Social media use, online experiences, and loneliness among young adults: A cohort study

The study examines whether and how digital technology use, particularly social media, is related to loneliness among young adults. Against a backdrop of widespread smartphone adoption and diversified online communication (including social media, forums, and dating apps), concerns have grown that digital interaction may contribute to loneliness in youth—an age group with higher reported loneliness than older adults. Theoretical perspectives offer differing predictions: displacement hypothesis (online time displaces more beneficial offline interactions); stimulation hypothesis (online use augments and enhances existing offline ties), including the rich-get-richer view; and compensatory perspectives suggesting lonely individuals may find online spaces to meet social needs. The heterogeneity of platforms (networking, content creation, passive consumption, dating) and user experiences (e.g., cybervictimization) likely moderates associations. The COVID-19 pandemic created a natural experiment, shifting social contact online, potentially altering or leveling associations between technology use and loneliness. The research questions focus on platform-specific use, online experiences, and user perceptions in relation to loneliness, accounting for prior loneliness, depression, and anxiety measured at age 18, and comparing responses before versus during the first UK lockdown.

The paper situates its inquiry within established theories and empirical findings on digital media and psychosocial outcomes. Displacement versus stimulation hypotheses have been tested in adolescent populations, with evidence that online communication can either reduce or fail to replace offline interactions depending on context. Prior work suggests associations with mental health (depression, anxiety, loneliness) depend more on how and why social media is used than on frequency alone. Research highlights platform heterogeneity: networking sites (e.g., Facebook) may build social capital; image-centric platforms (e.g., Instagram) can foster social comparison pressures; passive consumption platforms (e.g., YouTube) may relate differently to wellbeing. Studies underscore that problematic or compulsive use is linked to poorer outcomes, while supportive online interactions may buffer risk for certain groups. Cybervictimization has been identified as a distinct and significant risk factor for loneliness, with characteristics unique to online spaces (anonymity, persistence across time, spillover offline). The pandemic literature reports elevated loneliness among youth and reliance on digital tools to cope, but also notes that screen-based interactions may not fully substitute for in-person contact. Meta-analytic and review evidence indicates that item framing (e.g., measuring problematic use vs. neutral metrics like frequency) influences detected associations with wellbeing.

Design: Cohort study using the Environmental Risk (E-Risk) Longitudinal Twin Study of British twins born 1994–1995, followed from early childhood to young adulthood. Primary data were collected via the Social Media and Social Mobility (SM2) online survey (2019–2020) at mean age 26. Sample: 1632 participants (73.1% of original cohort; 76.6% of age-18 participants), with targeted efforts to maintain representation of males and low-SES individuals. The cohort reflects the full UK SES spectrum and geographic distribution. Measures:

• Loneliness at age 26: Four items from the UCLA Loneliness Scale v3 (lack companionship, left out, isolated, alone). Responses scored 0 (hardly ever), 1 (sometimes), 2 (often); summed 0–8 (M = 2.43, SD = 2.27; α = 0.84). Loneliness at age 18 assessed with the same short scale (M = 1.57, SD = 1.94; α = 0.83).
• Digital media use: Time spent on social media, watching TV, gaming, looking for information online, and total time online (1 = none to 8 = 7+ hours/day).
• Platform use and frequency: Use of Facebook, Instagram, Twitter, Snapchat, WhatsApp, YouTube, Reddit, and dating sites/apps (yes/no). Frequency for users (1 = less than once a month to 6 = more than 5 times/day).
• Perceptions and behaviors: Perceived effect of social media on loneliness (more lonely/less lonely/no difference). Frequency of actively posting and passively scrolling (1 = never to 7 = more than 5 times/day). Compulsive/problematic technology use: 7-item sum (M = 10.90, SD = 2.88; α = 0.76). Use of social media to improve mental/physical health; perceived emotional support online.
• Online victimization: Seven types (offensive names 36%, bothered/harassed 33%, physically threatened 13%, purposely embarrassed 28%, unwanted sexual questions 21%, explicit photos shared without consent 8%, received explicit photos without consent 27%). Coded 1 = never, 2 = once, 3 = multiple times. Composite indicators: any victimization at least once; any repeated victimization; multiple types of repeated victimization. Covariates: Biological sex; family SES (latent factor from income, education, occupation at age 5; categorized low/medium/high). Age 18 diagnoses of major depressive disorder and generalized anxiety disorder via structured clinical interview (DSM-IV). Analysis: Linear regressions with loneliness at age 26 as the dependent variable. Each predictor entered in separate models controlling for sex and SES (Model 1), then additionally for age-18 depression and anxiety (Model 2). Interaction tests with sex, age-18 loneliness, depression, and anxiety. Robust standard errors accounting for twin clustering (Stata v16, vce cluster). COVID-19 moderation: comparisons before (pre-March 23, 2020) versus during first UK lockdown using interaction terms; subgroup characteristics summarized separately. Open science: Preregistered analysis plan; code available online.

Perceptions:

• 71.4% reported social media made no difference to loneliness; 12.8% less lonely; 15.9% more lonely. Those endorsing no difference had lower age-18 loneliness (M = 1.34) than less lonely (M = 1.92; t(1592) = 3.47, p = 0.001) and more lonely groups (M = 2.48; t(1592) = 7.39, p < 0.001). Digital media time (Table 1):
• Social media time: no association with loneliness (B ≈ 0.00; p ≈ 0.95–0.97).
• TV: positive association (B = 0.10; 95% CI [0.02, 0.18]; p ≈ 0.010–0.011); stronger among those lonelier at 18 (interaction β = 0.07; 95% CI [0.02, 0.12]; p = 0.005).
• Gaming: positive association (B = 0.12→0.10; p < 0.001→0.001).
• Looking for information online: positive (B = 0.09→0.07; p = 0.007→0.043).
• Total time online: positive (B = 0.22→0.20; p < 0.001). Platforms used (Table 2):
• Networking platforms (Facebook, Instagram, Twitter, Snapchat): no significant associations.
• WhatsApp use associated with lower loneliness (B = −0.42→−0.38; p = 0.028→0.046).
• YouTube: positive association in Model 1 (B = 0.29; 95% CI [0.04, 0.55]; p = 0.023), attenuated in Model 2 (ns).
• Reddit: higher loneliness (B = 0.72→0.62; p < 0.001→0.001).
• Dating sites/apps: higher loneliness (B = 1.19→1.15; p < 0.001). Frequency:
• More time on YouTube related to higher loneliness (B = 0.20; 95% CI [0.14, 0.26]; p < 0.001), not explained by age-18 depression/anxiety.
• Greater WhatsApp frequency associated with lower loneliness (β = −0.20; 95% CI [−0.32, −0.08]; p = 0.001).
• Instagram frequency: marginal overall trend to lower loneliness (p = 0.061); significantly lower loneliness among those with age-18 anxiety (β = −0.24; 95% CI [−0.43, −0.05]; p = 0.014). Behaviors and experiences:
• Active posting and passive scrolling: no associations with loneliness.
• Compulsive/problematic technology use: higher loneliness (β = 0.37; 95% CI [0.32, 0.42]; p < 0.001).
• Using social media to improve mental/physical health: higher loneliness (β = 0.22; 95% CI [0.17, 0.27]; p < 0.001).
• Emotional support online: no overall association; linked to lower loneliness among those with age-18 anxiety (β = −0.20; p = 0.036) and those above-median loneliness at 18 (β = −0.08; p = 0.023). Cybervictimization:
• Any victimization at least once (β = 0.22; 95% CI [0.17, 0.27]; p < 0.001); any repeated victimization (β = 0.22; 95% CI [0.17, 0.27]; p < 0.001); multiple types of repeated victimization (β = 0.21; 95% CI [0.16, 0.25]; p < 0.001). Sex differences:
• Females had higher mean loneliness than males (t(1,591) = 2.15, p = 0.032).
• In males, increased time on Twitter associated with lower loneliness (β = −0.15; 95% CI [−0.28, −0.01]; p = 0.040); no effect in females.
• Repeated online victimization more strongly associated with loneliness in females (β = 0.26) than males (β = 0.16), both significant. COVID-19 period comparisons:
• Mean loneliness did not differ before vs. during first lockdown (M ≈ 2.46 vs. 2.39; p = 0.552). Perceptions shifted modestly (pre: 18.1% felt social media made them lonelier; during: 14.3% felt less lonely).
• Associations present before but not during: TV (β = 0.09, p = 0.008 before; ns during), information seeking (β = 0.07, p = 0.022 before; ns during), YouTube (β = 0.09, p = 0.011 before; ns during), Reddit (β = 0.14, p < 0.001 before; ns during). Increased Instagram frequency linked to lower loneliness before (β = −0.08, p = 0.032) but not during lockdown.

Findings indicate that overall time online relates to higher loneliness, but time spent specifically on social media does not. Platform-specific analyses reveal heterogeneity: networking/messaging platforms, notably WhatsApp, may support social connection and are associated with lower loneliness, while platforms characterized by passive consumption (YouTube) or large, open communities (Reddit) show positive associations with loneliness—patterns likely reflecting selection effects wherein lonelier individuals gravitate to certain online activities or communities. Use of dating apps was strongly associated with higher loneliness, plausibly indexing offline circumstances (e.g., being single). Beyond platform use, the nature and motivations of engagement matter: compulsive/problematic technology use and using social media for health coping were associated with higher loneliness, whereas perceived emotional support online related to lower loneliness among individuals with prior anxiety or elevated earlier loneliness, suggesting potential benefits for vulnerable subgroups. Cybervictimization remained a robust correlate of loneliness in young adulthood, extending earlier findings from adolescence. Sex-stratified analyses suggest some differential patterns (e.g., Twitter linked to lower loneliness among males; stronger victimization effects in females). COVID-19 comparisons suggest a leveling effect during lockdown: several associations between passive media use and loneliness evident pre-pandemic were not observed during lockdown, consistent with widespread increases in online time reducing the discriminatory power of usage patterns to differentiate lonely from non-lonely individuals. Overall, the results align with evidence that neutral metrics of social media use (time/frequency) show weak associations with wellbeing, while measures capturing problematic use or adverse experiences reveal stronger links.

Social media use per se, including frequency, was not a strong indicator of increased loneliness among young adults. Messaging/networking platforms (e.g., WhatsApp) may support connection and relate to lower loneliness, whereas platforms such as Reddit and dating apps were associated with higher loneliness, likely reflecting selection based on offline circumstances or needs. Compulsive/problematic technology use and experiences of online victimization were consistently linked to greater loneliness, identifying vulnerable groups for targeted support. Longitudinal and experimental studies are needed to assess causal pathways and to examine platform- and feature-specific effects over time, including evolving platforms (e.g., TikTok). Overall, social media appears to be a less powerful correlate of loneliness than other established risks (e.g., bullying/victimization).

The study is observational and cross-sectional at age 26; causal direction cannot be inferred and associations may be confounded by third variables (e.g., dating app use may proxy being single). Self-reported technology use is susceptible to subjectivity and recall bias; objective usage tracking was not available. Measures of posting/scrolling were general and did not capture within-platform heterogeneity; cybervictimization analyses were not platform-specific. Platform usage is correlated, complicating attribution of effects to individual platforms. The social media landscape evolves rapidly (e.g., lack of TikTok data at survey time). COVID-19 comparisons involved independent pre/during samples, limiting true pre/post inference, and data collection ended before subsequent lockdowns that may have altered loneliness trends. As a twin cohort, early-life sibling dynamics may influence baseline loneliness. Effect sizes observed were small and should be interpreted cautiously.