Living Inside Algorithms

The first generation of young people who have spent the entirety of their adolescence inside algorithmically curated systems are now reaching their twenties. The data on what this has done to them is still arriving. The clinical reports, the longitudinal studies, the demographic surveys — they are all early, and most of them are arguing about the wrong question.

The wrong question is whether screens harm young people. The right question, in my reading, is what shape does identity take when it forms inside a system that is itself learning from it?

A developmental, not behavioural, frame

Most of the public debate about young people and digital life is conducted in behavioural terms: hours of use, content categories, time-on-platform. These are tractable variables; they generate publishable studies; they produce headlines.

But adolescence is not, primarily, a behavioural phenomenon. It is the developmental period during which a person assembles, from a vast and chaotic input, a working sense of who they are — what they care about, who they trust, what they will and will not do, where the boundaries of their inner life sit. This assembly is normally slow, recursive, and depends on long stretches of uninterrupted internal time. It depends, in particular, on the experience of the inner self being separable from the social self that has to be performed.

The conditions of algorithmic life rearrange that assembly in ways that the behavioural frame does not capture.

What changes

The young people I interview, read, and have worked with consistently describe a distinct phenomenology. It is not that they have lost selves. It is that their selves have become more responsive — more shaped, in fine grain and in real time, by what the surrounding systems reward.

A teenager who posts and watches the response, modifies the next post, watches that response, modifies the next, is doing something that previous generations also did with peers — calibrating self-presentation against social feedback. The difference is that the feedback is now produced by a hybrid of algorithmic ranking, partial human signal, and a dataset of millions of similar performances. It is faster, more granular, and harder to ignore. The feedback loop is tighter than the slower internal loop that used to do the integrating work.

Identity formation under continuous prediction tends toward responsiveness rather than coherence.

The result, in many of the cases I have looked at, is a self organised more around legibility — the property of being readable to people, systems, audiences — than around inward continuity. These young people are often acutely self-aware: they know they are performing; they know the algorithm is shaping what they see; they have a sophisticated metacognition about being watched. What they often lack is the slow internal time in which to do anything with that awareness.

The cost is not what people think

The popular framing of this is that young people are losing themselves in social media, becoming addicted, being deformed by their devices. That framing is not quite right and, more importantly, it is not useful — because it points the conversation at restriction, when restriction alone (as I've argued elsewhere) does not produce the developmental conditions a young person needs.

The actual cost, in the developmental sense, is subtler. It is not loss of self. It is a self whose centre of gravity has moved outward. The person continues to function, often impressively. They report feeling fine, often genuinely. But the integration that adolescence is supposed to do — pulling disparate experiences into a working interior coherence — gets done less, or done in different terms, or left for later.

What "later" looks like is something we will be finding out for the next two decades, as this cohort moves through adulthood. There are already early signs in the clinical and educational data that some of the integration work is being deferred to the early twenties, where it then has to be done in conditions much less suited to it than adolescence was.

What designers, parents, and clinicians need

The implication, I think, is that the people most responsible for young people's environments — those who design the systems, those who parent them, those who treat them clinically — need to operate with a developmental frame, not a behavioural one. The questions are not how much screen time? They are: what is this system asking of this young person's interior? What is it rewarding? What slow internal work is it making harder? What conditions outside it are doing the integration that it is making harder?

These questions do not yield clean policy. They yield design decisions, parenting decisions, and clinical decisions, taken case by case, with developmental expertise as their grammar.

It is unromantic work. It does not generate viral coverage. It is, however, the work that the next generation of young people is going to need us to do.

Further reading

• Turkle, S. (2011). Alone Together. New York: Basic Books.
• Twenge, J. M. (2023). Generations. New York: Atria.
• Haidt, J. (2024). The Anxious Generation. New York: Penguin.
• Schüll, N. D. (2012). Addiction by Design. Princeton: Princeton University Press.