The experience is familiar by now. The phone is picked up with no particular intention. An application opens to a feed algorithmically ranked by predicted engagement. The thumb scrolls. A video plays, then another, then another. Time passes — ten minutes, thirty minutes, an hour. The phone is put down with no memory of what was watched and no conscious decision about any of it. The algorithm chose every frame, and the choice was accepted without resistance. This is not a failure of willpower. It is a structural feature of the environment that a generation has grown up inside, and its deepest cost is not the hours lost to scrolling. It is the gradual disappearance of the ability to want something independently and pursue it over time, because the algorithm has always provided the next thing before the wanting could form.

Wanting is not a passive state. It is a skill, developed through practice, maintained through use. It begins with noticing a gap between the current state and a desired one. The gap must be held in attention long enough to form an intention — I want to learn this, I want to build that, I want to become someone who can do the other. The intention must survive the inevitable distractions and difficulties that arise during pursuit. It must be revisited, reinforced, and sometimes revised. The process is effortful, and it strengthens the neural pathways that support self-directed action. Every time a person decides what they want and works toward it, the ability to do so again becomes slightly more available. Every time the decision is outsourced, the ability weakens slightly.

Algorithmic recommendation is the most effective outsourcing mechanism ever built. It does not wait for the user to want something. It presents something before the user has finished the previous consumption, filling every gap with predicted content. The gap between one video and the next is measured in milliseconds, which is less time than it takes for the question what do I want now to form. The gap between finishing a post and seeing the next is zero — the next post loads before the user has decided whether to stay or leave. The gap is the space where wanting happens, and the algorithm has been engineered to eliminate it. A person who grows up in this environment does not learn that wanting requires holding a gap open, because the gap is never allowed to exist.

The consequences extend beyond consumption into every domain that requires sustained self-directed effort. Setting a goal requires forming a desire that will not be gratified immediately. Studying for an exam, learning an instrument, building a project, maintaining a relationship — these all demand that the wanting survive periods where the algorithm would have offered a replacement. A person whose wanting has never been exercised beyond the length of a short-form video has no practice in sustaining desire across weeks or months. The difficulty is not that the goals are too ambitious. It is that the mechanism for holding a goal in attention through distraction has never developed. The algorithm filled every distraction with something more immediately gratifying, and the muscle of returning attention to the goal was never built.

Social media extends the same dynamic from consumption into relationships. A person posts content — a photo, a thought, an update — and waits for responses from an audience they cannot see. The responses arrive asynchronously, if they arrive at all. Someone replies hours later, if they remember. The conversation, if it can be called one, consists of alternating performances. Each person broadcasts; neither responds in the moment. The relationship becomes one-directional: I post, you watch; you post, I watch. The bilateral exchange that defines real interaction — I say something, you react now, I react to your reaction — is replaced by a sequence of broadcasts that never quite connect. A generation raised on this pattern learns to perform for approval rather than engage for understanding. The difference is structural. Performance optimizes for the reaction of an imagined audience. Engagement requires the vulnerability of responding to a real person in real time, without the ability to edit or delete. Performance can be perfected. Engagement cannot, and that is the point.

The loss of real interaction compounds the damage to wanting through a social mechanism. Wanting is shaped by observation — seeing someone else pursue something makes the pursuit feel possible. But the relationships that form through algorithmic feeds are not rich enough to transmit this kind of knowledge. A person sees the highlights of another’s life — the vacation photos, the promotion announcements, the curated achievements — but not the daily effort, the setbacks, the boredom, the doubt that accompanied the achievement. The highlights create the impression that success happens suddenly, without the long wanting that actually precedes it. The viewer compares their own uncurated, effort-filled reality with the curated, outcome-only feed of others and concludes that something is wrong with them. The feed suggests that everyone else has already arrived, which makes the wanting of the viewer feel pathological — why am I still struggling when everyone else has already won? The wanting is abandoned not because it was misguided but because it felt lonely.

Alongside the atrophy of desire runs a parallel erosion of objectivity. The algorithm shows more of what the user has already engaged with. A belief, once expressed through a click or a linger, is reinforced by a stream of similar content. A disagreement, encountered in the feed, can be dismissed with a scroll. The user never has to argue for their position, never has to hear the counterargument from someone who holds it sincerely, never has to sit with the discomfort of being wrong. In pre-algorithmic life, disagreement was friction that had to be resolved. A person who held an opinion had to justify it to others, and the act of justification revealed weaknesses in the reasoning. The algorithm removes this friction. Every scroll past disagreement is a missed opportunity to test a belief, and the accumulated effect of years of missed opportunities is a worldview that has never been challenged and therefore cannot be called a worldview at all — it is a set of unexamined preferences, reinforced by selection bias, mistaken for conviction.

The three mechanisms — the atrophy of wanting, the impoverishment of relationships, and the loss of friction — do not operate in isolation. They form an amplification loop that deepens with each cycle. Weaker desire leads to more passive consumption. Passive consumption occupies the time that could have been spent in real interaction. The absence of real interaction removes the social context that supports wanting. Without wanting, the user is more receptive to algorithmic suggestions. Each mechanism feeds the next, and the loop tightens over years. The loop is not detected by the person inside it because the environment feels normal — it is the only environment they have known. The constant availability of content, the absence of silence, the replacement of choice with suggestion — these are not experienced as deficits. They are the baseline. The question of whether something is missing never arises, because the missing thing has never been experienced.

The loop exists because the algorithm is not a neutral tool. It is a revenue engine, and its design is determined by the economics of surveillance advertising. The platforms that host the algorithm do not charge users for access in most cases. The product is free because the user is not the customer — the user is the raw material. Every minute spent on the platform generates data: what the user watches, how long they watch, what they skip, what they like, what they linger on despite not liking, what they search for afterwards, who they follow, who they mute, when they are most receptive, what emotional state correlates with the highest probability of clicking an advertisement. This data is the platform’s inventory. It is sold to advertisers who use it to target the user with products they are statistically likely to buy. The more time the user spends on the platform, the more data is generated, the more precisely the targeting can be calibrated, and the more the platform can charge for access to the user’s attention.

The algorithm is optimized for this business model. Engagement is the metric that correlates with revenue, and the algorithm is designed to maximize engagement by whatever means necessary. Emotional content performs better than neutral content because emotion drives interaction. Outrage performs better than calm because outrage compels a response. Short-form content performs better than long-form because short-form allows more transitions, more opportunities to measure reaction, more granular data points. The recommendation engine does not select content based on what is good for the user, what the user needs, or even what the user consciously wants. It selects content based on what keeps the user scrolling, because scrolling is the activity that generates data, and data is the product being sold. The user is being farmed. The intellectual degradation, the atrophy of desire, the impoverishment of relationships, the loss of objectivity — these are not side effects that the platform tolerates despite their harm. They are the predictable outcomes of a system designed to hold attention by any means, and attention held by any means is worth exactly as much as the data it produces.

The user in this system occupies a dual role that the design of the platform obscures. On one side of the transaction, the user is the source of data — the livestock whose behavior is measured, analysed, and packaged for sale. On the other side, the user is the target of advertising — the consumer whose attention is sold to the highest bidder. The platform occupies the middle, extracting value from both directions. It collects data from the user’s free labor (every scroll, every view, every interaction), processes it into a marketable asset (the targeting profile), and sells access back to the user in the form of advertisements that exploit the profile. The cycle is self-reinforcing. More data produces better targeting. Better targeting produces more effective advertisements. More effective advertisements produce more revenue, which is reinvested into better algorithms, which hold attention longer, which produce more data. The user is both the worker and the customer, and the platform owns both sides of the market.

This is not a conspiracy. It is the structural logic of advertising-funded platforms, and it operates regardless of the intentions of the engineers who build the features. The engineers who designed the infinite scroll, the autoplay, the personalized feed, and the notification system were solving for engagement metrics because the metrics were tied to revenue. The question of whether a person who scrolls for four hours a day would retain the ability to form a goal and pursue it over months was never part of the optimization function. The metrics did not ask that question, and the incentives did not reward asking it. The algorithm is not malevolent — it is indifferent. It optimizes for the signal it is given, and the signal it is given is engagement, because engagement is what the business model monetizes. If the business model monetized user satisfaction or long-term well-being, the algorithm would optimize for those instead. It does not, because the advertising economy does not pay for satisfied users. It pays for attentive ones, and the most reliable way to sustain attention is to keep the user in a state of low-grade wanting that is never quite fulfilled. The algorithm is designed to be addictive not by accident but because addiction is the most effective form of retention, and retention is revenue.

Reclaiming the ability to want is not a matter of deleting applications or switching to a dumb phone. Those actions change the external environment but do not rebuild the internal capacity. Wanting is restored only through practice — through choosing something, holding it in attention, and pursuing it past the point where an algorithm would have offered a replacement. The practice must happen in the same environment that weakened the capacity, because the environment is not going away. The algorithm will continue to whisper, and the user must learn to listen to themselves over the whisper, which means deliberately inserting friction where the algorithm has removed it — closing the feed before the next thing arrives, sitting with the gap, letting the wanting form. The gap is uncomfortable. It was designed to be. The discomfort is the feeling of a capacity being exercised, and the capacity, like any muscle, grows only through the discomfort.

This is not a prescription. It is an observation about what the algorithm has taken and what it would take to reclaim it. The algorithm took not just time but the structure of desire itself. It replaced wanting with receiving, pursuit with consumption, and relationship with performance. The replacement was gradual enough that the change was invisible to the people inside it. A generation that never learned to want did not notice the absence of wanting, because the algorithm always provided something to fill the space. But the space is the point. The space is where the person decides what matters to them. Without it, the algorithm decides. And the algorithm is very good at deciding — it just decides for its own metrics, not for the person whose attention it holds.