Joshua Lambert

I created that.

And I have a student who works with me, too, to help answer those questions.

And so what we did was we gave them, they would ask their questions, and then we would give them three responses back, one from me, one from my student, and then one from the chat bot.

We blinded them from it, so they didn't know which was which.

And then they had to basically rate them and then also ask or answer the question, which one was the chatbot?

So like you said, the results were they preferred it.

But one of the most interesting things that we found in the data was not only did they prefer it, but the one that they rated the worst, they consistently guessed was the chatbot.

Yeah, exactly.

So, I mean, ideally, the way that this would work for them is we could actually just set this up on a website where they could go to at any point, ask their questions, get some feedback.

And, you know, maybe that's sufficient for them.

If not, we're always here to help them, obviously.

Not only happy, all of them said they preferred

the chatbot's answers over my own and my students' answers.

That must not have made you feel amazing about yourself, side note.

Yeah, it was pretty humbling, right?

Well, the funny thing is that they never actually found out.

So they got the answers.

I had them fill out the response.

But we never actually revealed to them what the truth was.

So when they read this paper or watch our interview, they're going to find out for the first time that, oh, they guessed that the chat bot was the best answer.