Backstage Collapsed: Universal Recording and the Architecture of Courtship

A panelist on a recent broadcast conversation made the following argument. Young people across the wealthy world are not having children. Before they do not have children, they do not date. Before they do not date, they do not interact at the dances, the parties, the mixers their parents and grandparents used as the primary infrastructure for finding mates. Even when they show up at such gatherings, they hold the wall, not approaching, not asking, not risking the awkward overture that has been the entry-cost of human pairing for as long as human pairing has been formalized into ritual occasions. The panelist asked why, and answered himself. They are afraid of being recorded. They are afraid that any silly thing they say or any failed dance step or any drunk confession will be filmed and uploaded and used against them by people they cannot identify in advance. So they withdraw. The species, the panelist concluded, cannot continue under such conditions, and the only available remedy is to restrict the technology that produced those conditions.

The empirical observations are mostly correct. His explanation is wrong, his framing is suspect, his policy is impossible. And yet, beneath all that, there is a real diagnosis worth saving.

The empirical layer

Total fertility rates have fallen across most wealthy countries to levels well below the 2.1 replacement threshold demographers use to mark steady-state population. South Korea recorded a TFR of 0.72 in 2023, the lowest sustained fertility rate ever measured in a peacetime country. Italy, Japan, Spain, and large parts of Eastern Europe sit between 1.2 and 1.4. The United States hovers around 1.6, having declined more or less continuously since the 2008 financial crisis. The pattern is widespread, persistent, and resistant to short-term policy intervention. Hungary’s pronatalist subsidies, Singapore’s matchmaking initiatives, and various French family-allowance schemes have produced modest temporary effects without reversing the larger trend.

Adolescent and young-adult sexual activity, romantic partnering, and physical-presence socialization have also declined since approximately 2010. Jean Twenge’s analysis of the General Social Survey, the Monitoring the Future longitudinal study, and similar datasets in iGen (2017) and Generations (2023) establishes the broad pattern. Pew Research Center surveys show roughly half of Americans aged eighteen to twenty-nine reporting no current romantic partner. The CDC’s National Survey of Family Growth shows declining rates of partnered sexual activity among young adults. Jonathan Haidt’s The Anxious Generation (2024) brings together adolescent mental-health and social-withdrawal data with the smartphone-saturation hypothesis as causal frame.

So the panelist is right that something has shifted. He is right that the shift coincides with the smartphone era. He is right that fertility, dating, and physical-mixing behaviors have moved together rather than separately. The shift is large enough that even skeptics of the strong smartphone hypothesis (Candice Odgers in Nature and elsewhere has questioned the directionality of the causation) acknowledge the data showing the change.

The cause structure he ignores

The panelist’s mistake is the move from “this changed” to “this one thing caused it.” Declining fertility has many causes operating simultaneously. Housing costs in metropolitan areas where young adults work and look for partners have risen far faster than wages over the same period. Student debt obligations delay household formation by a decade or more in countries with the steepest tuition inflation. Educational tenure now extends into the late twenties for many couples who would, in earlier decades, have married before completing school. Antidepressant prescription rates have risen sharply, and SSRI sexual side effects are pervasive enough that any honest discussion of declining sexual activity has to consider pharmaceutical effects on libido. Pornography availability and the parasocial frame it cultivates compete with effortful real-world courtship in ways that have only recently become measurable.

Women’s economic independence has, more than any other single factor, removed the marriage imperative for women specifically. A woman in 1955 needed a husband to obtain a credit card, a mortgage, a stable apartment lease, and in many states a tubal ligation or any form of reliable contraception. A woman in 2025 needs none of those things, and the marriages that earlier women contracted under economic pressure are now contracted only when they are wanted on their own terms. The corresponding decline in marriages produces a corresponding decline in marital fertility, which in countries where extramarital fertility remains low (Korea, Japan, much of Southern Europe) translates almost directly into the headline TFR.

The civic and social infrastructure that earlier generations used to find mates has thinned dramatically. Robert Putnam’s Bowling Alone (2000) documented the collapse of bowling leagues, fraternal organizations, church congregations, neighborhood card games, and similar associational life. Ray Oldenburg’s The Great Good Place (1989) named the third-place categories (cafes, taverns, barbershops, beauty parlors) where unscripted social mixing happened, and most of them have either closed or reorganized into delivery-and-screen-time formats that no longer serve the third-place function. Eric Klinenberg’s Palaces for the People (2018) extended the analysis to libraries, parks, and public schools as social infrastructure. The connection point between this thinning and the smartphone era is real but partial: phones did not close the bowling leagues. The bowling leagues closed for reasons documented by Putnam in 2000, before smartphones existed.

Anne Helen Petersen’s Can’t Even (2020) traced the millennial economic precarity that delays every milestone that historically preceded childbearing. The labor market that produces childlessness in her account is the gig economy, the credentialed-but-precarious knowledge job, the impossibility of saving on stagnant wages, the medical bill as catastrophe. Bill McKibben and others have documented the climate-anxiety strand: a meaningful fraction of young adults in surveys cite climate uncertainty as a deterrent against having children they would otherwise want.

The smartphone is one cause. Treating it as the cause requires ignoring everything above. The panelist made that move. The move is dishonest, and once it is named, his policy proposal becomes legible as a search for a single intervention that will avoid the harder work of addressing housing, debt, infrastructure, labor markets, and gender-economic transformation. Of course someone making a panel argument prefers the simpler causal story. Of course it has been wrong every prior time someone made the same kind of argument about a new medium.

The framing problem

“Society needs people” sounds neutral. It is not. The phrase carries political content that the panelist did not announce and may not have fully acknowledged to himself. There is a legitimate version of the demographic concern, and there is an illegitimate version, and the slide from one to the other is the move that needs to be flagged whenever the argument appears.

The legitimate version is fiscal and care-economy structure. Aging populations require working-age tax bases to fund pensions, public health systems, and care labor for the elderly. The dependency ratio (the number of working-age adults per retiree) determines whether existing social-insurance commitments can be honored without crushing tax burdens, benefit cuts, or collapse. Japan’s experience over the past three decades shows what a low TFR combined with restrictive immigration produces: stagnant growth, rising elder-care costs, contracting rural communities, and political pressure to maintain commitments that the working population cannot fully fund. This is a real problem with real fiscal arithmetic. It deserves serious treatment.

The illegitimate version converts fiscal arithmetic into a metaphysical claim about human species advancement. The phrase “the species cannot continue” makes population a teleological category rather than a contingent one. It treats the survival of a particular cultural-national configuration as identical to the survival of humanity, which is a category error. Eight billion humans are alive on the planet, more than at any prior point in our species’ existence, and any honest discussion of fertility decline has to acknowledge that the collapse is happening to a particular post-1945 demographic configuration in particular wealthy countries, with the human species itself in no demographic difficulty whatsoever. The pronatalist register, especially in its more excited online versions, slides further still: from “we need workers” to “we need our children specifically” to “we need to outbreed migrants” in the explicitly xenophobic register that has become disturbingly common in European and American right-wing discourse over the past decade. Ross Douthat has been writing variants of this argument from a Catholic position for years, with varying degrees of awareness of the company the position keeps.

A panelist who says “society needs people” without flagging the slide is doing one of two things. Either he is unaware of the slide, in which case his argument is intellectually thin. Or he is aware and electing not to flag it, in which case his argument is rhetorically dishonest. Neither version deserves credence at the level of policy.

The policy impossibility

“Restrict technology” functions as a wish dressed up as a policy. Converting the wish into an actual policy requires answering a series of questions that the wish-form refuses. Which technology gets restricted? Smartphones? All cameras? Recording functions specifically? Live-streaming applications? Social media platforms that enable sharing of recordings? Encrypted messaging that prevents oversight of recording sharing? AirPods? Smart watches with audio capability? Each cut produces winners and losers and constitutional and economic consequences that the wish-form pretends do not exist.

By whom is the restriction enforced? National governments operating against multinational platforms? International treaty bodies that historically take a decade to ratify and another decade to implement and another decade to enforce? Local school districts that can prohibit phones in classrooms but not at the dance afterwards? Parents who have already lost the contest with their children’s social environments? Each enforcement venue has its own jurisdictional limits and its own track record of failure on similar prior tasks.

Against whose objection? Disability advocates whose clients depend on phones for accessibility? Journalists whose work requires recording capability against state and corporate power? Workers documenting unsafe labor conditions or police brutality? Whistleblowers? Students documenting harassment? Each of these constituencies will (rightly) object to any blanket restriction, and any restriction narrow enough to satisfy them will be narrow enough to fail at the panelist’s purpose.

The history of media-restriction policy responses contains nothing encouraging. The penny-dreadful panic of the 1830s through 1890s produced moralistic legislation that failed to suppress cheap fiction. A comic-book panic in the 1950s produced the Comics Code Authority, which constrained American comics for decades, slowed the medium’s artistic development, and accomplished nothing the panic claimed it would. Rock-and-roll panic produced the Parents Music Resource Center and the Tipper Gore hearings, which produced the parental-advisory sticker, which became a marketing asset for the music it was meant to suppress. A video-game panic peaked after Columbine in 1999, produced ratings systems and proposed bans, and accomplished almost nothing while gaming became the largest entertainment industry on the planet. Each panic was wrong about the medium it targeted, wrong about the youth behavior it claimed to explain, and wrong about the policy that would address either.

It is possible that the smartphone-and-social-media transition is different from those prior transitions. The Twenge and Haidt research lines argue that it is, on the grounds that the magnitude and timing of the adolescent mental-health changes coincides with smartphone saturation in ways the earlier transitions cannot match. The argument is worth taking seriously. But “different from prior moral panics” is a long way from “the panicked policy proposal will work this time.” The structural problems of jurisdiction, definition, enforcement, and unintended consequence apply equally to this transition, and the burden of proof for any restrictive intervention sits with the proponent.

The diagnosis salvaged

Here is where the panelist had something. Reconstruct his argument charitably and the version that survives is roughly: certain kinds of relational fields require low-stakes interaction zones in which social experimentation, error, recovery from error, and rehearsal are possible, and the architecture of universal recording collapses those zones. That version is defensible.

Erving Goffman’s The Presentation of Self in Everyday Life (1959) provides the analytical apparatus. Goffman analyzed social interaction through a dramaturgical frame, arguing that competent social performance requires the distinction between frontstage (where performance happens, witnessed and evaluated) and backstage (where performers prepare, rehearse, recover, drop the performance, and exist as something other than the role they perform on the frontstage). Both spaces are necessary. A performer who has only frontstage cannot perform competently for long; the role degrades into either hollow caricature or exhausted breakdown. A performer who has only backstage never performs at all and has no opportunity to develop the competence that backstage rehearsal exists to support.

The mixer, the school dance, the bar after work, the college party, the roller rink, the church social, the Friday-night cruise, the corner store with the magazine rack and the soda fountain, all functioned as graduated frontstage zones with reliable backstage proximity. A young person could try a clumsy approach, get rebuffed, retreat to the bathroom or the parking lot or the friend group across the room, recover, try again with a different person or a different opening. The performance was witnessed, but the witnesses were finite, locally embedded, and (within reason) socially obligated to forget. By morning, last night’s awkward overture was either discussed in a closed circle of friends or quietly forgotten. The archive of social trial-and-error was small and degraded fast.

What changed in roughly 2007 to 2012 was the arrival of an instrument that converted every frontstage zone into a potential permanent archive. Phones in pockets. Cameras on every device. Cloud storage that retained material indefinitely. Social platforms that incentivized the sharing of others’ embarrassments for engagement metrics. Faces searchable through reverse-image lookups. The tools for converting a clumsy ten-second moment into a months-long social problem became universally available within five years. Tools for managing the conversion remained, and remain, almost entirely absent.

Architecturally, this is the elimination of backstage. Not the reduction of backstage. The elimination. Any space that contains another person with a phone is potentially frontstage, including spaces that previously served as recovery zones (the bathroom, the parking lot, the friend group across the room). The competent young social performer has nowhere to retreat for the rehearsal-and-recovery that competent performance requires. The rational response is to refuse performance altogether, which is what a meaningful fraction of young people are doing.

That diagnosis is real. It does not depend on any nostalgic claim about how good things used to be. No metaphysical species-progress slide is required to make the case. No technology-restriction policy follows from it. The diagnosis identifies a specific architectural change with specific behavioral consequences and locates the consequences in a sociological framework that has been operative since 1959 and is well-established in the literature.

What the older architecture actually was

A point that any honest treatment of the question has to make. The pre-smartphone mixer was not a uniformly benign social environment whose absence we should mourn without qualification. Lisa Wade’s American Hookup (2017), drawing on her sociological research at multiple American universities, documents the assault, coercion, exclusion, racial sorting, and gender-asymmetric harm that pre-smartphone party culture produced reliably and abundantly. The college fraternity party of 1985 was a site of date rape, of binge-induced injury, of homophobic violence, of the public humiliation of women whose desire was either pretended or refused. A high-school dance of 1955 sorted by race in ways that excluded entire populations from any romantic possibility within the dominant social field. Corner stores and soda fountains, in many parts of the country, were segregated by law or by custom, and the “graduated frontstage zones with reliable backstage proximity” were available only to the demographic groups the architecture was built for.

The surveillance architecture has, in some respects, exposed and reduced these older harms. A phone that records the racial slur or the unwanted advance creates a documentary record that the older architecture’s systematic forgetting was specifically designed to prevent. The clearest version of the argument concerns cell-phone footage of police violence. Eric Garner’s killing in 2014 was filmed by Ramsey Orta. Walter Scott’s killing in 2015 was filmed by Feidin Santana. George Floyd’s killing in 2020 was filmed by Darnella Frazier, then seventeen years old, whose footage was the central evidence in the criminal conviction of Officer Derek Chauvin in April 2021. Without those phones in those hands, the killings would have been recorded as the police described them in their reports, and the families would have had nothing. MeToo as a movement was made possible, in part, by the same recording infrastructure: the audio recording Ambra Battilana Gutierrez wore to her second Weinstein meeting, the screenshots and documented exchanges that survivors compiled and shared, the platforms that allowed dispersed witnesses to compare accounts and recognize a pattern. Documenting the harassment is one of the few effective remedies the harassed have ever had, and a young woman or a young Black man in 2025 has tools their grandmothers and grandfathers in 1965 would have understood as miraculous if anyone had bothered to explain them.

So the argument is not that the older architecture was good and the newer is bad. The argument is that the older architecture had a feature (graduated frontstage with reliable backstage) that the newer architecture has eliminated, and that this elimination has costs that fall most heavily on the people whose social competence was still in development. The older architecture had compensating costs that fell most heavily on people the architecture’s beneficiaries chose to exclude. Both are true. Romanticizing the older architecture is dishonest. So is dismissing the loss of its useful feature on the grounds that it had other harmful features. The interesting question is whether the useful feature can be reproduced without the harmful ones.

The actual interesting question

What kinds of low-stakes interaction zones can be deliberately constructed inside the surveillance architecture, rather than imagining the architecture away?

Phone-free zones with social cost rather than legal mandate. A salon that announces a no-phones policy at the door, asks attendees to deposit devices in a locked box, and reliably enforces the rule, has produced a temporary backstage. The enforcement mechanism is social rather than legal: those who refuse the deposit do not enter, and those who violate the rule are not invited back. The model is not new. Quaker meetings, twelve-step rooms, certain therapy groups, professional confidentiality settings, and some religious services already operate this way. Extending the model to social mixing requires hosts willing to do the curating work and attendees willing to accept the constraint as the price of entry. Both have been increasingly available as the failure of the unrestricted-phone model has become legible to participants.

Skill-based interaction structures where the encounter is about an activity rather than about the encounter. Contra dancing, climbing-gym sessions, choir practice, recreational softball, beekeeping clubs, community gardening, kayaking groups, and similar contexts produce social interaction as a byproduct of shared task. The phone is hard to operate while doing the task. The social interaction happens at lower stakes because the explicit purpose is the task, and recovery from awkward social moments is easier when the conversational frame can return to the activity. Klinenberg’s social infrastructure argument applies here directly: the institutions that produce these contexts (community recreation centers, public parks, library meeting rooms, school gymnasiums available for adult use) are the infrastructure that has been disinvested, and reinvestment in them produces social returns that are difficult to capture through markets but are nonetheless real.

Reputation systems that punish the publication of others’ embarrassments rather than rewarding it. The current platform incentive structure rewards exposure of others. A platform incentive structure could, in principle, reward discretion. The problem is that the existing platforms are not built for this and have no commercial reason to rebuild for it. The intervention has to come from the social norms governing platform use rather than from the platforms themselves: the friend group that does not film, the host who removes guests who do, the social cost imposed by peers on the person who uploads someone else’s drunk dance. Norms can be built. They take work and time, and they have to be enforced by the people who want them, but they are not impossible.

Direct construction of new third places. The disappearance of Oldenburg’s third places is not irreversible. The barbershop that refuses screens, the coffeehouse that limits laptop time, the bar with no televisions, the bookstore with chairs and a refusal of the photograph-and-tag culture, all already exist in scattered locations. Scaling them requires entrepreneurs willing to take the commercial risk and patrons willing to pay slightly more for the curated environment. This is happening, slowly, in cities where the demand has become legible. The interesting policy question is what tax structure or zoning regime makes such third places easier to operate, not what restriction on technology might make their absence less painful.

These are not solutions in the sense of fixes that restore the prior architecture. They are constructions: deliberate work to build new low-stakes zones inside the surveillance environment rather than waiting for the environment to become someone else’s policy problem. The work is already underway in scattered places. It is unevenly distributed and unevenly resourced. It is also more honest than the panelist’s wish for restriction, because it accepts the architecture as a given and asks what can be built within it.

A predictable objection: curated phone-free environments cost money to operate and to attend, which puts them out of reach of the disinvested communities most affected by social-infrastructure collapse. The objection is real and the response has to be specific. A market layer (the curated coffeehouse, the salon dinner, the boutique gym) sits at one tier of the response and is available, by design or by accident, to those who can pay. An infrastructure layer (the public library meeting room, the parks-and-recreation gymnasium, the community-college evening class, the union hall, the church basement, the tribal community center) sits at the other tier, and the political fight worth having concerns the funding and staffing of these institutions rather than the question of phone restriction. Both political parties have disinvested public social infrastructure for forty years. Klinenberg’s argument applies: libraries that close and rec centers that lose their evening hours are spaces where the recovery infrastructure of the older architecture was supposed to live for everyone who was not paying for it privately. Reinvestment in public social infrastructure is the equity dimension of the construction project. The wealthy will build their phone-free salons either way; the question is whether the public sector builds anything for everyone else.

Closing

Return to the panel. The panelist had real data and a partial diagnosis and a hopeless policy. An honest version of his argument would have separated the data from the diagnosis from the policy, would have credited the multi-causal structure of the underlying phenomenon, would have refused the metaphysical species-progress slide, and would have asked the construction question instead of the restriction question. He did none of those things, because the panel format does not reward any of them, and because the rhetorical satisfaction of “blame the phones, restrict the phones” is greater than the rhetorical satisfaction of “we have a structural problem with multiple causes and the most useful interventions are slow, local, and unglamorous.”

Eight billion humans are alive on this planet. Not all of them will reproduce. Some of them will, and some of them will work and pay taxes and care for the old. The fiscal-demographic problem is real and deserves attention. The socialization-architecture problem is also real and deserves the kind of attention this essay has tried to give it. Neither problem is solved by restricting cameras. Both are addressed, in part, by building the third places, the phone-free zones, the skill-based contexts, the reputation systems, and the social norms that the failed older architecture also depended on, even when its participants could not name what the architecture was doing for them.

The young people holding the wall at the dance are not the problem. They are responding rationally to an environment that has eliminated the conditions under which trying-and-failing is recoverable. Build them better environments and they will dance.