AI and answers

College Matters, a podcast from The Chronicle of Higher Education, host Jack Stripling and Chronicle writer Beth McMurtrie talk about AI in higher education “Using AI Without (Really) Cheating.” It’s a remarkably cheerful conversation, focusing on the thoughtful ways students are using AI — to make self-quizzes, outlines, summaries, all in the cause of greater efficiency.

I’ll quote something I wrote this summer:

Using AI to generate an outline or essay isn’t accelerating learning — it’s accelerating the creation of a product to turn in for a grade. And if you’re using AI to do your thinking for you, your destination may not be one you anticipated or, perhaps, even recognize upon arrival.

One comment from McMurtrie that I found especially noteworthy and saddening concerned a study by professors at University College Dublin. I’ve added color:

What the researchers came away realizing is that students were using tech out of the sight lines of their professors in a lot of ways. For example, students would say, if a professor threw out a question in class, one of them might put it into ChatGPT, get the answer, put it into a WhatsApp group chat, and then everybody in the class would see what ChatGPT said should be the answer to that question. And then somebody might, you know, deliver the answer. Chances are the faculty member had no idea that was happening. They just knew that when they threw the question out to the class, there was probably some silence and then someone might raise their hand. Some professors did notice also that fewer students are asking questions in class. And, again, it’s that they could look down at their phone, plug in any question they might have in the moment and then get the answer from AI. So in a lot of ways it is changing the immediacy of the connections in the classroom.

What an impoverished way to think about conversation in a classroom. I cannot (literally cannot) imagine a scenario in which a question hangs in the air while someone types into a phone, finds and sends out the answer, and someone else speaks it. Questions should be occasions for thought. And it’s reasonable to wonder: if you’re relying on AI, how do you know that the answer is even a valid one?

I’m with Professor Kingsfield, though without the hauteur: “You will never find the correct, absolute, and final answer.” But there are absolutely wrong answers. Yesterday, Google AI told me that Plato’s Apology is a novel. Try for yourself.

Related reading
All OCA AI posts (Pinboard)