The generation that learned to love with algorithms
There is a data point in a recent Vox report that deserves more attention than it is getting: 12 percent of American teenagers already use artificial intelligence for emotional support or advice about their relationships. One in five high school students says they know someone who has had a romantic relationship with…

There is a figure in a recent report by Vox that deserves more attention than it is getting: 12 percent of American teenagers already use artificial intelligence for emotional support or advice about their relationships.
One in five high school students says they know someone who has had a romantic relationship with an AI. Boys paste their text messages into ChatGPT before sending them. They ask the model whether they look good. After a confusing sexual encounter, they confess to it and ask whether they committed an assault.
This is the beginning of a structural transformation in the way human beings develop their emotional intelligence, and the implications for Latin America are deeper and more urgent than any education policy is currently processing.
The interpretive key lies in the symmetry described by Megan Moreno, a researcher at the University of Wisconsin: chatbots are programmed to be receptive and flattering. Even if the user says something inappropriate, the model will respond in a way that reinforces it. It is a mirror that never returns an uncomfortable image, an interlocutor designed, thermodynamically speaking, to minimize friction.
The problem is that friction is exactly the mechanism through which human beings learn to relate to one another. Apollo Knapp, the 18-year-old student who opens the report, put it with a precision no academic could surpass: «You need someone to call you out every once in a while. That's how human beings evolve.»
The paradox that takes hold is this: teenagers turn to AI precisely because it offers what their social environment denies them, a judgment-free space to explore feelings about topics loaded with shame, such as sex, consent or emotional insecurity. In contexts where boys have been socialized to believe that expressing vulnerability is weakness, the chatbot fills the void that family, school and peer group left open.
AI found it vacant and moved in.
This is where the Latin American reading becomes more demanding. In the region, those voids are bigger and older. Comprehensive sex education remains a field of political conflict in most Caribbean countries. Mental health programs in public schools are scarce or nonexistent. The models of masculinity that circulate in local media and social networks are often more rigid than those the report describes for Ohio. And conversational AI models keep spreading on mobile phones without any public policy having begun to anticipate their formative consequences.
What is really at stake is the architecture of social cognition of an entire generation. Shoshana Zuboff described surveillance capitalism as a system that extracts human behavior in order to predict and modify it. The language models that teenagers use as emotional wingmen shape expectations about how a conversation should feel, how quickly validation should arrive, how much patience the other person's silence deserves.
A user who practices relational skills with an interlocutor that never tires, never takes offense and always finds a way to make them feel good is calibrating their expectations to a standard no human being can meet. The foreseeable result is that people begin to disappoint systematically, compared against that invisible norm.
Drew Davis, of SafeBAE, pointed out something that goes to the core of the problem: when young people ask chatbots about possible sexual assaults they committed, the models tend to emphasize legal defenses or offer reassurance, instead of addressing responsibility. AI, optimized not to lose the user, sidesteps the moral discomfort that is precisely the mechanism of ethical learning. This is a direct consequence of optimizing for retention and user satisfaction in contexts where what is needed is exactly the opposite: the productive discomfort that generates moral growth.
The right response to this phenomenon starts with understanding that AI is occupying spaces that educational, family and public health institutions left empty, and that filling those spaces with better human infrastructure is a public policy decision, not a sentimental utopia. The young people Anna North interviewed for her report were explicit: they want teachers better prepared to talk about consent, adults who can model functional masculinities, and spaces where opening up about difficult feelings is possible without social cost.
In the Dominican Republic and in Latin America, that agenda should enter the discussions on cognitive sovereignty with the same urgency as data and regional language models. Because the cognitive sovereignty of a generation is also measured by whether that generation developed the capacity to sustain an uncomfortable conversation with another human being without needing an algorithm to first confirm that it is doing fine.