Two symptom bridges link social media problems and depression in students

For years, the research headline has been deceptively simple: problematic social media use goes hand in hand with depression. Meta-analyses pooling dozens of studies and tens of thousands of participants have repeatedly found that students who score high on measures of compulsive, poorly controlled social media engagement also tend to report more depressive symptoms. But a new study from Poland argues that this familiar correlation hides more than it reveals, and its findings may change how scientists, clinicians, and universities think about the relationship between our feeds and our moods.

Published in Current Psychology, the study was led by Sławomir Rębisz of the University of Rzeszów, together with Paweł Grygiel and Sylwia Opozda-Suder of the Jagiellonian University in Kraków and Roman Dolata of the University of Warsaw. The team surveyed 936 students from the University of Rzeszów, 78.1 percent of them female, with an average age of 21.64 years, drawn from 65 different academic programs. Rather than comparing two questionnaire totals, the researchers treated every single symptom as a node in a network and mapped the conditional connections between them, asking a question that total-score studies cannot answer: which specific components of problematic social media use are actually tied to which specific depressive symptoms?

The technical machinery behind the study is worth unpacking, because it represents a broader shift in how psychopathology is studied. The students completed the Bergen Social Media Addiction Scale, or BSMAS, which measures six addiction-like components derived from Griffiths’ components model of addiction: salience, tolerance, mood modification, withdrawal, conflict, and relapse. They also completed the Patient Health Questionnaire-9, or PHQ-9, whose nine items correspond to the DSM symptom criteria for major depression, ranging from anhedonia and depressed mood to sleep disturbance, fatigue, feelings of worthlessness, concentration difficulties, psychomotor changes, and suicidal ideation. The researchers then estimated a Graphical Gaussian Model with LASSO regularization, a statistical technique that shrinks weak partial correlations toward zero and sets some of them exactly to zero, so that only robust conditional associations survive. The Extended Bayesian Information Criterion selected the optimal regularization parameter, and nonparametric bootstrapping, including case-dropping subset bootstrapping, verified that the network’s central nodes remained stable when portions of the data were removed.

The first striking result concerns the architecture of the network itself. Using seven different community detection algorithms, including Louvain, Walktrap, Fast Greedy, Edge Betweenness, Label Propagation, Leading Eigen, and Infomap, the researchers tested whether the fifteen questionnaire items cluster into distinct communities. The answer was unambiguous on one side and revealingly messy on the other. The six BSMAS indicators formed a coherent, stable community in every single solution, structurally distinguishable from depressive symptoms. The PHQ-9 symptoms, by contrast, showed a heterogeneous organization that shifted depending on the algorithm. Some solutions divided depression into a somatic-vegetative component, containing insomnia, fatigue, and appetite changes, and a broader affective-cognitive-psychomotor component. Others placed concentration difficulties and psychomotor agitation or retardation with the somatic symptoms instead, or even carved them out as a separate small community. This boundary-like position of concentration problems and psychomotor symptoms mirrors a long-standing debate in the psychometric literature about whether the PHQ-9 is best described as essentially unidimensional or as two strongly related factors.

Within each domain, the centrality analysis identified the symptoms most strongly embedded in the internal structure of their own community. On the social media side, tolerance, the sense of needing increasing amounts of time online to achieve the same effect, emerged as the most central indicator, connecting most strongly with salience, the preoccupation with social media, and also with mood modification and withdrawal. On the depression side, sadness and guilt, the latter capturing dissatisfaction with oneself and feeling like a failure, were the most central symptoms, and they also formed the strongest connected pair in the entire PHQ-9. Yet here lies a subtle and important finding: tolerance, despite being the hub of the problematic-use community, did not form any direct connection with depressive symptoms at all. High centrality within one’s own domain, the study shows, does not automatically translate into relevance for cross-domain associations.

And that brings us to the study’s headline result, the one that gives the familiar correlation its precise shape. When the researchers examined bridge edges, the direct cross-domain connections between the two communities, they found exactly two. Not a diffuse web of associations linking every compulsive-use indicator to every depressive symptom, but two selective points of contact. The first connected sadness, one of the core symptoms of depression, with mood modification, the BSMAS component describing the use of social media to forget about personal problems. The second connected concentration difficulties with conflict, the component describing the disruptive impact of social media use on studying or work. Every other PHQ-9 symptom and every other BSMAS indicator was connected only within its own domain or showed no nonzero connection in the regularized network.

Each of these two bridges tells a different story. The sadness-mood modification edge is an affective-regulatory point of contact, and it fits neatly with the compensatory internet use theory proposed by Kardefelt-Winther, according to which online activity is undertaken in response to emotional difficulties or offline problems. It also aligns with longitudinal research showing that depressive symptoms may predict later increases in using the internet for mood regulation. But the researchers are careful: using social media to improve mood is not maladaptive in itself, and the edge may partly reflect broader negative affectivity or dispositional factors such as neuroticism, which is strongly associated with both depression and problematic technology use. The concentration problems-conflict edge, meanwhile, is a functional-cognitive point of contact, consistent with the deficient self-regulation perspective, in which uncontrolled internet use arises from difficulties in monitoring and regulating one’s own behavior. Attentional difficulties on the depression side were tied to the component of problematic use concerning disruption of academic and occupational responsibilities, a connection that previous network studies on problematic internet use and platform-specific use have also highlighted.

The practical implications are as counterintuitive as they are important. The findings do not support the simple conclusion that reducing time spent on social media, or temporary abstinence, is a sufficient strategy for improving students’ mental health. Recent meta-analytic evidence is mixed in this respect: one meta-analysis of randomized controlled trials found only marginal beneficial effects of social media restriction on subjective well-being, while a meta-analysis focusing specifically on social media abstinence found no clear evidence that abstinence affected positive affect, negative affect, or life satisfaction. Instead, the study points toward helping students recognize when, why, in response to which emotions, and with what consequences they use social media. The aim should not be to discourage using social media to improve mood, but to broaden the repertoire of emotion-regulation and coping strategies: offline social contacts, behavioral activation, physical activity, problem-focused coping, and more intentional use of digital media. On the functional side, assessment should focus on the quality of concentration, sleep rhythm, task postponement, difficulty completing work, and repeated unsuccessful attempts to reduce use despite negative consequences.

The researchers are equally candid about the limits of their evidence. The study is cross-sectional, so the network edges describe conditional associations rather than temporal or causal relations; it is impossible to determine whether depressive symptoms precede an increase in problematic use, whether problematic use precedes an increase in depressive symptoms, or whether the relations are reciprocal. The sample came from a single university in a single country, the measures were self-reported, and the study did not assess specific patterns of social media activity, such as active versus passive use, platform-specific behaviors, or objective time spent online. The sadness-mood modification edge in particular may partly reflect the affective proximity of the two items or shared vulnerability factors rather than a construct-specific association. The authors frame their two bridge edges as hypothesis-generating findings that future longitudinal studies, cross-lagged panel network models, and ecological momentary assessment designs should test over time, ideally after accounting for neuroticism, self-control, stress, and offline social functioning.

What the study ultimately delivers is a more differentiated picture of a relationship that public discourse has often flattened into a simple cause-and-effect narrative. Problematic social media use and depression are structurally distinct, the new analysis shows, and their association is not a broad, diffuse entanglement but a concentrated structure with two selective points of contact: depressed mood tied to escapist use, and attentional difficulties tied to functional disruption. As of April 2026, active social media user identities worldwide had reached 5.79 billion, equivalent to 69.9 percent of the global population, making the stakes of getting this science right enormous. The message for universities, clinicians, and students themselves is not that social media is a poison to be abstained from, nor that the familiar correlation is meaningless, but that the meaningful questions are now symptom-specific: which emotions drive use, which functions it serves, and whether it remains under control or begins to interfere with the tasks of everyday life.

,

Subject of Research: Mapping the symptom-level network between problematic social media use and depression among university students

Article Title: Mapping the symptom-level network between problematic social media use and depression among university students

Article References: Rębisz, S., Grygiel, P., Opozda-Suder, S., & Dolata, R. (2026). Mapping the symptom-level network between problematic social media use and depression among university students. Current Psychology, 45(20), Article 1628. https://doi.org/10.1007/s12144-026-10200-x

Image Credits: AI Generated

DOI: 10.1007/s12144-026-10200-x

Keywords: Mapping, symptom-level, network, between, problematic, social, media, depression, university, students, scientific research