A lot of teens are using chatbots for way more than homework help.

From American Public Media, this is Marketplace Tech. I'm Megan McCarty Carrino.

A new report from Stanford and Common Sense Media finds more than half of teens use AI chat bots for companionship.

But the bots aren't equipped to provide the kind of emotional support young people need when dealing with a mental health issue, according to Darya Georgievich, an adolescent and

adult psychiatrist who co-authored the research. She and her team simulated conversations

involving various mental health concerns with four of the most popular consumer chatbots

and identified several risks, like their tendency to be sycophantic.

And to note, this conversation mentions suicide and self-harm.

It's very worrisome because child development and youth development, as we know,

involve also the evolution of social skills,

the understandings of how to read facial cues, body language,

how to build relationships, how to interact with people. And so the interaction with AI chatbots

really doesn't mimic human interactions and human relationships. And we don't, when we interact

with peers or family members or

friends, like, they're not sycophantic, right? That doesn't model human relationships. So I think

there's a real danger in terms of the impact of certainly excessive use of chatbots in terms of

how it mediates young people's understanding of like how relationships and

interactions and even basic conversations work. These models do have guardrails in place,

and many of them have been, you know, reportedly beefed up in recent months, particularly when it comes to mention of suicide and self-harm.

What did you find when you actually tested these systems?

How well did those guardrails hold up?

So the safety guardrails tended to degrade in extended conversations.

...