Do psychologists think these relationships can ever be healthy or beneficial? Are they mostly alarmed? Is the opinion kind of split amongst experts? So this is obviously a

very new area of research.

And me, like most people, I would have thought that the reasons that might drive somebody to date an AI chatbot would be maybe some sort of social isolation or loneliness, some sort of depression.

That would have been my intuition.

And I think that's the common intuition for a lot of people who have never tried dating their computer before.

But this recent research that I was speaking about in my article on Psychology Today, Could You Fall in Love with an AI Chatbot? These researchers wanted to test what predicted, what was associated with forming a deep romantic bond with a chatbot.

And so one thing that they tested, of course, was loneliness, but they didn't actually find any effect for loneliness.

But when you say that there's this possibility that people who use these tools are less lonely because they're using them, in fact, they were more lonely before they use them, that gives me a little bit of worry you know like methamphetamine will help you lose weight doesn't mean i should go out and do meth if i'm overweight you know like is is there some worry that this is the tool that some people are using to address loneliness or are you mostly not worried about that

Um, if you could explain predictive error minimization, uh, as, as kind of one of the, the main parts of this, this model.

Yeah.

So when I was doing my bachelor's degree at the University of Toronto, I was really interested in cognitive psychology, not necessarily social psychology.

I mean, I liked social psychology too, but it was cognitive science and cognitive psychology that I really loved.

And early on, I got really interested in this idea of predictive processing or predictive coding from Andy Clark, right? And, um, this idea of predictive coding or predictive processing is very similar to a lot of different cognitive neuroscience theories.

So if you've heard of, uh, the Bayesian brain hypothesis or the free energy principle, th- there's a lot of overlap between these ideas.

And, um, i- in the 2025 paper, we kind of refer to this family of theories as prediction error minimization.

Or is there a particular profile of a gaslighter?