It seems that ChatGPT has inserted itself into every nook and cranny of the classroom experience. With easy answers (and entire term papers) at their fingertips, students are no longer doing deep thinking required to learn something well.
This effect will reverberate in the coming years, as the “ChatGPT babies,” now in lower grades, will one day make their way to college — if they even attend college. With an attenuated level of thinking, they’ll struggle to process information the way that today’s college students take for granted. In other words, it seems if we are at the beginning days of a knowledge apocalypse.
Or so we might be tempted to think.
The reason for this is simple: we are letting students lead the way with how ChatGPT and other AI are used. We are letting them prescribe, unwittingly or not, what we as educators end up perceiving as the limits of the technology. In short: if all we see are students using AI to cheat and take shortcuts, it’s easy to bandy about terms like “knowledge apocalypse.”
Instead, educators must understand the power — not just the perils — Frontier AI models like ChatGPT; these models can actually enhance the classroom learning experience. Rather than shutting down or proscribing its use, teachers should actually invite ChatGPT into the classroom. Indeed, they should make AI an integral part of the learning experience.
What do the studies say?
If this sounds far-fetched, if not downright misguided, consider the following studies. In one study, researchers determined whether using ChatGPT for feedback on their writing could help ESL students become better at writing in English. The study (Mahapatra, 2024) found “a significant positive impact of ChatGPT on students’ academic writing skills.” Students in this study also found the impact of ChatGPT as being “overwhelmingly positive.”
Another study by two MIT researchers (Noy and Zhang, 2023) discovered similar effects. Here, they found that subjects who used ChatGPT to aid them in the writing process (not do the actual writing for them) produced work at a rate 40% faster and of a quality of 18% better. While that latter result might be a little more nebulous to pin down, that 18% translated into writers producing work that was as good as highly-skilled writers in their workplace. Essentially, ChatGPT can make you a significantly better–and faster–writer.
Now, had I removed the name ChatGPT from the studies above and replaced it with “a new technology,” educators would likely be very intrigued to, at the very least, have students experiment with it in their classes. However, as I mentioned, ChatGPT is so freighted with baggage around plagiarism and student cheating that any reportings around its ability to significantly boost learning outcomes is often reflexively dismissed.
These studies, if objectively engaged with, open up the exciting possibility of ChatGPT as a “thought partner,” not a “thought thief.” And while it is understandable that many educators continue to cast it in the light of the latter, the sooner teachers embrace ChatGPT as an extra brain in the classroom, the sooner they are poised to help students learn and master material potentially faster than ever before.
It’s not that easy
Of course, just dropping ChatGPT into the classroom and expecting learning to be transformed is woefully misguided. Firstly, teachers themselves will have to learn how this powerful technology can be used to help their aim of educating students in a particular subject. Secondly, they’ll have to understand how to weave AI into their pedagogical workflow so that they are empowered, not undermined, by the technology.
Lastly, the looming specter of cheating and plagiarism isn’t going away, no matter how well-meaning our intentions. We will need creative solutions to address it.
In this article, I will discuss five bold and innovative ideas that show how ChatGPT can be integrated in the classroom to enhance learning, decrease cheating, and empower educators. The endgame here is a world of collaborative learning between teachers, students, and AI, rather than a world where teachers fear their classrooms have become a Wild West of shadowy AI use.
Idea 1: Turn in homework as ChatGPT thread
Teacher workloads seem to be becoming more grueling, much of it often spent reviewing homework. With the advent of ChatGPT, many have given up even trying to monitor students’ work. The inconsistency of AI checkers certainly doesn’t help, leading to both false positives (suspecting a student of plagiarizing when they haven’t) and false negatives (not catching students who do cheat.)
Imagine a different world, one in which students are encouraged to use ChatGPT. (Yes, I realize this may take quite a bit of imagining for some of you, but bear with me for a moment.)
Here, students are taught how to use ChatGPT to help them better ideate, better generate that first draft, and better iterate along the way, with the AI providing useful feedback and encouragement. This entire process will be documented in the thread that a student submits to a teacher, something that will be even more practical to online learning programs. Students will then be graded on how they interacted with the AI along the various stages, as well as their final output.
The advantages here are twofold: first, submitting a thread of the actual ChatGPT interaction makes cheating a lot more difficult. Students can’t just have ChatGPT create a fake dialogue. Having another LLM come up with an output to be plugged back in becomes cumbersome over the long exchange required for such an assignment. It’ll take a very determined student to cheat (for one, they’d also have to check the outputs of each LLM.) The net effect should be a significant drop in those students using the AI dishonestly.
Secondly, using this method can boost learning outcomes. For one, it can help meet students where they are at, helping them fill in any knowledge gaps. And, like the ESL study, it can help the students improve upon their writing, by getting direct feedback, thereby personalizing learning. Whoever is grading the student-AI collaboration will be able to see this iteration of writing happening.
As for those grading the papers, they, too, will be able to use AI, grading the thread and evaluating. Of course, they can and should read the final output. But AI can at the very least help in maintaining the integrity of the assignment and can sometimes even surface valid feedback that a teacher might miss.
And while those boosts might not turn a ‘C’ student into an ‘A’ one, they have the potential to turn that student’s writing output into a ‘B.’ Remember, this isn’t ChatGPT writing a ‘B’; it’s a student asking for specific feedback to take their writing from a ‘C’ to a ‘B.’’ The best part is this newfound writing know-how won’t be lost, as each student is able to apply, going forward, the writing knowledge they learned in their interaction with the AI.
Idea 2: Train the educators now on how to use ChatGPT to enhance learning
Idea 1 will be really hard to pull off without first training teachers on how to use ChatGPT to boost learning outcomes and as a thought partner. This might sound daunting; after all, any new technology requires a learning curve. This can often be steep, depending on the technology.
The good news is that generative AI trailblazer and thought leader Ethan Mollick (author of the popular book Co-Intelligence) says that all you need is one long sleepless week to get up to speed on how to use Generative AI. By his well-informed estimate, becoming relatively proficient at these models takes about 10-ish hours of playing around and experimenting with them.
To many that can seem daunting. The good news for educators is that they don’t need quite that long, given that a specialized training around most of their specific use cases might only take around 5 hours — and take them much further than 10 hours of experimenting by themselves likely would.
5 hours is not too much of a lift and, most excitingly, it positions educators as leaders with ChatGPT in the classroom. Some might argue that students are light years ahead. While it is true that they’ve probably spent more time using and tinkering around with these tools than the average educator, they have not done so with learning in mind. Educators who receive specific, properly targeted training will bring with it their deep pedagogical insights, something an AI most certainly can’t replace.
But by “partnering” with an AI, educators will show students how this exciting technology can be used to enhance the learning experience. They may, for example, give in-class assignments where students compete to create the most clever prompts or the most successful output. Instructors could also ask their students to figure out innovative ways to learn a concept; or to come up with an assignment that will best poke at a student’s learning. And these hypothetical use-cases are all just the tip of the AI-infused learning iceberg.
Idea 3: Have students check their knowledge state using AI
Most syllabi have something in common: they presuppose that all students start from the same point. Of course, the reality is very different; inevitably, students’ knowledge of a subject varies widely.
Most courses move at a rapid clip, and students can find it difficult to fill in their personal knowledge gaps. Often, some get left behind and can’t quite make it over the bump, even though they might pore over the textbooks. Retreading course materials won’t necessarily help, because students aren’t going to be aware of where their knowledge gaps lie. For so many hard-working teachers and professors, helping individual students figure out all of their exact gaps simply isn’t viable. Short of hiring an expensive tutor, students struggling to discover and overcome their knowledge gaps may feel they have no clear path forward.
All of that can change with ChatGPT. If you are a student, you can feed ChatGPT the syllabus, ask it to quiz you on a specific area, input your own knowledge on that specific area — your “knowledge state” — and it’ll be able to tell you where some gaps in your knowledge might lie. From there, students can burrow down further into those areas in their conversation with the AI. Suddenly students have a tool to hyper-personalize their learning to such an extent that they are able to plug in their knowledge gaps by simply asking ChatGPT where those gaps lie.
While this is certainly a powerful technique for particularly capable solo learners, learning to leverage ChatGPR in this way in a classroom and having an instructor or TA mediate this process is a huge win for those who might otherwise get left behind.
These kinds of AI-supported knowledge state check-ins aren’t just useful at the outset of a class. Throughout the quarter or semester, whether to prepare for a big exam or just to ensure that reading sticks, a student can work with ChatGPT and similar AI to make sure that they are where they need to be.
There, of course, is a little more nuance to it than the basic process described above, but that is exactly where an educator comes in. Basically, they can help ensure that a student is not only thinking of ways to determine whether learning is sticking but also how to partner with AI for support when learning isn’t progressing as fast as expected.
A powerful technique around assessing one’s “knowledge state” that both teachers and students can use is what I call a “thread synopsis.” After engaging with ChatGPT or a similar AI, ask it to see if it can identify and summarize any of your knowledge gaps — the longer the thread, the more powerful this method. Upon request, this kind of AI platform can comb through your interaction with it and find any areas in your understanding that might be a little shaky. Once it identifies these trouble spots, you can home in on them and shore up any weak areas.
Teachers can incorporate this into the syllabi by reminding students to do periodic knowledge checks after an assignment. Additionally, students can always follow up with a teacher or TA to see if the AI missed anything or didn’t focus enough on a certain aspect of the syllabus that will be very important to that student’s final grade.
Idea 4 — Future proof students for our AI-infused world
We are moving towards a future (assuming we aren’t there already) where what will matter most is the final output a student is able to coax from the AI. Granted, for many in the educational sphere, this may be far from an ideal outcome. The idea that a student is partnering with an “alien intelligence” to help them do their work violates the time-honored tradition of self-sufficiency that comes with learning.
Yet, unless someone puts the lid back on the AI-genie, we likely will be living in a new AI-inflected reality for the foreseeable future. It should also be noted that students have long collaborated with others; collaborating with AI, when not outright cheating, is just another form that, in the coming years, will likely become increasingly normalized.
And rather than see our relationship with AI as a net negative on learning and self-actualization, I suggest a reframing: our relationship with AI is, when it’s at its best, is a “thought partnership.” The AI meets us where we are currently at — our strengths, our foibles, our habits, our quirks — and when used ethically and responsibly, allows us to be an even better version of ourselves.
Those who realize this sooner and who are better able to leverage the models more quickly for these ends stand the best chance of success in the new, AI-infused future.. In essence, they will be able to get the best final “outputs” — be that landing a highly sought after job or being able to accomplish important tasks skillfully and efficiently. And what better place to fine tune this ability than the very place that has long served to prepare the younger generation to enter the workforce?
The way students tend to use AI right now — and I realize I might come across judgemental or over-generalized– is not to use it to make themselves more prepared for the future. Rather, by turning to AI for cheating and lazy solutions, students are making themselves less prepared. Or, in AI parlance, their future outputs will be generic. That is, if students are allowed to use ChatGPT to do the majority of the work for them, their ability to compete or even simply function in the real world will be unequivocally diminished.
Instead, students should use ChatGPT and other platforms like it in the ways espoused in this article. And there are no better guides to the use of AI to learn than educators. But first the educators must themselves be trained on how to responsibly and effectively use AI for learning, as outlined in Idea 2. If we fail to take this path, we potentially set up future generations with the incorrect expectation that AI can do most of the work for them.
Conclusion
We are currently at an inflection point in education, one where we can dig in our heels and push back against the seeming onslaught of AI. The other possibility might seem to capitulate, resigning oneself to the perceived reality that nothing can be done.
But there’s another path: we invite AI into the classroom, into the learning process, so that both can be enhanced. Granted, such a transition will need to be managed wisely. For one, educators themselves, many who’ve pushed back against the technology, will need to learn and even be trained on how to best leverage LLMs like ChatGPT for learning, thought partnership, and the creative use cases outlined above.
Speaking of these use cases, I believe they represent just a small (though important) part of how we will savvily bring AI into the classroom. In the coming months, we will need to think of yet more creative ways to bring ChatGPT into the classroom. We will need to collaborate, share notes, and even invite ChatGPT and other LLMs into the conversation.
The alternative, as we’ve seen over the last year or so, is one in which both students and teachers are losing out — the former in potentially undermining their learning, the latter in pushing back against a tool that can, with the right know-how, actually be used to enhance the very mission of education: to better equip students for the world of tomorrow.
Comentarios