of cheating because there is no incontrovertible test for LLM use." Second, the sentence "it leaves me with the disturbing thought that even my engaged students might not be what they seem" is basically the exact argument I try to convey in my piece on deception. Once enough students use AI to cheat, it becomes impossible to trust even good students. After all, if someone deceives you (e.g., your partner cheats on you), it becomes that much harder to trust others. Third, I just really like the section heading that reads, "Do or do not, there is no AI." If you're familiar with the classic Yoda quote, "Do or do not, there is no try," you'll appreciate this heading as much as me. If not, the joke is basically that using AI is the same as doing nothing, though I think Scott would argue that it's worse because it wastes everyone's time. I agree wholeheartedly. After all, I am legitimately starting to wonder if students are passing my feedback directly into an LLM. Like, are students even reading their feedback? Fourth, I really like when Scott asks, "what’s the point of building formative assessments into a course if they’re just handed off to an LLM?" I identify with this at a spiritual level. My course has 37 written formative assessments. I always warn my students that this is a lot (and not my idea), but that they're there to help them learn. As a result, they're graded on completion. However, recently, I shifted to a "graded on professionalism" model. Basically, if the assignment doesn't look like an honest attempt (i.e., clearly AI or just generally half-assed) and/or the submission is not well formed, then we—as in the graders and I—don't give them credit for it. Though, I do let them resubmit. I figure the assignment is worth so few points that the student won't complain if we call them out for AI usage, and usually it is obvious (e.g., drawing perfect binary search trees with ASCII characters or calling methods from an API that doesn't exist). Fifth, I also really like when Scott asks, "should instructors preserve these sorts of assignments for students who want to benefit from them and accept the cheating, or should they eliminate the learning opportunity just to prevent cheating?" Again, I think this is the problem now. I had a similar concern when the government started telling university educators to meet certain accessibility requirements with their classroom resources. I get the premise, but we're not equipped with the skills to do it. As a result, I fully expect a lot of faculty to stop sharing certain resources because they don't meet the requirements. I suspect a similar thing is going to happen with assessments in relation to the effects of AI. Before long, we'll be back to oral tradition. Sixth, I love this throwaway line Scott shares, "as if the Internet wasn’t bursting at the seams with human writing that one could critique!" Here, he's making fun of the one "effective use" case of LLMs for assignments, which is apparently to have it spit out work and critique it. This is, of course, hilarious because the work produced by an LLM is supposed to be perfect yet bad enough to invite critique. Anyway, I like the little throwaway line here because it's kind of silly to be like: "use the AI to produce slop that you can critique when there is perfectly good human work worth discussing." What are we honestly doing here? Lastly, while there are several more nuggets I would love to discuss, I'd hate to just rehash the entire article with you when you can just read it yourself. Instead, I'll finish with this: "it too often feels like the only winning move is not to play." This is the really sad reality. I've personally kind of given up on fighting LLM use in the classroom, and I wouldn't be surprised if I was eventually asked to teach students how to use generative AI. I am, of course, in computer science where there isn't an ounce of critical thinking on this issue. Despite the public's negative perception of AI, pushing against AI in computer science kind of feels like being a heretic believing in heliocentrism in the 17th century (yes, I'm referencing Orb). Any Recommendations? As I mentioned in the previous section, we're fortunately at a time where a lot of people hate AI. People are sick of the data centers popping up in their towns, the misinformation showing up in their feeds and searches, and the layoffs affecting their careers. For a time, I was worried that maybe educators wouldn't get the memo either, but Scott helped convince me otherwise. After all, take a look at the comments: It really is not always about making things easier, the work, the journey, is more often what we need to become more fulfilled and human not the destination or goal. GraphicH, 2026-04-13 The problem with LLMs is that they take the old "they wouldn't print it if it weren't true" argument and amplify it by a million. Not only are they untrustworthy plagiarism machines, they also tend to lead people to abandon critical thinking and assume that the LLM output is correct. (It often isn't, as you well know.) JohnDeL, 2026-04-13 People using AI aren’t doing work. They’re commissioning it, just like a rich patron commissions art. Of course they learn nothing. They did nothing. markgo "Yes, education got destroyed, along with a lot of other things that makes societies work. But for a beautiful moment in time we made a few tech billionaires the richest people in history." ubercurmudgeon, 2026-04-13 Disheartening when assignments that used to return creative, good products (and lead to good questions) now return substantially the same midrange answers and phrasing. NedKrist, 2026-04-13 I'm a big fan of that last quote. Generative AI does feel like it's just averaging all of the students; the great equalizer, some might say (tongue-in-cheek, of course). Work quality is improved for the worst students and worsened for the best students, while every student becomes a worse version of themselves. There is just no way this is sustainable long term. Anyway, the point is that the comments are generally on the side of the other. The few "it's not as bad as you think" folks are getting clowned in the comments. Like, this comment had 206 downvotes compared to 13 upvotes when I saw it: I think we should just look back to how we did this 25 years ago. We all used Sparknotes, and most problems asked by professors were answered in the spark notes. You could literally copy/paste whole sections and just move sentences around and change some words and get a good grade. Fraternities and academic societies have maintained “test banks” for years.Which is all to say, this is probably not as big a problem as it seems. Exelius, 2026-04-13 Like, is it not obvious that friction matters, as both Scott and Hank Green have said. There was probably a time when the only way to cheat was to go through the humiliation ritual of asking for someone's help, and that was infinitely more valuable to you than an LLM could ever be. Even then, often it would be easier to just do the work. And guess what? Developing expertise makes future work easier to do. Funny how that works. I swear my article on how to "ethically" cheat seems like a genuine learning strategy these days. Anyway, this is the point in the article where I ask you for recommendations. If you liked this piece (as in Scott's), I'm happy I shared it! Now's your chance to share one with me, preferably on Discord. Though, you can find other ways to get a hold of me on my list of ways to grow the site. In the meantime, I have plenty more for you! Though, this series is fairly new, so here's all I have from it so far: Recommended Reading: “Tech Bullshit Explained: Uncle Bob” by Melissa McEwen Recommended Reading: “Some Final Thoughts on the SAT and ACT” by Jon Boeckenstedt If you send me more to read, I'll keep the series going! Otherwise, take care. I wanted to sneak a side note down here where I echo Scott's point about not being able to even trust good students anymore. It's what makes it really hard to even trust that Scott's article is legitimate. After all, I've heard (though search engines are useless now) that there might even be bot farms generating anti-AI pieces, so that's cool I guess. Hell, the pope's anti-ai manifesto might even be AI (though, I think these kinds of witch hunts using AI tools to assess AI writing are an exercise in absurdity). Anyway, I bring this up because this concern around constantly being deceived puts me in a state of hypervigilance. For example, what happens if I just spent all this time praising the work of a bot or agent? That doesn't feel good. It makes me feel stupid. It lends credence to all the AI sycophants who say things like, "who cares if a human wrote it? The content is all that matters" (and yes, I saw this exact argument recently when someone claimed they didn't care if a story was written by a bot as long as it was good). But, I don't think the solution is to disengage entirely. It seems bad for a society to assume that people are trying to deceive you, even if these "thinking machines" are trying to advance that idea. If I'm wrong, there's no shame in that. I'd rather continue to build empathy, trust, and revolutionary optimism (sometimes with bots accidentally) than succumb to cynicism. Otherwise, a better world is not possible. I hope others will take that trade as well.