Education Meets AI: It’s Good — Really.
At the risk of making tired analogies, when I was in middle school we learned how to use a slide rule. Not much later the electronic calculator was introduced, and the slide-rule became irrelevant for all but those who had an attachment to its tactile qualities. By the 1980s, slide rules skills and manual arithmetic were mostly not useful. A similar transformation is underway, only it’s much bigger and more exciting.
The sudden emergence of artificial intelligence into our collective popular consciousness, the “zeitgeist,” in late 2022 has shaken us up as if space aliens had finally landed on Earth. AI has been covered in the media for both its beneficial and potentially scary consequences. Concerned technical, business, and academic leaders have been warning that AI could do more harm than good and that those who are building these digital minds need to take responsibility for preventing very bad things from happening. That’s an entire subject on its own, but for now let’s talk about how AI is already affecting education and where education may be heading.
Let’s start with the big problem: cheating. Already, the majority of high school students when polled anonymously, admit that they are using AI to complete assignments. That could be mostly reports and essays, but they could also be using AI to complete math problems and other types of work. I am the technical manager of an online digital high school, and I can tell you that students are trying to pass off AI compositions as their own, and with no common sense. It’s obvious that when a student completes dozens of assignments for a single course within 24 hours, they haven’t actually completed either the writing or the extensive reading on which that writing is supposed to be based. When we recognize this kind of behavior, we need to take action. It just isn’t fair to other students who do the work, or to the students sidestepping effort themselves, who are only impoverishing their own intellectual development. I’m not convinced, however, that monitoring millions of students, with or without technological detection, is going to solve the problem.
Get Scott Jarol’s stories in your inbox
Join Medium for free to get updates from this writer.
When I first accepted my position, my CEO asked me to concentrate on how to prevent cheating. I told him that it’s a losing battle — or a lost battle. Lost war? All kinds of clever people are claiming they have systems that can detect material written by AI. They can’t. Not reliably. I can ask the most popular online AI agent, ChatGPT, to write a book report using the grammar and vocabulary of a ninth grader, and to sprinkle in a few decoy errors. When we tested one system’s ability to differentiate it got the answers backwards — it identified the human composition as AI generated, and the AI generated essay as written by a human. In our non-scientific experiments, detection has been 50/50, equivalent to random guesses. Unless all AI, by every organization working on it incorporates some kind of identification system, such as “watermarking,” there will be no way to enforce academic integrity. Instead of engaging in escalating leapfrog, we need to change our approach to teaching and our expectations of what students need to learn to be successful in their lives, work, and relationships.
Which brings me back to slide-rules and calculators. When electronic calculators we shifted math education away from an obsession with accurate calculation, and toward teaching mathematical methods. At the high school level, many students may have questioned why they would ever need algebra, trigonometry, geometry, or calculus, but as they advanced toward various careers, some significant percentage of them discovered how useful these skills could be. From carpentry to rocket science, math is useful, arithmetic is best left to machines, which make fewer mistakes and speed up the final steps of applied math. A carpenter building a house with odd angles doesn’t need to waste her time on the numbers. She only needs to understand how to use basic concepts like the Pythagorean Theorem to punch in the right calculations.
Maybe AI is the equivalent of the electronic calculator for writing, manipulating words for us instead of numbers. It’s heretical to challenge the idea that kids may not need to develop strong writing skills, but there is an important distinction between thinking and expression. It’s an imperfect analogy to math because I’ve just argued that math skills are valuable across a spectrum of professions, and I could just as easily argue that even then AI could do the work. We transitioned from manual calculations to slide rules, to calculators, to personal computers, and then to miniature computers we carry in our pockets. Computing tools can do for verbal communication what they’ve already done for math.
If a student can identify the essential ideas in a novel they’ve read, or a complex social problem, or a scientific question, and they can also name the points of an argument they may wish to make in favor of one position or theory or another, then is it necessary for them to express their argument in well-crafted prose? They could just as easily collaborate with an AI agent and use that system’s extensively trained language model to organize their thoughts into a coherent argument. A scientist proposes an hypothesis, and then tests it by designing an experiment or making observations, which become data. The data either supports or refutes their hypothesis. To present the outcome, they write a publishable scientific paper. Some scientists are great writers and their papers are not only informative, but even entertaining. Others write dry accounts. Regardless, the narrative is not the point of the paper. The interpretation is what matters, and it would not, and already does not require a human author to draft the text.
If you haven’t used an AI agent yet, you may not be aware of how advanced these systems have become, seemingly overnight. I am amazed myself every day. When I have a program to debug and I can’t spot the error, I copy my code into ChatGPT and ask it, and I mean literally, as in “Why does the following Python code fail? <followed by my pasted program code>” and it not only explains my error(s), but rewrites the code for me so I can copy and paste it back into my programming editor. It almost always works. ChatGPT and other AI programming tools have saved me days or weeks of wasted effort. Our team also uses ChatGPT to write business documents, including drafts of legal agreements, and we’re about to start using it to help us design our curriculum because testing has shown us that it does a brilliant job.
Students need to learn how to use these new tools to their advantage, not for cheating, but for self-expression. AI has landed and its shockwaves require radical changes in how humans use their own brains. This is not concession or abdication to our robot overlords. It’s just another technological revolution introducing tools that can free us from more of our less interesting tasks so we can apply ourselves to what matters most. Resisting these changes (I am not going to quote the Star Trek villainous Borg queen), is hopeless. We’re already here. I, for one, am less interested in catching our students at bad behavior than I am at figuring out how AI will help them learn with speed and agility so they can be brilliant at being human.
