Abstract

ABSTRACT. Background

Social media surveillance behaviours are increasingly prevalent yet poorly operationalised in psychological research. Existing measures often fail to differentiate normative engagement from problematic monitoring. The Social Media Surveillance Scale (SMSS) was developed to address this gap by capturing behaviours along a continuum from everyday “creeping” to more intrusive behaviours such as “cyberstalking.” Previous stages involved systematic item generation and expert content validity review to ensure theoretical coverage and face validity. Building on these foundations, the current phase aimed to refine the initial 113-item pool and validate its structure, with construct validity testing now underway.

Method

A large and diverse sample of social media users (N = 1,641; Mean age = 42.65) were recruited via Prolific. Inclusion criteria required participants to be aged 18+ and active social media users. The refined 113-item SMSS was administered online via Qualtrics. Items were rated on 7-point Likert scale assessing behavioural likelihood and motivational agreement.

Data were screened for missingness, coding errors, and assumption violations. Given ordinal data and non-normality (Mardia’s test, p < .001), polychoric correlations and robust estimation methods were applied. To minimise capitalization on chance, the dataset was randomly split using the Solomon method: Subsample A (n = 821) for Exploratory Factor Analysis (EFA) and Subsample B (n = 820) for Confirmatory Factor Analysis (CFA). EFA employed Principal Axis Factoring with oblimin rotation to allow correlated factors. Factor retention decisions were guided by parallel analysis, scree plot inspection, eigenvalues > 1, and theoretical interpretability. Items were retained if they demonstrated primary loadings ≥ .40, cross-loadings < .30, and communalities ≥ .30. Internal consistency was assessed using Cronbach’s α and McDonald’s ω.

All participants provided informed consent, and the study was conducted in accordance with the British Psychological Society (BPS) Code of Ethics and approved by the University of Sunderland Ethics Committee.

Results

Initial EFA suggested an eight-factor solution, which was iteratively refined to achieve improved fit and interpretability. Parallel analysis and scree plot inspection supported a six-factor structure comprising 48 items. The six dimensions were labelled based on thematic inspection: Preoccupied Engagement, Investigative Searching, Deceptive Identity Monitoring, Privacy Intrusion, Affiliative Social Interaction, and Relational Following. Internal consistency was excellent across factors (Cronbach’s α = .82–.95; ω = .84–.95).

CFA on Subsample B confirmed the six-factor model as the best-fitting representation of the data (χ²[1065] = 11,347.39, CFI = .98, TLI = .98, RMSEA = .11), supporting the multidimensional nature of social media surveillance behaviours. Item loadings demonstrated conceptual clarity with minimal cross-loadings, and reliability indices exceeded conventional thresholds. The final 48-item SMSS therefore provides a parsimonious yet comprehensive measure of online surveillance behaviours.

Current Work and Future Directions

Construct validity testing is currently in progress and will be completed before the conference. This phase examines convergent validity with theoretically related constructs (e.g., anxiety, dark personality traits, problematic social media use) and divergent validity with opposing constructs (e.g., empathy, prosocial tendencies). Preliminary hypotheses predict positive associations with GAD-7, Short Dark Triad, and Bergen Social Media Addiction Scale, and negative associations with the Toronto Empathy Questionnaire and Prosocial Tendencies Measure. These analyses will aim to establish the SMSS within a broader nomological network, providing evidence of its theoretical and practical utility.