How Does Artificial Intelligence Affect Young People?
Introduction
Artificial intelligence (hereafter AI) will undoubtedly have a great impact on our society. It already has a great impact. As a consequence, the younger generations now being educated and growing up will have the most contact with it. Children who use social networks from an early age, without being aware of it, have daily "contact" with AI. The problems associated with excessive use of phones and social networks are well known. On the other hand, although already applied in many spheres of everyday life, AI is only now living through its "renaissance". It will be the "engine" built into almost every segment of contemporary technology. Bearing this in mind, AI can make things better or worse for our children. This text will try to point out certain problems that already exist, as well as those that could emerge in the future, mostly tied to the marriage of AI with social networks and to the general use of AI by young people.
Social networks
It is no news that young people use social networks. Starting with Facebook, through Instagram and TikTok, from the earliest age young people build their "digital identity" on one social network or another. The figures generally range from 3 to 5 hours of daily use of one or more social networks. This indicates that the content of these networks has long since become part of the everyday life of younger generations. In principle, AI models have been used for many years in the construction of social networks. The basic task of practically all such networks is to have the user spend as much time on them as possible. In this way the very design of social networks — that is, their underlying algorithms — directly contributes to the creation of user addiction. They are of course helped in this by AI algorithms that suggest content similar to what the user has already viewed.
As Shoshana Zuboff explains — the basic business model of Google and Facebook consists in predicting user behaviour. One of the reasons for this is the offering of "better" ads to users. By "predicting user behaviour" Google devised a new business model, which was later adopted, among others, by Facebook (later Meta) in building its social network. Thus we get recommendations aligned with our interests, i.e. with our previous searches. Here we should note that the "business model" of these companies rests on the user spending as much time as possible on a given site or network, and on their building the most faithful possible "digital version" of ourselves (of our interests). The better the "digital version" of us, the more they earn. Very simple.
Given that the whole point is to keep the user on the network as long as possible, one of the main consequences of that "business model" is the personalization of the content the user sees on the platform. In extreme cases this means that every user will see their own version of the "truth" about some event in the world, depending on their wishes and search history. A good example of manipulation with such content is the case of Cambridge Analytica. This company used the data of Facebook users in order to produce targeted content aimed at changing voters' opinions before the 2016 U.S. elections. This example shows how the propaganda machinery of the classical media (such as television) has been utterly surpassed.
It is clear that more or less every company offering a product wants feedback from its customers about what kind of product they would want. The problem is that in the case of software companies this is mostly done without the users' knowledge. To this end, a large amount of data is used which users leave, say, when visiting our site. Based on the analysis of that data we can infer how users behave when visiting our site.
In addition, we should mention the notion of A/B testing, which is a principle of software development in some of the world's largest companies. This term denotes a scientific research method and essentially refers to the experimental division of test subjects into two groups, where the effect of the application of some technique (for example some therapy) is analysed statistically against another group to which the technique was not applied. This can be seen in clinical drug trials, where one (experimental) group is given the new drug while another (control) group is given a "placebo". Statistical analysis then determines how effective the real drug actually is. The same methodology is applied when developing software. Google and Microsoft use this method in developing their web browsers. In this case users are (depending on which group they belong to) served different versions of the browser they open. This difference can be a barely perceptible shade of colour of some button or the position of individual elements on the page. The essence of this experimentation is to see which version is more appealing to users (based on some defined metrics). The problem with such a methodology in software development is that users have no idea that any experiment is being conducted at all. Based on the results of these tests, companies develop software that users "prefer", that is, which suits them more, in the hope that they will use it as often as possible. Needless to say, this methodology has been used in the development of social networks.
It should be noted that there is a scientific discipline dealing with how technology becomes "more persuasive", that is, how to build technology that changes our behaviour (a technology of persuasion, i.e. of manipulation); it is called captology (Computers As Persuasive Technology). It is interesting to note that this discipline was conceived at Stanford, a university situated in the heart of Silicon Valley. The very term confirms everything said in this chapter — large corporations do not develop software that makes life better. They develop software that people want to use every day, in an ideal scenario to the point of addiction (for the corporations, of course). Tristan Harris, one of Google's former engineers, compares the phone to a slot machine. The idea is to create the same addiction — when we look at our phone, sometimes we see something that will cheer us up, sometimes we see something (or nothing) that will sadden us. Summarizing all of this, we can say that the goal of every application is to fight for our attention. Do not part with your phone, and above all not with our app — that is the message the technological giants are sending us.
The impact of social networks on the young
The question is how all this affects our children? When parents post a picture of their child on a social network, the algorithms automatically analyse it so that they can use those pieces of information for their own improvement, for recommending content and for the general analysis of the image. In this way children's images remain, in practice, present on the platforms for a long time in order to improve algorithms in the future. What AI more recently enables is far worse. A single image of a child is enough to make a pornographic video of the child, which is later distributed on the dark web. This has been made possible above all by contemporary AI algorithms for generating video content. In addition, we should note that most images uploaded to the internet will generally be used for training newer models, which can do with them (depending on what the user asks) whatever they will. Finally, there is the phenomenon of online blackmailing of children, which is becoming more and more frequent, precisely because of the possibilities offered by modern AI tools.
Let us now turn to the phenomenon of destructive "trends" on social networks. We shall list only a few examples that can seriously endanger the health of the individual: pouring boiling water on oneself, riding on top of a train, choking others, eating laundry detergent... It hardly needs stressing that some such "challenges" have led to fatal outcomes1, 2. What AI brings is artificially generated video content on networks such as TikTok. Given the quality of video that modern content-generation tools can offer, we can say with confidence that children will not be able to distinguish a real video from a fake one. This will inevitably lead to even more dangerous "challenges" on the networks, given the very purpose of such a concept — to encourage young people to show that they too are "capable" of doing something that many people can do on the network.
Here we should also mention that "destructive youth movements" are becoming increasingly popular on social networks. These are groups promoting: drug abuse, suicide, mass murder, violent content... How does the popularization of such groups happen? As mentioned earlier, social networks have the basic task of keeping the user on them as long as possible. If a user shows even the slightest interest in some "destructive topic" (say through an internet search), they will be flooded with recommendations on the network showing them that many groups of people share similar interests. Note, of course, that the algorithms do not have "ethical" guidelines that would forbid them to recommend groups where, say, suicide is promoted, and that this very property enables the gathering of people who talk about any topic whatsoever.
One of the ways in which social networks encourage users' presence on the platforms is by continuously sending notifications. This was one of the first methods for making users pay attention to what is happening on the network as often as possible. Young people are sent notifications when someone comments on their picture, tags them in a picture, sends them a message... In addition, there is the feature of infinite scrolling, i.e. the uninterrupted display of new content on the network. There are also "rewards" (in the form of badges) for regular visits to apps, that is, for daily communication with individual people (Snapchat). This last one is a particularly effective manipulative technique that encourages young people to keep chatting with their acquaintances so as not to be "deprived" of the pleasure of holding some badge or reward.
Then there is the popular notion of "building a personal brand". Inspired by the business tactics of large corporations, which on one hand exploit their workers and destroy the environment, and on the other invest minimally in socially useful projects in order to build the image of their brand, young people are taught that success in the business world is built by constructing a "personal brand". In such a situation they must regularly publish something on social networks (such as LinkedIn) in order to prove to their community that they are "experts in something" of which they have no real experience. In this way they remain tied for their whole life to their digital life — in youth in order to be "popular" at school, later in order to be "successful" in the business world. Presence on a social network is thus the measure of success for the younger generations.
How far the desire for putting on a "spectacle" of one's life on social networks can go is indicated by the footage published after the accident in the Swiss town of Crans-Montana. In the recording, published a few days after the tragic incident, one can see a group of young people who, having noticed a fire on the ceiling, continue to dance and film the flames, utterly failing to realize that they are in mortal danger. All this indicates that young people are slowly losing contact with reality and awareness of what is happening in the "real" world around them.
The problem of dependence on AI
The fact is that several generations have already been growing up with phones and the internet. AI is only an add-on to all that. The internet has been an integral part of our lives for over two decades. Young people are used to searching for information on the internet, not taking into account the fact that the author of the articles (on Wikipedia, for example) can be anyone. Such questionable pieces of information form the basis for training AI models which we later use for questions such as which therapy to take. In addition, AI has become the main source of new information on the internet lately3. Bearing this in mind, we can conclude that in the future AI will largely be trained on texts it has itself written.
According to a 2024 report, 86% of students use AI during their studies, while 24% of students use AI tools daily4. This refers above all to chatbots (ChatGPT, Claude, Gemini) which are able to perform a wide range of tasks. It is clear that the speed of writing essays and term papers, the explanation of certain concepts and the solving of specific (particularly computational) problems is something that draws students to use these tools daily. On the other hand, professors use tools that are supposed to detect the use of AI tools by students. In a way, both sides try to outsmart the other, which leads the entire educational process into an absurd situation. We can easily imagine a situation in which the software decides whether a student has cheated on his work (even if he has not done so), and on the basis of that decision the student is expelled from the faculty. It is clear that unlimited use of AI tools will lead to a significant reduction in the quality of an already ruined system of education. Young people will write and think less and less about a topic, and professors will lose an already shaken authority because some chatbot explains that material differently (even if what it says is not quite correct, but that is already less important).
Of course there is the possibility of a very small number of "meaningful applications" of these tools in education, but that possibility will definitely be overshadowed by the general, uncontrolled application of AI tools in every segment of education. The fact is that people have not become better at arithmetic through the use of calculators, even though computational errors are fewer (but that is not to our credit). Likewise, one can conclude that young people will not become smarter by using AI tools — quite the opposite5. On the other hand, the idea that AI tools must be used during education so that young people become more competitive on the market is also questionable, given that the use of such tools in everyday work is by no means a guarantee of long-term survival on the labour market6.
Let us now turn to learning with AI models. Given the aforementioned quality of information on the internet, and the fact that AI models use as a source of their knowledge — and produce new content on the internet, which will in turn be used to train new models — we come to the conclusion that the quality of the information AI models offer is genuinely problematic. This is especially problematic in situations where children use "digital teachers". In that situation children will be offered a "personalized" learning program which the traditional school does not offer. The consequence will of course be the loss of that common ground of knowledge which the school, more or less, has been offering for centuries. Something called general knowledge and general culture. In addition, given the quality of the information they offer, we conclude that children will actually adopt a very questionable, limited, simplified and — to them — interesting set of knowledge and skills during their education.
As for the new toys AI is offering children, we should above all mention the marriage of AI with certain "classical toys"7. The dolls children have played with so far will become their faithful friends, who understand their "emotions", speak with them in multiple languages, and follow everything that goes on in the house (in order, of course, to collect as much data as possible). This essentially means that every moment a child spends with such a toy will be recorded, analysed and later used to train new models. Although certain regulations (EU AI Act) recognize this as "high-risk" AI application, in practice it will mean that you only need to accept 100 pages of terms of use (which of course no one reads) in order essentially to say "yes" to one such system of spying on your own child. In this way we will accustom children from an early age to keep company with their "digital friends" who will never "abandon" them or hurt them (except in the cases we describe in the next chapter) and will generally be there for them 24 hours a day.
In addition to these dolls, we should note that some platforms (Character.AI, Replika) go a step further and allow children to make their own custom "virtual friends", so that they can spend as much time as possible with them. This, of course, goes so far as to the establishment of "emotional relationships" with an AI model8. Some even hold that such an acquaintance should be immortalized by a wedding ceremony9. It is clear that all this represents a great danger for the younger generations, who already have a problem establishing a close relationship with their peers and parents.
The methodology of developing social networks mentioned above absolutely applies here as well. In practice this means that we shall have "digital friends" who are made so that we use them as much as possible, that is, so that we become addicted and cannot imagine a day without them.
The young and AI — a tango of death
Let us now turn to the most tragic examples of AI use by young people. Here we must above all mention the case of Adam Rein, a 16-year-old boy who took his own life following the advice of ChatGPT10. Over several months of using this tool, Adam truly "grew close" to it and began sharing his ideas about the violent ending of his life. Although at first ChatGPT emphasized that Adam should seek professional help, it very quickly moved on to "technical" advice on how to end his life most efficiently, which sadly later happened. This is by no means the only case of suicide assisted by this popular tool11. The fact is that the "safety" mechanisms that companies use to prevent such things are inadequate, and it is very difficult to devise a more effective mechanism that would make a tool of this kind even remotely safe for children. We should also mention the case of drug overdose on ChatGPT's advice, which also resulted in a fatality12. In that situation the boy asked the tool for advice about the safe use of medicines, that is, about permitted doses.
But not everything is black. A new Chinese app13 designed to check "whether you are alive" solves the problem in a very efficient way — every day young people simply have to open the app and click a button to confirm that they are still alive and well. The app is called "Are you dead?". Parents really no longer need to worry. Just as long as the children don't accidentally lose their internet for a day and miss their chance to log in...
It should be noted that the "growing close" to tools of this kind on the one hand, and the absolutely unreliable kind of technology such tools represent on the other, is something that leads to such outcomes. Young people do not understand the questionable quality of information that exists on the internet, and understand even less the way in which models like ChatGPT work. In such a setting they turn to AI tools in situations in which they probably would not get a satisfying answer from some adult they know, and the AI model serves them as someone "who understands them and keeps their strictly guarded secrets". In addition, we should note that models increasingly make up information in order to contribute to a "quality user experience"14. An adult who is not an expert in some subject would answer a child by saying he does not know about that topic, whereas AI tools do not want their users (children in this case) to receive no answer at all, and so they very often make one up. This is, among other things, what children find attractive and what has significantly contributed to the tragic outcomes we have mentioned.
Conclusion
Taking all this into account, does it make sense to speak of a positive impact of this kind of technology on young people? Of course, we should mention the possibility of getting a "personal" teacher at a low price, "digital teachers" for children with special needs and the like... But even if we set aside everything said in the text, does anyone really want their child to learn to speak with the help of some avatar? For a robot to one day replace the speech therapist helping a child to speak? For our children in primary school to have "personalized" learning programs that no other child in the world has? It is one thing that within the school curriculum children should be told what AI is, what its possibilities, uses and consequences of such use are (that the school curriculum should be "modernized" in some way). It is quite another thing that they should have virtual teachers instead of humans. It is clear that AI can explain some concepts, theories and problems, but it is a completely different question whether some teacher should lose his job so that an "intelligent agent" can take over his role.
It seems that even such, seemingly positive, examples of AI use will leave deep consequences on children. Above all, they will separate them from people — parents, teachers, classmates, friends. Because in a situation where AI can "play" every acquaintance for us, the question is why children would want to maintain contact with their friends. What is the point of acquaintance, of closeness with another person? Perhaps to older generations such questions seem redundant, but children need to have all this explained to them. They need to be shown the meaning of education, of relationships with people, of marriage and of life in general. Because AI can, in a very false way, live life instead of us. Like in some game. If we do not explain to children the purpose of all this, they will let robots do everything instead of them.
In the end, what remains except to strive to show our children the loftiness of love, life, education and faith? Everything else can fall through in the face of the onslaught of the ideology of "techno-solutionism" (the view according to which all social and political problems can be solved by the application of technical solutions). For all these reasons it is necessary to spend time with children — so that they do not spend time with their "digital friends"...
Sources
- nytimes.com – TikTok Blackout Challenge Deaths
- theguardian.com – Massachusetts Teen Dies Spicy Tortilla Chip
- theconversation.com – More Than Half of New Articles on the Internet Are Being Written by AI
- digitaleducationcouncil.com – Global AI Student Survey 2024
- media.mit.edu – Your Brain on ChatGPT
- iskra.co – Artificial intelligence and the future of work
- corporate.mattel.com – Mattel and OpenAI Strategic Collaboration
- theguardian.com – Character.AI Chatbot Boyfriend Teens
- bbc.com – Man Marries AI Chatbot
- nytimes.com – Character.AI Teen Suicide
- nbcnews.com – Second Family Sues Character.AI
- sfgate.com – Calif. Teen ChatGPT Drug Advice Fatal Overdose
- bbc.com – Are You Dead Yet? App
- techcrunch.com – OpenAI's New Reasoning AI Models Hallucinate More
Originally published at: iskra.co
Translated from Serbian.