Laura Vandenbosch is an associate professor at KU Leuven (Belgium) and director of the Media Psychology Lab. The core subject of her research is the relationship between media and well-being, leading to international publications in several fields including developmental psychology, sexology, body image, social relationships and communication theory. 

Over the last 5 years, she worked on the EU-funded European Research Council (ERC) research initiative “Malleability in mediated ideals: A paradigm to understand effects of contemporary media in adolescents’ well-being”, also known as the MIMIc project.

---

One of your studies found that younger adolescents are more likely to use AI companions and experience more positive emotions than older adolescents. What makes younger users particularly susceptible to these interactions?

Laura Vandenbosch: We particularly examined whether AI interactions made them feel negative emotions. Our focus was mainly on whether they already experienced negative feelings, as they generally use AI for positive purposes like gaining information, entertainment, social companionship, and other reasons. 

Sometimes, the AI itself prompts such feelings with standard questions like "How are you?" or inquiries about their mood. We observed that some adolescents did feel negative emotions after conversing with an AI agent. 

It's important to remember that while their reasons for using AI are often positive, the gratification they receive is likely not always positive. This is especially relevant given their developmental stage and the biases present within AI. Even when they aren't feeling negative, gender biases and stereotypes related to appearance can be concerning, especially because AI agents often behave more like humans. 

During entertainment or casual chats, users may not think critically, an issue not only for adolescents but for adults as well. It's difficult to regulate this since people aren't always in the mindset to think critically, making them more vulnerable in those moments. This presents a challenge for the future as AI agents continue to evolve.

That said, you can use AI in a task-supportive way in which it also supports your learning skills. Yet in some ways, it may reduce your motivation to learn in the future. 

Additionally, AI can provide helpful advice in complex social situations, but it often offers one-sided guidance. This can likely boost your self-esteem temporarily, yet over time, it might hinder your ability to build meaningful relationships with other humans, especially since people at that age are still developing many social skills.

---

Snapchat's ‘My AI’ is integrated into a platform where teenagers already interact with friends every day. How does this shape the way they perceive and engage with an AI companion?

Laura Vandenbosch: We know that children use AI, especially ChatGPT is popular; similarly, Snapchat’s AI conversation agent is also used by this age group. Incorporating AI automatically encourages children to converse with these agents. However, direct insights into the nature of teen conversations with these tools are only just emerging. Before deploying AI features on children's social media platforms, it's important to evaluate their ethical and well-being implications, risks, and benefits to ensure they provide advantages and as little as possible harms. 

This issue is part of a broader discussion about rapid technological developments outpacing research. I believe it would be beneficial to have a regulation that requires an independent committee to review a product before it is approved for release to teenagers, rather than relying solely on companies' self-assessment, which is often profit-driven rather than child-centric. With the rapid rate of innovation, parents cannot keep up with verifying whether each new trend is child-appropriate.

---

As AI becomes more conversational and emotionally responsive, could it influence how teenagers develop social skills and build relationships offline?

Laura Vandenbosch: Because of their developmental stage, adolescents have less developed critical-thinking skills. They also are still developing media literacy, and the developed media literacy skills further vary depending on their age, whether they are 12, 15, or 18 years old. At these ages, they are still acquiring skills like emotional intelligence and the ability to think abstractly. Empathy, in particular, helps them recognise social dynamics that may seem unusual. All of these abilities are still in development. 

Additionally, nowadays, adolescents are growing up in a generation where AI is new to everyone. If you use social media today, your parents have already used it before. As a teen today, you're more familiar with social media than a teen was 10 years ago. The same applies to AI; it's new to all of us. Therefore, there are many aspects of AI that none of us probably fully understand.

This makes us somewhat concerned. Therefore, laws like the European Digital Services Act are important to ensure we have tools today to enforce mitigation measures and hold developers more responsible if their technology is exploiting minors. While AI offers substantial advantages, we also face risks, both short-term and long-term. Research makes it clear that these risks highlight the need to educate all AI users so they are better informed about what they are using.

---

Much of the public debate focuses on misinformation and privacy. Do you think we are underestimating the psychological impact of AI companions on adolescents?

Laura Vandenbosch: We're beginning to analyse this, and we notice that, for example, in social relationships, AI tools can provide companionship and reduce loneliness. However, questions arise - as with other media that often offer instant gratification. 

For instance, if I feel lonely today, I talk to my AI agent for companionship, or if I feel insecure about my appearance, it tells me I am beautiful. This quickly addresses immediate discomfort, as they are built to be biased towards comforting you, but the bigger question is whether, in the long run, it can offer advice that genuinely helps improve psychological issues. 

For example, if the AI tells you you're beautiful but then suggests skincare products, it may boost your confidence immediately. Still, it could also lead to a greater long-term focus on appearance. Research shows that frequent focus on appearance, through skincare or conversations, can actually worsen one's focus on their looks over time. So, while it might alleviate instant discomfort, the long-term effects might be different.

---

The AI Act and the Digital Services Act introduce safeguards for young users. From a developmental psychology perspective, are there important protections still missing?

Laura Vandenbosch: Europe is moving in the right direction to make the internet safer and create a playing field that is much more democratic and balanced, less dependent on what that one company thinks we should be using, and with many players involved and perspectives discussed. So, I think with AI we need to look in the same direction.

In addition, I believe the European population is also changing. For example, France is developing its own AI. Such emphasis on own creations relates to the idea that developers' attitudes influence the kind of AI tools that are eventually created. This suggests that focusing on developing Europe's own AI tools and AI tools specifically developed for minors and tested by independent child wellbeing experts could be a promising future direction. Such tools would particularly benefit minors and adults alike, helping users discover their autonomy and access information in a way that doesn't make them lazy or less inclined to learn. 

This approach would enhance overall safety in AI usage. Building your own AI within the EU framework seems like a strategic move.

---

Sometimes regulation can move more slowly than technology and changes in society. So, without waiting for new regulations to catch up, what can schools and parents do today?

Laura Vandenbosch: For parents, it’s very important to talk openly with children about these tools and make it a daily routine, rather than a once-a-year conversation, since children likely spend some time a day on technology. This discussion is a way to show interest in their everyday experiences, and those everyday life experiences, just like your school day, vary a lot. One day at school you played soccer, and you lost the game, and you're sad about that. The next day at school, you play chess, and you find out you didn't understand the rules, and this makes you unhappy.

Social media and AI usage extend beyond just sex, alcohol, or similar topics. They can involve other varied subjects. For example, if a friend shares a picture, you might put it into AI and modify it to get a different result. This raises questions: did you have your friend's permission? What modifications did you request? Did you share the edited image later? 

So, it definitely makes sense to talk about what happens in that part of life, and as tools evolve rapidly, your adult perspective can help identify potential issues, whether social, health, gender-related, or privacy concerns. Children or teenagers often lack the developmental skills to recognise these layers, and you don't need to understand everything behind the technology to navigate these conversations effectively.

Related article