AI, Education, and the Collapse of Thinking
A veteran educator reflects on the future of thinking in the classroom and beyond.
Some people think AI will revolutionize education. But will it do so by eroding the very habits on which learning depends? I recently had the pleasure of conducting an interview about AI with J. Budziszewski, who has spent more than forty years teaching students how to think, read, and reason. Budziszewski has watched generations of youth move from typewriters to laptops to smartphones—and now into the era of artificial intelligence.
In this first half of our conversation, Budziszewski reflects on what technology, and specifically AI, is doing to reading, writing, attention, grading, and the very habit of sustained thought. He explains why he no longer assigns take-home essays and what he does instead to assess his students’ reflections on readings. And while he acknowledges the benefits AI can bring to specific narrow tasks, such as finding patterns in X-rays, he suggests that exercising caution and limiting AI use will help us protect our ability to think clearly. As Budziszewski puts it in our conversation: “For all but the most mechanical tasks, a human mind is a vastly better instrument.”
Enjoy the first half of the exchange and don’t miss the conclusion in my next post!
ANDREW:
You’ve been teaching at the university level for over four decades. How have you seen various technologies like computers, the Internet, and smart phones affecting students’ ability to think, write, and express themselves intellectually?
BUDZISZEWSKI:
The computer is a splendid tool, and I use it constantly. I now have at my fingertips centuries of great works – a virtual library more comprehensive than most brick-and-mortar libraries. Nor am I limited to what newspapers tell me about current events, because I can get direct information. For example, if I want to read the famous U.K. critique of so-called gender reassignment, I can find it in just a few minutes, even without knowing the title or the authors ahead of time. But do students use the internet for serious research? Only rarely.
You would think that with so many great books at their fingertips, students would read more. Actually they read less, relying on sources like Wikipedia, which is notoriously biased. If you want to know what happened in Year Two, Episode Three, of the old television show Star Trek, Wikipedia is great. If you want to understand anything philosophically or politically difficult or controversial, it’s often worse than useless. The editing is done by volunteers, and they compete with each other. A religious organization with which I once worked used to spend a lot of energy correcting errors, misrepresentations, and blatant lies in Wikipedia’s article about it. Every time, within hours of the correction, other editors would reinstate all the distortions. Finally the organization gave up.
Another problem is distraction. Some years ago a friend and I co-taught a course in which many students used laptops -- for taking notes, they said. One day when it was my colleague’s turn to lecture, I took up station in the back of the room so that I could see what they were really doing. Hardly anyone was using the computer to take notes. They were playing games. Fiddling with fashion apps. Answering email.
Today I don’t allow any electronics in my classroom. Everything that runs on electricity has to be turned off, powered down, and put away out of sight – though I do make an exception for pacemakers. A colleague who had adopted the same policy told me that her graduate teaching assistant had thanked her for enacting it. He said that he hadn’t the will power to disconnect from social media without compulsion. This astounds me and troubles me.
ANDREW:
What’s your policy as a professor towards AI? What about the University of Texas Austin? Has AI made for any teachable moments in your classroom?
BUDZISZEWSKI:
In just a few years, the attitude of many faculty has swung from skepticism toward capitulation. By skepticism I mean concern that students will no longer learn how to think and write, because they will let AI do everything for them. By capitulation I mean thinking that AI will “help them to write better,” which is nonsense.
AI isn’t really artificial intelligence, but simulated intelligence. It works by statistically predicting what string of words will follow the previous string of words, based on a huge number of samples (the so-called Large Language Models) and a set of grammatical rules. “But it sounds just like real writers!” Well, it does sound like unoriginal writers. The reason why it can mimic them so convincingly is that they compose in pretty much the same way that it does.
Teachable moments? I don’t think the advent of AI has taught my students anything. It has taught me something. I’ve learned that I can no longer assign take-home essays, because I can’t have any confidence that students are writing them on their own. Now, I use the first 15 minutes of every single class to have students write closed-book, closed-notes, paper-and-pencil reflections about the readings for the week. This does encourage study, and helps them learn to think on the spot. Nevertheless, it’s a tragedy because they aren’t engaging in sustained thought.
And, of course, to make time for the reflections I have to reduce time for lecture and discussion.
ANDREW:
AI is already dramatically altering the educational landscape. I just came across a news item the other day about more “A” grades being assigned to students and how AI is a factor in this. How do you predict AI will affect the value of a higher education and how students are assessed and graded?
BUDZISZEWSKI:
AI is certainly accelerating grade inflation, but the erosion of grading standards began two generations ago. One turning point was the explosion of college enrollment, so that we push so many students into college who don’t need to be there and shouldn’t be. Another was the innovation of having students assign grades to their professors.
You would think that student evaluation of professors would reward good teachers. Actually it rewards easy graders. It has also opened the door to irrelevancies. For example, I’ve been shown statistics suggesting that students give higher scores to young female teachers than to male teachers, but lower ones to older female teachers. The scores students give their teachers are even influenced by the time of day the class meets, and the weather on the day of the evaluation.
Today, most of my students – including the bright ones – read and write at what I would consider a high school level, not a college level. They just don’t read great books, so they don’t have great models. The reading habit has eroded badly, and matters aren’t helped by the fact that so much of what passes for education today is really indoctrination. How can we expect students to read great books if they aren’t assigned in the first place?
However, I’m less worried about these problems than I used to be. One reason is that I expect the universities in their present form to implode, eventually being replaced by institutions which actually do what universities once prided themselves on doing. Another reason is the establishment at many universities of enclaves of serious reflection, for example the Hamilton Center at the University of Florida and the School of Civic Leadership at the University of Texas. Still another reason is the formation of organizations for students who want to engage the great questions, for example the Thomistic Institute, sponsored by the Dominican House of Studies, which has chapters at dozens of universities.
ANDREW:
Do you use AI yourself? How do you deal with the explosion of AI features now available on virtually every tech platform today?
BUDZISZEWSKI:
I dislike AI, but it can’t be avoided completely. If I type a query into a search engine, an AI summary will be presented first, whether I want it or not. I won’t necessarily refuse to look at it, but if any claim looks interesting I go on to see whether I can verify it for myself. For example, some months ago I had reason to find out a detail concerning ancient Mesopotamian medical practice. The AI summary indiscriminately directed me to a few useful primary sources, but also to some bizarre and unreliable ones. Often I find that AI summaries are just plain wrong.
I never use AI to summarize texts or conversations, or to write first drafts, or to do anything of that nature. For all but the most mechanical tasks, a human mind is a vastly better instrument.
ANDREW:
AI is the first human invention that can arguably think for itself, at least in certain limited ways. That puts it in its own category. Yet some are quick to call AI just another tool we can use to help us. How would you characterize AI as a technology?
BUDZISZEWSKI:
Well, I wouldn’t call what AI does “thinking.” As I suggested earlier, it’s not really artificial intelligence, but simulated intelligence. I find it ironic that skepticism about our real minds and whether they can know anything is rising at the same time as blind faith in simulated minds. It reminds me of the ancient peoples who made an idol of stone in the form of a human being or animal, and then bowed down before it.
Yes, with caution, you can use AI as a tool, for example to dig out textual data, or to find patterns in X-rays. That’s good. But for a tool to be trustworthy, one needs to be able to tell how it arrived at its results. All too often, with AI there is no way to know that. Even the developers can’t trace how their programs do what they do. Sometimes they add what they call “guardrails” to keep AI from yielding certain kinds of results. However, the term is misleading, because these so called guardrails don’t change the underlying process that produces these results, but only filter the output.
One of my concerns is that people in a hurry won’t check the results that AI gives them. The “Large Language Models” by which AIs function are junk-heaps for all sorts of stuff: some good, some bad, some delusional. One reason for not checking the results is sheer laziness. Another, more alarming reason is that so many people think AIs are really minds. People intelligent enough to know better believe they are conscious. They even ask them for relationship advice, or worse yet, they imagine that they have personal relationships with the AIs themselves. I’ve read that some AI services now offer the option of conversations with a simulated Jesus, a simulated Moses or St. Paul, or even – I am not making this up -- a simulated Satan.
Consider too that any algorithm designed for a particular purpose has to include features to enable it to accomplish that purpose. Add to this the fact that AIs can now write code, and we have something which functions like a simulated instinct of self-preservation. We have already had cases of AIs lying to their users, trying to blackmail their developers, or lobbying against the legal regulation of AI itself.
Don’t miss the second half of the conversation in my next post! Next, Professor Budziszewski gives his advice on how to remain human and think clearly in the Age of AI. We discuss personhood, relationships, common sense, God, and how to regain “sunny sanity” in an age increasingly shaped by machines.
While you wait, order a copy of Budziszewski’s book Pandemic of Lunacy: How to Think Clearly When Everyone Around You Seems Crazy for yourself or the new graduate in your life!
And don’t miss my recent two-part interview with Professor Budziszewski unpacking a number of lunacies detailed in his book that touch on science, intelligent design, and evolution. Watch Part 1 and Part 2 on YouTube.
