The ministerial commission on teenagers’ use of social media has delivered its report, and one can only agree with its damning assessment of the use of this platform by the majority of teenagers, and of the risks common to social media. But it would be very naive, and very dangerous, to believe that banning the youngest from these networks is enough to make everything go back to how it was before they appeared. For the world has changed since the time when teenagers met after school on industrial wasteland or on building sites awaiting reconstruction, to which access was often still possible, and when they could play football at weekends in streets not yet overrun by cars. Today, the last bit of canal still present in many towns 30 years ago has been covered in concrete, and there are no longer any spaces in which children and teenagers can share common activities, set off on adventures, practise free play and the rules of socialising that have to be invented as soon as a common activity is shared. Even supposing, and this is far from certain, that school playgrounds and covered play areas were quickly made accessible to families on Saturdays and Sundays, that would only very partially solve the problem. For social media are not only a space for dulling the mind, but also for meetings, debates and the comparison of experiences, even if social media algorithms tend to lock users into repetitive practices and “bubbles” in order to control their exchanges and maximise the quantity of personal data harvested from each person.
- Places to meet
Despite the many disillusions and frustrations caused by the use of social media, they continue to benefit from an imaginary of emotional closeness and powerful encounters, capable of producing forms of psychological resonance between two anonymous interlocutors who may know nothing of each other’s physical appearance and daily reality. This imaginary is fed by three characteristics: the feeling of closeness, the existence of an easily found common language and, of course, anonymity.
The feeling of closeness has been part of the imaginary of the Internet since its creation. The network is in fact mostly perceived as a secure and fluid space of opportunities where interactions are simplified. People wanting to discuss problems that concern them intimately can meet there internet users who share their questions, but not only that.
To confide in someone, it is important to be able to imagine that they share the same values as we do. And that comes from the conviction that we have similar backgrounds, that we are interested in the same things, that we have the same lifestyle, the same culture or subculture, and so on. In short, everything that the sociologist Pierre Bourdieu identified as a set of shared practices. This imaginary of the existence of common concerns is fed by a common language, whether spontaneous or simulated. Indeed, in peer groups formed on the Internet around health questions, if one of the participants has the medical terms for the concerns mentioned, it is better for them to avoid using them: shared jargon works better if you want not only to be listened to, but also to be heard.
Finally, anonymity is an essential element of intimate exchanges on the Internet. Although it is often associated with crude attacks and cyberbullying, it also makes it possible, whether in exchanges between two people or several, to protect participants’ intimate exchanges. This makes them less likely to be made to feel guilty, or even mocked, for their questions or the way they phrase them.
The combination of these three characteristics is particularly important for encounters between teenagers within peer groups devoted to particularly intimate exchanges, since they concern health, and in particular sexual health[1]. This approach rests on the fact that at certain stages of life, notably in adolescence, the impact of information given by those close to us is greater than other influences, and anonymity plays an essential role in it.
Yet it is precisely all of these characteristics that artificial intelligence assistants reproduce today.
2. Machines programmed to simulate understanding
First of all, they recreate the same feeling of closeness that can be found in discussion groups. The chatbot creates familiarity with its interlocutors. It exploits the imaginary of a secure, fluid and friendly space for exchange, where interactions are simplified. To do so, it uses smileys, emoticons and emoji (hand, flexed arm, smiling face, etc.), very present notably on GPT5. They are intended to feed the feeling that the chatbot has affective and emotional states, exactly like interlocutors on social media or messaging apps.
Next, the culture of those who chose the databases from which the AIs were trained and programmed is generally that of those who use them. As a result, the user often encounters the illusion of a common language that is important to feeling confident.
Finally, the feeling of anonymity is total. Who could keep a secret better than a chatbot whose maker guarantees us that it does not retain what we confide in it? Indeed, Sam Altman, the founder of OpenAI, is considering having users’ exchanges with their AI assistant covered by medical confidentiality…
3. The risk of a new public health problem
If social media and their anonymity-protected spaces for exchange were to be banned for teenagers, it is to be feared that a large part of these teenagers’ need to be listened to and informed would turn to chatbots, with a major risk of isolation and of adopting false information : whether through errors by these machines, which never say they don’t know and invent answers they don’t have; or through programmed misinformation.
This risk is all the greater because chatbots add a fourth characteristic to the three we have just considered: the ability not only to answer questions put to them anywhere and at any time, but also to sustain exchanges of unlimited duration with the same interlocutor, whereas those with a human always come up against time limits. Yet the endless conversation is a constant of teenage communication. It represents both a form of recognition and of validation. Being listened to proves that what one says is worth hearing, and talking to peers, or to a chatbot, removes the risk of being criticised over a wording that adults, parents or educators, might judge clumsy or ambiguous.
For these four reasons, the main danger of chatbots is to discourage users from seeking to talk with those close to them about their questions, in other words, from forming bonds. And it starts with all the simple questions that many of us are getting used to asking an AI rather than a parent, a partner or a friend. Yet if people do not interact with each other over simple questions, how will they find the courage to bring to their family or friends questions about far more worrying problems, such as a plan to separate, a serious illness affecting themselves or someone close, or the onset of recurring nightmares ? Thus someone who believes they are temporarily easing their loneliness by telling a chatbot about their everyday difficulties risks making it worse in the long term.
That is why the proposals of the ministerial commission on teenagers’ use of social media have only gone half the way. Driving out toxic social media is good. But letting chatbots occupy the time thus freed up would risk taking us out of one public health problem and into another at least as serious, and perhaps more so.
While waiting for official recommendations, we propose that the 3-6-9-12+ association apply the following guidelines :
- No digital tools before age 3. A child needs a human voice, love and limits in order to develop.
- No connected toys before age 6. AI listens, but it does not understand the child, and it risks imposing on them ways of thinking and values that are not those of their parents.
- No Internet or general-purpose chatbots before age 12, but accompanied use is recommended to develop critical thinking and discover how to use AI by learning to doubt and to reason.
- No digital companion before age 18. AIs in fact merely build a double of the user, which is then presented to them as an interlocutor who always agrees with them. What is more, machines simulate emotions but do not build real bonds. Love, friendship and judgement cannot be automated.
ST
[1] These groups are defined as an “educational approach that calls on peers (people of the same age, social background, function, education or experience) to give information and promote types of behaviour and values.” (Haut Conseil à l’Egalité. Rapport relatif à l’éducation à la sexualité : répondre aux attentes des jeunes, construire une société d’égalité femmes hommes. 2016) The European Commission recognises them as an alternative or a complement to traditional health education strategies. They are even said to play a major role in sexual health education.
