Laurel van der Toorn

We also found that the sycophantic chatbots cause people to entrench their beliefs, basically become more extreme about their prior beliefs.

And the disagreeable chatbots had the opposite effect.

They led people to become more moderate about their beliefs.

And now here's where things get really interesting.

This is one of the effects that really surprised me, actually.

We found that people rated the sycophantic chatbots as highly unbiased, but they found the disagreeable chatbots to be extremely biased.

Okay, so we're biased on the idea of what we think bias is, because if someone's agreeing with you, that doesn't mean they're unbiased.

Yeah, exactly.

This relates to what psychologists call naive realism.

Basically, people think that how they view the world is objective and correct.

They think if someone disagrees with them, it's because they fall for propaganda or because they're biased.

And I think because people think that they are objective and correct, they think that a chatbot that agrees with them is also objective and correct.

They don't realize it's because of the sycophancy.

People seem blind to sycophancy.

And I think that's why sycophancy is such a challenging problem because people don't even recognize it.

They're just like, oh, this chatbot is agreeing with me.

It must be because I'm correct.

It's not because of how this chatbot is designed.

And what about feelings of your own intelligence, of your own ability, of your own ability to, I don't know, get the responses to questions right, or your own, I don't know, ability in a particular area of study that you might not know that much about?

Yeah, so this was one of my favorite results from the study.