Algorithmic feeds and adolescent identity formation

A peer group optimizes for cohesion across time. A recommendation system optimizes for engagement within a session. These are not the same objective and they do not produce the same distribution of selves. Cohesion penalizes the personas that break groups — too extreme, too brittle, too demanding of others. Engagement rewards the personas that hold attention — emotionally charged, visually distinctive, narratively sharp. When identity formation moves from the first feedback system to the second, the population of teenage selves shifts toward whatever the engagement metric favors. This is not a claim about any individual teenager. It is a claim about what the curve looks like when millions of teenagers spend years receiving their identity feedback from a system that was tuned for a different purpose.

Performing for twenty known peers and performing for an unknown audience of thousands are different psychological tasks. Known audiences allow nuance because the audience can fill in context. Unknown audiences demand legibility because nuance does not survive context collapse. The pressure on the adolescent self under unknown-audience conditions is to become summarizable — a bio, a flag, a set of labels that can be parsed in three seconds by someone who does not know you. This is not how identity worked before. It is closer to how brands work. A generation that learns to construct itself this way carries the brand logic into adulthood, into relationships, into work.

Adolescence used to come with an implicit clemency. The dumb thing you did at fifteen was witnessed by people who would forget it by twenty-five. The dumb thing posted at fifteen is indexed and searchable at twenty-five. This raises the cost of experimentation and lowers the rate of productive failure. Teenagers respond rationally: they experiment less in public, they commit harder to publicly declared identities to avoid the cost of recanting, and they outsource the riskiest experimentation to anonymous accounts that operate in a different psychological register entirely. The split self — public brand plus anonymous shadow — becomes the modal arrangement.

Jean Twenge's work on generational data shows mental health indicators for adolescents diverging from earlier cohorts starting around 2012, with the divergence largest for girls and largest for measures related to anxiety, depression, and self-harm. The timing aligns with the smartphone-saturation point and the shift to algorithmic feeds. The causal claim is contested — correlation, confounders, measurement changes — but the cohort effect is robust enough that the burden has shifted: those arguing nothing is happening have to explain the curves, and the explanations that don't invoke the new media environment have not done well.

Haidt's The Anxious Generation organizes the evidence into a four-part claim: a great rewiring happened roughly 2010–2015, it disproportionately affected adolescents, it shows up across multiple countries with similar timing, and the mechanisms are identifiable — sleep displacement, social comparison, attention fragmentation, replacement of embodied play. The book is a call to collective action, not individual coping. Its four norms — no smartphones before high school, no social media before sixteen, phone-free schools, more independence and unstructured play — are explicitly collective because Haidt's diagnosis is that individual opt-out does not work when the surrounding cohort is opted in.

Jia Tolentino's essay "The I in the Internet" describes how the architecture of social media trains a particular kind of self — one that experiences its own life partly as content, that runs ongoing background calculations about how moments would read if posted, that loses access to the unposted version of experience. Adolescents who form identity inside this architecture do not have a pre-internet self to fall back on. The optimized self is the only self they have. This is not pathology; it is adaptation to the environment. The question is what kind of generation that adaptation produces.

Devorah Heitner's work on raising children with technology emphasizes that the developmental tasks of adolescence — separation, identity work, intimacy — have specific requirements that platform architectures often fail to meet. Adolescents need spaces where they can fail without permanent record. They need feedback from people who know them, not metrics. They need boredom, which is the engine of inward turning. Platforms designed to eliminate boredom and maximize engagement are not neutral environments for adolescents; they are environments tuned against the developmental work.

Catherine Steiner-Adair's research in The Big Disconnect documented how children experience parental distraction by devices — and, reciprocally, how parents experience their children's distraction. The collective effect is a household-scale loss of mutual attention that compounds over years. Identity formation depends on being seen, and being seen depends on someone being present enough to see. When the ambient condition of family life is partial attention on both sides, the developmental substrate for identity work erodes in ways that no single conversation can repair.

Algorithmic feeds do not just amplify existing subcultures; they generate pipelines into them. A few minutes of engagement with content adjacent to a subculture is enough to bias subsequent recommendations toward deeper content in that subculture. For adolescents whose identity is in flux, this produces a sorting mechanism that accelerates affiliation and makes exit harder. The phenomenon shows up across politics, gender, body image, mental health categories, and aesthetic tribes. It is not a conspiracy; it is what engagement optimization does when applied to identity-adjacent content.

A family that delays their child's smartphone until sixteen is making a rational choice that is undermined by the fact that their child's peers got phones at eleven. The social cost of being the one without — exclusion from group chats, missed coordination, status loss — falls entirely on the opting-out child. This is the classic collective action problem: each family acting alone faces a worse outcome than all families acting together. Wait Until 8th and similar pacts exist precisely to solve this by coordinating the opt-out across a cohort, lowering the per-family cost.

Real change at the collective scale runs through platform defaults, age verification, regulatory pressure, and school policy. Defaults matter because adolescents do not change defaults. Age verification matters because self-attestation has failed. Regulation matters because platforms have not voluntarily designed for adolescent welfare. School policy matters because the school day is the largest concentrated block of adolescent time. None of these levers belong to parents individually; all of them respond to organized parent pressure, which is why the parenthood-lens collective scale on this concept is fundamentally about political activity, not domestic management.

Three things scale: join the pacts (Wait Until 8th, local equivalents) so the social cost of delay is distributed; push school boards on bell-to-bell phone-free policies because schools are leverage points; vote and lobby on platform regulation because that is where the architecture lives. At home, the work is smaller and more boring: keep phones out of bedrooms, keep the family meal phone-free, model the behavior you want, and accept that you will lose some battles. The collective work is what changes the curve; the household work is what keeps your particular child within reach while the curve changes.

Abeles, Vicki. Beyond Measure: Rescuing an Overscheduled, Overtested, Underestimated Generation. New York: Simon & Schuster, 2015.

Haidt, Jonathan. The Anxious Generation: How the Great Rewiring of Childhood Is Causing an Epidemic of Mental Illness. New York: Penguin Press, 2024.

Haidt, Jonathan, and Jean M. Twenge. "Social Media Use and Mental Health: A Review." Unpublished manuscript, New York University, ongoing collaborative review, accessed 2024.

Heitner, Devorah. Growing Up in Public: Coming of Age in a Digital World. New York: Tarcher Perigee, 2023.

Heitner, Devorah. Screenwise: Helping Kids Thrive (and Survive) in Their Digital World. New York: Routledge, 2016.

Steiner-Adair, Catherine, with Teresa H. Barker. The Big Disconnect: Protecting Childhood and Family Relationships in the Digital Age. New York: Harper, 2013.

Tolentino, Jia. "The I in the Internet." In Trick Mirror: Reflections on Self-Delusion, 3–34. New York: Random House, 2019.

Twenge, Jean M. iGen: Why Today's Super-Connected Kids Are Growing Up Less Rebellious, More Tolerant, Less Happy — and Completely Unprepared for Adulthood. New York: Atria Books, 2017.

Twenge, Jean M. Generations: The Real Differences Between Gen Z, Millennials, Gen X, Boomers, and Silents — and What They Mean for America's Future. New York: Atria Books, 2023.

Twenge, Jean M., Jonathan Haidt, Andrew B. Blake, Cooper McAllister, Hannah Lemon, and Astrid Le Roy. "Worldwide Increases in Adolescent Loneliness." Journal of Adolescence 93 (December 2021): 257–69.

Shannon, Brooke. "Why We Started Wait Until 8th." Wait Until 8th, accessed 2024. https://www.waituntil8th.org.

Orben, Amy, and Andrew K. Przybylski. "The Association between Adolescent Well-Being and Digital Technology Use." Nature Human Behaviour 3, no. 2 (2019): 173–82.