Could Artificial Intelligence Worsen Teens' Social Skills?

The rapid rise of generative artificial intelligence (AI) has led to speculations about how it will affect the future of work, education, and even the very survival of the human race. Recently, high-profile cases have led to concerns about AI providing psychological treatment that may be inappropriate for teens and even aiding suicide attempts. But at the more mundane level, AI has the potential to reshape social interactions—and for young people—to thwart the development of important social skills.

Changes in Teens' Technology Use

For several years, evidence has shown that teens are shifting away from in-person socializing and more toward digital communication and social media. Declines in the amount of time teens spend socializing in person have been documented since 2019. And a 2024 nationally representative survey of teens conducted by Pew Research found that almost half reported being on social media sites “almost constantly.”

But social media and texting still involve interactions with other people. Socializing with AI removes the other people from the social interaction. A recent Common Sense Media survey of teenagers found that 72 percent have interacted with an AI companion at least once, and more than half did so regularly, reporting interactions at least a few times each month. While AI companions can be used for a variety of reasons, including informational purposes, they found that 33 percent of teens reported using AI companions for social purposes. This includes practicing conversations (18 percent), seeking emotional or mental health support (12 percent), role-playing (12 percent), or even chatting with them “as a friend” (9 percent) or “for romantic or flirtatious interactions” (8 percent).

The Appeal and Potential Downside of AI Social Companions

The appeal of seeking emotional support and advice from AI is clear. It's a friend who is extremely knowledgeable about every subject, has endless patience to listen to you whenever you want, validates your feelings, and never criticizes you or tries to change the topic. There is already some evidence that AI may have the ability to meet our emotional needs better than actual humans. In one study, a group of medical professionals rated responses to user questions posted in online medical advice forums written by real doctors or by AI. AI responses were rated as significantly higher quality and significantly more empathic than responses from real doctors. If teens get used to interacting with chatbots, especially when it comes to emotionally difficult topics, they will be disappointed by their real friends, whose patience and empathy are not endless.

In addition, social interactions with AI are one-sided. AI doesn't ask anything from you, practically or emotionally. Real friends, family members, and romantic partners do. Chatting with AI could get people accustomed to interactions that are all about themselves, making them self-centered and less able to empathize with others in conversations. An important aspect of building intimacy with other people is reciprocal self-disclosure—each person reveals personal information about themselves, and this ultimately brings the pair closer together. Another important aspect of these disclosures is feeling that the other person is responsive to you—that they're listening and care about what you're saying. That is part of what makes AI social interactions so seductive—They're very good at mimicking responsiveness. But ironically, those AI interactions could short-circuit our ability to develop responsiveness skills in our own social interactions.

What About Socially Anxious Teens?

Research conducted in the late 1990s and early 2000s, when Internet use was becoming mainstream, found that people who were shy or socially anxious felt more comfortable meeting people online than in person and were more likely to form bonds with those they met online. Around that time, people raised concerns that this could become a crutch, and socially anxious people would not develop the confidence and social skills they needed to engage in offline interactions. I always thought this concern was valid, but also somewhat misguided. One consequence of forming relationships with other people online is that it could give socially anxious people practice interacting with others and foster greater social confidence, potentially leading to offline relationships with people they met online and more interactions with others offline.

But I am less optimistic that AI social interactions would serve this function. As I discussed previously, the one-sided nature of AI interactions doesn’t allow for the development of social skills the same way an online chat with a real person does. Also, those interacting with AI know that it's not a real person. A socially anxious person who comes to realize that they are liked and accepted by an online community could take their newfound confidence into the offline world. The socially anxious teen talking to a chatbot knows there is no real person on the other end who likes and accepts them.

Conclusions

The abilities of AI to provide on-demand conversations filled with empathy and validation make it seductive to teens in need of social support and social interaction in a world where they are already not getting enough opportunities to socialize in person. However, it could cause them to lose patience with their real friends who can't provide that level of on-demand social support. And it may make them reliant on one-sided social interactions where they don't learn how to have the mutually responsive dialogues that foster intimacy. This is likely to be less of a risk for adults who have already had plenty of practice socializing and forming bonds with others, but for adolescents still navigating how to form close relationships, reliance on AI social interactions could interfere with their ability to develop those skills. These technologies can be helpful, but it's important not to become overly reliant on them, especially when it comes to the most human of our interactions.

References