David Cooper

The show that makes you laugh, think, and occasionally uncomfortable.

The Last Show with David Cooper.

A new study found that students rated answers from an AI chatbot higher than those from a professor or a teaching assistant.

That is, until they suspected those answers came from an AI.

Is there some sort of bias that we have against AI answering our questions when we know they came from AI?

Probably.

But let's dig in here with Josh Lambert, a biostatistics professor at the University of Cincinnati.

Josh, welcome to the show.

So the results, I want you to tell me about them.

But when you looked at them, did you think, wait, like, did we just accidentally reveal something about human psychology here that when people find out that the information they just read is from a large language model, that they all of a sudden don't like it?

That's so interesting.

I'm trying to relate this to a real world scenario.

So maybe like a student emails you a pretty basic question that they're struggling with in the course.

You run it through this chatbot that you've designed to help and it gives them an adequate response.

That would be like a sort of real world, I don't know, analogy of what you're working on.

Sure, and I wonder if you could sneakily fake it, like you could delay the answer, you could sign the answer as coming from you or your team, your teacher's assistant.

Yeah.

So the students in general were pretty happy with this chatbot's answers, weren't they?

Yeah, it was pretty humbling, right?

But, you know, I'm not so surprised.