Surfing into trouble? How internet use influences early adolescent behavior through diverse pathways?

Adolescents comprise a substantial share of global Internet users, including in China, where by end-2021 over a billion people were online and about 18% were under 17. Early adolescence (approximately ages 11–14) is marked by heightened arousal and challenges in emotion and behavior regulation, making youth susceptible to externalizing behaviors (e.g., aggression, rule-breaking, substance use). Problem behavior theory emphasizes the role of environmental influences; excessive Internet use, as an environmental factor, has been linked to poorer academic performance, disrupted daily routines and family relationships, and elevated emotional/behavioral problems. Although prior research has examined Internet use and adolescent problems, gaps remain for early adolescents in China and for externalizing outcomes specifically. The literature lacks clarity on mechanisms linking Internet use to externalizing behaviors and on heterogeneity across demographic groups (e.g., gender, urban–rural). Given the developmental significance and consequences of externalizing behaviors, this study investigates whether Internet use predicts externalizing problem behaviors in early Chinese adolescents and examines mediating mechanisms (self-education expectations and peer relationships) and heterogeneity by gender and residence. The work aims to inform interventions and policies that support healthier digital engagement among youth.

Guided by multiple frameworks, the review links Internet use to early adolescent externalizing behaviors. Social Learning Theory (SLT) posits learning via observation and modeling in social contexts; online environments expose adolescents to role models and behaviors (including antisocial or aggressive content) and offer reinforcement through likes and comments, with anonymity potentially emboldening risky actions. Empirical studies associate exposure to online aggression and antisocial communities with increased physical aggression and rule-breaking. Self-education expectation as mediator: Self-Determination Theory (SDT) distinguishes intrinsic versus extrinsic motivation. When Internet use aligns with intrinsic goals (e.g., self-education), it supports autonomy, competence, and relatedness; low self-education expectations or extrinsically motivated use may undermine these needs and increase externalizing behaviors. Evidence suggests higher educational expectations predict better academic outcomes and fewer problem behaviors, whereas excessive Internet use can diminish self-education expectations. Peer relationships as mediator: Drawing on Attachment Theory (AT) and ecosystem perspectives, peer relationships in school microsystems are central to adolescent adjustment. Internet use may both support and disrupt peer bonds; problematic or addictive use relates to poorer peer relations, social isolation, and loneliness. High-quality peer relationships foster prosocial behaviors and reduce problem behaviors; poor peer relations increase risks for aggression and rule-breaking. Thus, peer relationships can mediate Internet use effects on externalizing behaviors. Chain mediation (self-education expectation → peer relationships): Ecological Systems Theory (EST) suggests adapted self-education expectations can influence interactions with peers in the microsystem and across mesosystem linkages. High self-education expectations may enhance collaborative academic engagement and peer support; overly high, fixed expectations could conversely strain relationships. Therefore, Internet use may affect externalizing behaviors indirectly through sequential effects on self-education expectations and then peer relationships. Heterogeneity: Gender differences in Internet use patterns and externalizing tendencies suggest stronger adverse effects for males. Urban–rural differences in online activities (urban: information/communication; rural: entertainment) imply potentially greater risk in rural contexts. Hypotheses: H1 Internet use positively predicts externalizing problem behaviors. H2 Self-education expectation mediates the association. H3 Peer relationships mediate the association. H4 Self-education expectation and peer relationships exert sequential mediation. H5 Effects differ by gender. H6 Effects differ by urban–rural residence.

Design and sampling: Cross-sectional survey using stratified and random cluster sampling. Data were collected in March 2023 from three junior high and three primary schools across Liaoning, Henan, and Sichuan provinces in China. Permissions were obtained from schools; written informed consent was secured from students and guardians. Anonymity/confidentiality and right to withdraw were ensured. Trained assistants administered pencil-and-paper questionnaires during school hours; participants received a small gift. Participants: Of 1191 students approached, 1155 valid responses remained after excluding 36 with substantial missing data (response rate 97%). Ages 11–14 years (M=12.36, SD=0.553); 48.2% boys (n=557); 51.8% girls (n=598); 44.9% rural (n=519); 55.1% urban (n=636). Measures: Externalizing problem behaviors assessed by the Youth Self-Report (YSR) Externalizing scale (Aggressive Behavior and Rule-Breaking; 26 items described; 3-point Likert; higher mean = more externalizing; Cronbach’s α=0.930). Internet use measured as average daily time online on weekdays and weekends with a 6-point scale (1 <30 min to 6 >4 h); a weighted average across 7 days computed (validated previously). Self-education expectation assessed by desired highest education level (1=drop out to 9=PhD). Peer relationships measured by PROMIS Pediatric Peer Relationship 8-item scale (1 never to 5 almost always; higher mean = better relationships; Cronbach’s α=0.887). Control variables included family book collection, parents’ education, parental educational expectations, family economic conditions, gender, nationality, age, household registration (urban/rural), and physical health. Descriptive statistics (selected): Externalizing behaviors M=0.375 (SD=0.315); Internet use M=2.344 (SD=1.210); Self-education expectation M=7.010 (SD=1.460); Peer relationships M=3.935 (SD=0.857). Analytic strategy: Analyses conducted with SPSS 23.0, STATA 16.0, and Mplus 8.3. Steps included: (1) Harman’s single-factor test for common method bias (largest factor=30.02% <40% threshold). (2) Correlations among key variables. (3) OLS regressions predicting externalizing behaviors from Internet use with controls. (4) Generalized propensity score-based dose–response modeling (GPSM) to visualize the dose–response function. (5) Structural equation modeling (SEM) to test direct and indirect (mediation and chain mediation) effects with bootstrapped SEs and 95% CIs; fit indices reported (CFI=1, TLI=1, RMSEA=0, SRMR=0). (6) Heterogeneity analyses by gender and urban–rural residence with Fisher’s empirical p-values (5,000 bootstrap resamples) to test coefficient differences across groups.

- Correlations: Externalizing behaviors correlated positively with Internet use and negatively with self-education expectation and peer relationships (reported significant associations at p<0.01; e.g., total effect below). - OLS regressions: Internet use positively predicted externalizing behaviors across specifications; coefficients approximately 0.061***, 0.057***, and 0.044*** (robust SEs reported), controlling for covariates. - Dose–response (GPSM): The dose–response curve indicated a linear, positive relationship between Internet use and externalizing behaviors with 95% CIs enclosing the estimated function. - SEM path coefficients: Internet use → Externalizing behaviors: β=0.159, t=5.667, p<0.01. Internet use → Self-education expectation: β=−0.220, t=−7.302, p<0.01. Self-education expectation → Externalizing: β=−0.158, t=−5.595, p<0.01. Internet use → Peer relationships: β=−0.091, t=−3.006, p<0.01. Peer relationships → Externalizing: β=−0.315, t=−12.295, p<0.01. Self-education expectation → Peer relationships: β=0.146, t=4.700, p<0.01. - Mediation effects (bootstrapped): Indirect via self-education expectation: β=0.035*** (95% CI: 0.020–0.053). Indirect via peer relationships: β=0.029** (95% CI: 0.010–0.048). Sequential mediation (self-education expectation → peer relationships): β=0.010** (95% CI: 0.006–0.016). Direct effect: β=0.159*** (95% CI: 0.103–0.213). Total effect: β=0.233*** (95% CI: 0.170–0.292). - Heterogeneity by gender: Effect of Internet use on externalizing was stronger for males (β≈0.058**, SE=0.012) than females (β≈0.025*, SE=0.010); Fisher empirical p=0.025** (difference significant). - Heterogeneity by residence: Effect stronger for rural (β≈0.057***, SE=0.013) than urban (β≈0.033***, SE=0.010); Fisher empirical p=0.062* (difference marginally significant). Overall, H1–H4 supported (positive association and both mediators, including chain mediation). H5 and H6 supported (significant gender and urban–rural differences).

Findings show that greater Internet use is linked to higher externalizing problem behaviors in early adolescence, consistent with Social Learning Theory: digital exposure can normalize or reinforce aggression, rule-breaking, and other externalizing patterns, with online anonymity and feedback mechanisms potentially amplifying such behaviors. The mediating analyses clarify mechanisms. First, Internet use was associated with lower self-education expectations, which in turn predicted fewer externalizing behaviors when higher; this aligns with Self-Determination Theory, whereby unstructured or extrinsically motivated online activity may undermine autonomy and academic motivation, increasing behavioral problems. Second, Internet use was linked to poorer peer relationships, which predicted higher externalizing problems. Consistent with Attachment Theory and ecosystem perspectives, disruptions to peer connectedness—via reduced face-to-face interaction or lower social competence—can elevate conduct problems. The sequential mediation indicates that Internet use may reduce self-education expectations, which then weakens peer relationships, cumulatively increasing externalizing behaviors. Heterogeneity analyses revealed stronger adverse effects for males (potentially due to higher engagement and risk-prone online activities) and for rural adolescents (potentially reflecting more entertainment-focused use, less supervision, and contextual constraints). Collectively, these results affirm the multi-pathway influence of Internet use on externalizing behaviors and underscore the importance of both individual motivational factors and social relational contexts in shaping behavioral outcomes.

This study demonstrates that Internet use is positively associated with externalizing problem behaviors among Chinese early adolescents. Both self-education expectations and peer relationships independently mediate this association, and a sequential pathway (self-education expectation → peer relationships) further explains how Internet use translates into behavioral outcomes. Effects are stronger in males than females and in rural versus urban youth. The work advances theory by integrating SLT, SDT, AT, and EST to explain multi-level mechanisms linking digital engagement to behavior and offers practical guidance: enhance adolescents’ Internet literacy and self-regulation, support academic aspirations, and strengthen peer connectedness, with gender- and context-sensitive strategies. Future research should employ longitudinal or experimental designs, broaden samples beyond three provinces, and incorporate multi-informant/objective measures to validate causal pathways and explore additional mediators/moderators.

- Geographic scope: Data from three Chinese provinces (Liaoning, Henan, Sichuan) may limit generalizability; a nationally representative sample is needed. - Self-report bias: Reliance on adolescent self-reports risks social desirability and recall biases; future work should include parent/teacher reports and objective indicators. - Cross-sectional design: Limits causal inference; longitudinal designs are needed to establish temporal ordering and causality. - Model scope: Focused on a specific mediation pathway; alternative mechanisms and additional variables should be tested using longitudinal or experimental approaches. - Omitted variables: Not all determinants of externalizing behaviors were measured; future research should include broader individual, family, school, and community factors.