What the Loneliness Data Cannot Say About Technology

The claim appears in op-eds, conference talks, and policy proposals with a confidence the underlying data does not support: smartphones and social media caused the rise in loneliness. The timing is suggestive — smartphone adoption accelerated through the 2010s while several major surveys show loneliness climbing over roughly the same period. But timing is not evidence of mechanism, and a look at how these surveys actually measure technology use shows why the causal version of the claim runs well ahead of what has been tested.

What gets asked, and what does not

The American Enterprise Institute’s 2021 survey on American friendship, drawing on a large national sample, documents a striking decline: the share of Americans reporting no close friends rose from 3% in 1990 to 12% in 2021, and the share of men with six or more close friends fell from 55% to 27% over the same period. This is a real and well-measured trend in friendship quantity. It says nothing directly about technology, because the survey did not ask respondents to attribute the decline to any cause, nor did it measure their technology use alongside their friendship counts.

The Harvard Graduate School of Education’s Making Caring Common survey, published in 2021, found that 61% of young adults aged 18 to 25 reported serious loneliness, with 43% saying their loneliness had increased since the pandemic began. That report is candid about what it can and cannot establish: it is a single cross-sectional snapshot of self-reported feeling, not a study of technology use, and it does not include a measure of screen time, platform use, or online-to-offline contact ratios that would let a reader connect the loneliness figure to any specific technology behavior.

Cigna’s 2020 workplace loneliness report found that 73% of workers aged 18 to 22 report loneliness, and that lonely workers miss work roughly twice as often for illness and five times as often for stress-related reasons. Again, useful and specific numbers — about loneliness and its correlates in the workplace. Not a technology study.

This is the pattern across nearly the entire loneliness literature: technology is the implied backdrop, rarely the measured variable. Surveys ask “how often do you feel lonely” using validated instruments like the UCLA Loneliness Scale, cited in the AARP Foundation’s 2018 survey of adults 45 and older. They far less often ask “how many hours did you spend on which platform, doing what, with whom” in a way precise enough to test against loneliness scores in the same respondents.

Self-report on top of self-report

Where technology use is measured, it is almost always by self-report — “how often do you use social media” — rather than by device logs or platform data. Self-reported technology use is subject to the same recall and framing problems as self-reported loneliness, and the two self-reports are usually collected in the same survey instrument, at the same moment, from the same person. That creates a structural problem for any causal claim: a lonely respondent may over- or under-report their technology use in ways correlated with their mood, independent of what they actually did. A study that wants to show technology use predicts later loneliness needs technology data collected separately from, and prior to, the loneliness measurement. Almost none of the widely cited loneliness surveys were designed this way, because loneliness measurement, not technology measurement, was their primary purpose.

The instrument mismatch problem, applied to technology

The Center’s usual complaint about instrument mismatch — that different loneliness surveys ask different questions and so cannot be compared — applies with more force to technology. There is no equivalent of the UCLA Loneliness Scale for technology-mediated connection: no single, validated instrument that most researchers use, that has been checked for reliability across populations, that distinguishes between scrolling passively, messaging a close friend, and coordinating an in-person meetup through an app. “Screen time” as typically measured collapses all of these into one number. A person who spends two hours a day in a group chat with three close friends and a person who spends two hours a day watching short videos alone are recorded identically by most existing instruments. Any resulting correlation with loneliness is close to uninterpretable, because it averages over behaviors that plausible theory would expect to have opposite effects.

What a structural account would need

Robert Putnam’s 2000 account of declining American civic participation offers a useful contrast in method. Putnam relied on decades of repeated measures of specific behaviors — club membership, union rolls, church attendance, card-playing — that could be tracked as continuous series and cross-checked against each other. The technology-and-loneliness literature has no comparable backbone: it has scattered cross-sectional surveys, taken with different instruments at different times, asking different populations different questions, sitting next to adoption curves for products that themselves changed substantially year to year. A researcher wanting to make Putnam’s kind of structural argument about technology would need repeated, comparable measures of specific technology behaviors over time, linked to the same individuals’ social contact and loneliness scores — not a series of one-off snapshots loosely correlated by year.

Robin Dunbar’s social brain hypothesis, and his continuing argument that stable social relationships are capped at roughly 150 regardless of contact medium, points to a further complication. If the constraint on meaningful relationships is cognitive rather than technological — a claim Dunbar has made explicitly in arguing that his number has held up across three decades of scrutiny — then the number of contacts a platform enables may be close to irrelevant to loneliness, while the quality and layering of contact within that fixed capacity is what matters. Almost no widely cited loneliness survey captures layering: whether a person’s contacts are concentrated in five intimates or spread thin across five hundred followers. A technology measure built around hours or friend-counts cannot distinguish these cases, even though the theory says they should produce different outcomes.

Why the age pattern matters for how this gets read

The National Academies’ 2020 consensus report on older adults notes that roughly a quarter of adults 65 and older are socially isolated, using structural measures of network contact rather than subjective feeling. Younger adults, by contrast, report the highest loneliness rates of any age group in the Harvard survey. If technology were driving loneliness in a simple, direct way, the age pattern is not obviously the one theory would predict, since technology adoption and use is heaviest among the young. This does not rule out a technology effect — it may operate through different mechanisms at different ages, or interact with life stage in ways not captured by the current data — but it is a pattern that any causal claim about technology and loneliness needs to explain rather than skip past.

What would settle more of this

A study built to answer the technology question directly would need three things the current literature mostly lacks: a validated instrument that separates types of technology-mediated contact rather than aggregating them into screen time; repeated measurement of the same individuals over years rather than a single cross-section; and technology data collected independently of, and before, the loneliness measurement, so that later loneliness can be modeled as an outcome rather than a simultaneous correlate. Until such a design exists at scale, the honest position is that the available surveys can describe co-occurring trends in loneliness and technology adoption, and can rule out some crude versions of the causal story, but cannot support the strong claim now doing most of the work in public argument.