This commentary complements the arguments by Tam et al. (2025) and offers a comprehensive approach when considering light conditioning in the context of the development and maintenance of problematic Internet use. Thereby, we illustrate the impact of light on humans also within the context of theoretical considerations and empirical studies. We agree with Tam et al. (2025) that there is a need for a better understanding of habit formation within addictive behaviors and we acknowledge the empirical challenges which are proposed in the article mentioned. At the same time, this commentary emphasizes that light should not be considered as an isolated reinforcer, but rather as a complementary component of other conditioned and cognitive tendencies (e.g., use expectancies), as well within the interplay of further processes in addiction research such as craving, attentional biases, and self-control abilities.
Low reproducibility in brain-behavior association studies has prompted calls for larger samples, but rigorous quality control (QC) of behavioral measures may also strengthen the robustness of observed associations. This study examined how varying levels of behavioral QC affect brain-behavior associations using delay discounting, a widely studied measure of immediate reward orientation. Data were drawn from the Adolescent Brain and Cognitive Development Study (n = 10,936), and delay discounting data were subjected to eight increasingly stringent QC levels. Associations between delay discounting and brain structural morphometry were evaluated in 15 bilateral a priori regions of interest (ROIs), and changes in effect size, significant associations, and multivariate behavioral prediction were quantified across QC levels. Greater QC stringency substantially amplified association strength, with effect sizes increasing by an average of 188.52% across ROIs, the number of statistically significant associations more than doubling, and multivariate prediction improving by 267%. Although higher QC levels reduced sample size by up to 65%, strengthening of associations plateaued at intermediate levels, suggesting asymptotic improvements in resolution. Overall, stringent behavioral QC substantially increased the magnitude of brain-behavior associations, suggesting this is a critical methodological consideration for improving reproducibility in cognitive neuroscience.
Digitalisation and rapid technological advancements have intensified socio-cultural pressures surrounding body ideals and performance, contributing to the increased use of Image and Performance Enhancing Drugs (IPED). Body dissatisfaction, muscularity-oriented concerns, and distorted self-perception represent key drivers of this phenomenon and are further amplified by digital media, which increasingly promote IPED use. Although IPED use has been consistently associated with body image disturbances and eating-related psychopathology, its relationship with behavioural addictions remains largely underexplored. This represents a critical knowledge gap, given that IPEDs are often used to shape the body, regulate mood, sustain productivity, sharpen cognition, and intensify sexual experiences, all of which may contribute to compulsive patterns of behaviour. This narrative review synthesises existing literature examining IPED use in relation to behavioural addictions, including exercise addiction, eating disorders, work addiction, gaming disorder, gambling disorder, and compulsive sexual behaviour disorder. Moving beyond traditional doping frameworks, it examines how IPED use is shaped by appearance-related disorders, compulsive behaviours, and social validation pressures. Particular attention is given to the role of digital environments and online fora in facilitating access to IPEDs, normalising their use, and interacting with reinforcement processes in these conditions. By integrating previously fragmented lines of research, the review proposes a novel framework that conceptualises IPED use as recurring across diverse behavioural addictions rather than an isolated phenomenon confined to sport. Understanding the mechanisms linking IPED use, behavioural addictions, and digital contexts is crucial for informing targeted public health strategies and the development of effective prevention and intervention approaches.
Patterns of smartphone use vary across ages; however, adolescents and young adults may be at particular risk, with more behavioral addictions and adverse health effects. This study explored the prevalence of smartphone addictions among health adolescent professional students and examined how problematic smartphone usage interferes with their level of physical activity as well as health-related quality of life. A cross-sectional Analytical study based on self-perceived outcome measures such as the smartphone addiction scale-short version, the 'International Physical Activity Questionnaire-short form', and 'Patient-Reported Outcomes Measurement Information System 29'-item profile was done with a sample of 400 participants. A total of 400 individuals (125 Males & 275 females) with mean age being 20.8 + 2.06 years recruited for the study. Smartphone addiction was most prevalent in dentistry students (43 %), followed by medicine (32 %) and allied health science (30.5 %), with no statistically significant differences in the addiction rate among the three programs. Compared with smartphone-addicted individuals, nonaddicted individuals had marginally greater physical function (mean difference =0.670, p<0.001), and those addicted to smartphones had significantly higher. anxiety (mean difference = 2.776, p<0.001), depression (mean difference =2.264, p< 0.001), and fatigue (mean difference =2.264, p<0.001). Physical activity was found to have no statistically significant difference between addicted and non-addicted individuals and except for sleep disturbance, none of the domains of PROMISE-29 showed any statistically significant correlation with physical activity. The findings highlight the need for recommendation for setting a time limit for the usage of smartphones for formal and informal academic activities, as well as policy measures to promote normal smartphone use.
Although the frequency of compulsory admissions (CA) to inpatient psychiatric care in Italy is low, it remains a sensitive component of psychiatric practice, together with coercion and mechanical restraint (MR). In such practices, the staff's attitude may play a role. We wanted to identify risk factors for CAs and MRs in Italian psychiatric services. A total of 18 General Hospital Psychiatric Wards (GHPWs) collected data over 12 months, reporting GHPWs' features, staff, training on CAs reduction, restrictions, protocols with Emergency, user involvement and right advocacy, number and characteristics of CAs and number of MRs. The Italian version of the Staff Attitude to Coercion Scale - SACS assessed staff attitudes. 1246 CAs (19% of all admissions) occurred in the period (median 19.2), involving 1032 individuals. No correlations between GHPW characteristics and CAs emerged. MR was used in 18% of CAs. Three GHPWs reported no MR use during the study period. MR risk was higher for males, longer CAs, personality disorder, CA confirmation in the Emergency. MRs were fewer in GHPWs with a room for meetings and no restricted access to common spaces. Total SACS score and its subscales correlated with a greater frequency of MRs. The prevalence of CAs in Italy warrants careful monitoring. Although limited acceptance of coercion among professionals, MRs remain frequent and related to contextual factors. Benchmarking would support mobilizing stakeholders' will in improving the quality of care, human rights respect and professionals' well-being in mental health settings.
This study investigates the linguistic, structural, and behavioral characteristics of "brain rot" content on TikTok, an emerging phenomenon defined by cognitively shallow, hyper-stimulating videos optimized for maximum passive engagement. The study aims to operationalize and detect "brain rot" content and to examine how such content relates to user engagement and interaction quality. Drawing on a large-scale dataset of 78,452 TikTok videos, we employed a mixed-methods design combining Natural Language Processing, regression modeling, and qualitative coding. This approach enabled the identification of key textual, structural, and behavioral markers associated with low-cognitive-value media content. The findings show that videos under 15 seconds, clickbait language, high transition frequency, and emotionally triggering content are significantly associated with higher view counts but lower comment depth. The results also indicate that verified and unverified creators contribute to the diffusion of "brain rot" content across different topic areas, suggesting that platform architecture shapes both content visibility and content quality. This study contributes a novel operational framework for identifying "brain rot" content and advances theoretical understanding of attention economics and algorithmic cultivation in short-form video environments. The findings highlight the broader implications of digital content strategies for user cognition, media literacy, and platform governance.
Cognitive reserve (CR) is a protective factor in first-episode psychosis (FEP), influencing cognitive, clinical, and functional outcomes. CR is shaped by a combination of genetic, clinical, and environmental factors, yet the extent of their respective contributions remains unclear. This study investigates the influence of polygenic risk scores (PRS), clinical and environmental variables on CR in FEP. A cohort of 174 individuals with non-affective FEP, aged 25.5 (SD=5.3), was analyzed. CR was assessed using a socio-behavioral proxy. PRS for educational attainment (PRSEA), intelligence (PRSIQ), cognitive performance (PRSCP), occupational attainment (PRSOA), physical activity (PRSPA), and schizophrenia (PRSSZ) were calculated. Age at onset, socioeconomic status, birth weight, and family history of psychosis were considered. Multiple regression models were employed to evaluate the impact of the different predictors on CR. PRSEA (p=0.002), age at onset (p=5.32x10-5), and family history of psychosis (p=0.001) emerged as the strongest contributors to CR. Higher PRSEA was associated with higher levels of CR, while earlier age at onset and positive family history were associated with lower CR. The model incorporating environmental, clinical, and genetic variables explained 17.7% of the variance in CR, and the one without PRS explained 13.5%. The inclusion of PRSEA in the model improved the explanatory power (Δadj.R2=0.042) and predictive accuracy (ΔRMSE=-0.288). These findings highlight the role of precision psychiatry in better understanding CR. Early identification of individuals with earlier onset, family history of psychosis, and lower genetic predisposition to educational attainment may help characterize those with lower CR.
The Bergen Facebook Addiction Scale (BFAS) has served as a foundation for a series of adapted measures designed to assess social media-related behavioral addictions. Despite widespread application of these instruments, a systematic evaluation of their psychometric properties is lacking. This review aimed to evaluate the measurement properties of the BFAS and its adaptations using the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) methodology. A systematic search was conducted to identify psychometric studies of the BFAS and its adapted versions, including the BFAS-18, Bergen Social Media Addiction Scale (BSMAS), Bergen Mukbang Addiction Scale, Social Media Addiction during COVID-19 Pandemic scale, and Social Networks Addiction Scale-6 Symptoms. Eligible studies were assessed using the COSMIN Risk of Bias checklist, and the quality of evidence was graded according to the COSMIN-Grading of Recommendations Assessment, Development and Evaluation approach. A total of 55 studies were included. The BFAS and BSMAS demonstrated strong evidence for structural validity, internal consistency, measurement invariance, and hypothesis testing, with high-quality evidence across multiple domains. The BFAS-18 and Bergen Mukbang Addiction Scale received more limited and inconsistent support, while the Social Media Addiction during COVID-19 Pandemic scale and Social Networks Addiction Scale-6 Symptoms remain underexplored with very low-quality evidence. Across all scales, evidence for content validity, reliability, measurement error, and responsiveness was sparse, highlighting important gaps. The BFAS and BSMAS currently represent the most robust instruments for assessing Facebook and social media addiction, respectively. However, additional research is required to strengthen evidence for other adaptations, particularly in relation to content validity, measurement error, and responsiveness, as well as to evaluate linguistic and cultural invariance. (PsycInfo Database Record (c) 2026 APA, all rights reserved).
Millions of people struggle with compulsive/problematic pornography use (i.e., PPU) or self-perceived addiction (i.e., SPA) to pornography. Despite pornography's global availability and PPU's recent official recognition as a manifestation of compulsive sexual behavior disorder, these problems remain poorly understood, primarily due to a lack of rigorous research beyond small, homogeneous samples. To address this gap and provide evidence that is both robust and generalizable, we tested the applicability of the Moral Incongruence Model of Pornography Use across diverse populations. This model posits that individuals develop problems with their pornography use due to behavioral dysregulation (i.e., PPU) and/or when their use conflicts with their moral values (i.e., moral incongruence), often influenced by religious beliefs, resulting in SPA. Using data from the International Sex Survey (N = 66,994; 50.8% women), we examined the associations between religiosity, pornography use frequency, PPU, and SPA, considering moral incongruence as a moderator. We employed multi-group structural equation models across 34 countries, three genders, and seven religious affiliations. Results indicated that the model was invariant across all countries, genders, and religious affiliations, with weak and positive associations between religiosity and PPU/SPA, and moderate-to-strong associations between pornography use frequency and PPU/SPA. Moral incongruence moderated the relationship between pornography use frequency and PPU/SPA, with stronger associations at higher levels of moral disapproval. These findings suggest that, regardless of cultural background, gender, or religion, the same mechanisms may underlie PPU/SPA, supporting the generalizability of the Moral Incongruence Model and its relevance to current international diagnostic guidelines.
The Saini-Hodgins Addiction Risk Potential of Games (SHARP-G) is a recently developed instrument designed to assess the addictive potential of video games based on their structural characteristics. While the novel instrument is highly valuable, its applicability may pose several challenges. In this work, we introduce the SHARP-G-R, a revised version featuring modifications of selected items and an overall simplification of language to improve clarity and accessibility. This updated iteration enhances the interpretability of the items, removes those with limited empirical support, and introduces new items that capture additional relevant features that may contribute to the addictive potential of games. The commentary outlines all modifications made and provides a detailed rationale for each of them. By refining the instrument, the authors aim to contribute to the discourse on the role of game design in the development and maintenance of gaming-related harms, highlighting the ethical responsibilities of the gaming industry in fostering safer gaming environments.
Problematic smartphone use (PSU) features remain debated. This study aimed to identify both parallels and distinctions between PSU and established addiction criteria (DSM-5 and ICD-11), to critically examine the appropriateness of applying these criteria to PSU, and to explore potential etiological factors. Semi-structured interviews were conducted with 28 university students from 17 European countries who scored above the established cut-off values for problematic smartphone use on the Smartphone Addiction Scale-Short Version (SAS-SV; >31 for males and >33 for females). Thematic analysis was applied to explore patterns of meaning across participants' accounts. Participants reported addiction-like symptoms, including craving, salience and preoccupation, negative consequences, loss of control, coping, and tolerance- and withdrawal-like experiences. In addition, the smartphone's vital roles in daily life, habitual use, and its role as a disruptive distraction were highlighted. A substantial proportion of participants described experiences consistent with multiple DSM-5 or ICD-11 addiction criteria; however, all participants emphasized substantial positive consequences of smartphone use, often outweighing the negative. A central "trade-off process" between positive and negative consequences emerged, alongside motives and perceived social norms, as key elements in PSU etiology. Findings suggest that directly and uncritically applying existing DSM-5 or ICD-11 addiction criteria to PSU warrants careful empirical and conceptual scrutiny. Future research should move beyond existing addiction criteria to also address the dynamic balance between benefits and harms, and the roles of motives and social norms in shaping use patterns.
Research on gender differences in self-stigma among individuals with psychosis is limited. Gender differences have been found in factors associated with self-stigma, such as self-esteem, quality of life, and cognitive insight. This study aimed to determine predictors of self-stigma reduction in men and women after receiving Metacognitive Training (MCT) or a combined intervention delivering MCT alongside cognitive rehabilitation. Randomized clinical trial with two intervention branches. Participants were randomly allocated to one of two conditions: MCT or MCT alongside cognitive rehabilitation. Assessments were carried out before and after the intervention by blinded professionals. Linear mixed models, correlation analyses, and regressions were employed to examine the effect of treatment, relationships between variables, and potential predictors of treatment response in both groups for men and women, respectively. Self-stigma was reduced in both intervention groups, with no significant differences between them, suggesting both interventions were effective. Descriptive analyses revealed women reported higher self-stigma (n = 37, M = 63.78, SD = 14.01) than men (n = 40, M = 59.40, SD = 14.52). Regression analyses suggest that post-treatment self-stigma was predicted by different predictors according to sex. In women, self-esteem (R2 = 0.25, β = -2.31, p = 0.023) was a significant predictor of post-treatment self-stigma, while in men, quality of life significantly predicted self-stigma scores post-treatment (R2 = 0.37, β = -0.62, p < 0.001). Cognitive insight was not a significant predictor for either sex. Distinct pathways of self-stigma recovery emerged for women and men, underscoring the need for gender-sensitive interventions. These findings underscore the need for tailored stigma-reduction interventions that consider gender-specific factors influencing mental health outcomes. Longitudinal research should assess durability and further clarify gender-related mechanisms. https://clinicaltrials.gov/study/NCT06423651, Registration Number NCT06423651, 23/05/24.
This commentary is written in response to the paper entitled "Taxonomy of toxic behaviors in multiplayer gaming environments: An extension of the context of peer aggression" by Zsila and Demetrovics (2025). It seeks to support and expand the discussion in two central aspects: first, the conceptual boundaries between toxicity, cyberbullying, and trolling; and second, the structure of the proposed taxonomy, acknowledging its strengths while suggesting clarifications. Conceptual overlaps with cyberbullying and trolling highlight the importance of context and situational features when distinguishing toxic behaviors. We argue that the taxonomy could benefit from refinements that explicitly consider the role of intentionality, repetition, and the interplay between in-game dynamics and mental health consequences. These adjustments may also allow inclusion of behaviors extending beyond direct in-game actions. The taxonomy presented by the authors represents a significant step forward in systematizing research on toxic behaviors in video games. Future work should integrate psychological and contextual factors that shape both the experience of toxicity and the strategies players adopt to cope with it.
Rhinitis medicamentosa (RM) refers to persistent nasal congestion and related symptoms that result from prolonged use of nasal decongestants. Re-examination of the components model of addiction as applied to RM, does not support a suggestion that RM should be conceptualised as a substance-related type of behavioural addiction. Anticipation of a highly distressing nasal congestion upon the cessation of nasal decongestants might be driving their ongoing use. Instead of unnecessarily proposing yet another behavioural addiction, it might be more appropriate and useful to posit that RM represents misuse of nasal decongestants because their prolonged administration runs contrary to medical advice.
Youth-onset diabetes mellitus (DM) is an increasing public health concern, especially among adolescents facing socioeconomic challenges. Food insecurity (FI), a key social determinant of health, has been linked to adverse metabolic outcomes, but its association with youth-onset DM and the role of preventive care remain unclear. Using repeated cross-sectional data from the National Survey of Children's Health (2016-2021), we examined the association between FI status (food secure, mild FI and moderate-to-severe FI) and youth-onset DM among adolescents aged 10-17 years. We also evaluated whether usual preventive care modified this association. Propensity score weighting (PSW) was used to reduce the imbalance of measured confounders across FI groups. These weights were applied in the least absolute shrinkage and selection operator logistic regression for variable selection and in the final multivariable logistic regression models. In the PSW analysis of 66 560 adolescents, 432 had youth-onset DM. Moderate-to-severe FI was associated with higher odds of DM than food-secure adolescents (OR 1.55, 95% CI 1.35 to 1.77; p<0.0001), whereas mild FI was not associated with youth-onset DM. Among adolescents with preventive care visits, both mild FI (OR 1.19, 95% CI 1.04 to 1.37; p=0.0126) and moderate-to-severe FI (OR 1.67, 95% CI 1.46 to 1.92; p<0.0001) were associated with higher odds of DM. Among those without preventive care visits, FI was associated with lower odds of having a reported DM diagnosis, especially in the moderate-to-severe FI group (OR 0.11, 95% CI 0.03 to 0.35; p=0.0002). Moderate-to-severe FI was associated with youth-onset DM, and this relationship varied by access to preventive care. Adolescents with FI who lack regular preventive care may have fewer opportunities for DM diagnosis. Public health strategies should integrate FI screening, equitable access to preventive services and culturally responsive approaches to chronic disease prevention.
Many countries have legalised gambling, negatively impacting public health. Stakeholders are particularly concerned about online gambling advertising harming young adults. The Netherlands has legalised offline and online gambling, but subsequently introduced restrictions on advertising. Specifically, advertisements must not target young adults aged between 18 and 23. The EU Digital Services Act requires major social media platforms to publish databases of all paid adverts shown and disclose which demographic groups were the intended targets and how many users were eventually reached. We assessed whether Dutch gambling ads on social media illegally targeted under-24s. We systematically collected recent advertisements (N = 277) published by both online and offline Dutch gambling licensees and associated demographic targeting and reach data using Meta's Ad Library. We then assessed whether the age range of users targeted by those ads included prohibited targets. The compliance rate was 92.7% for online gambling licensee ads, but only 70.2% for offline licensee ads. A significant minority of gambling ads illegally targeted Dutch young adults under 24. Whilst compliance is relatively high, building some confidence, the regulator must robustly enforce the rule against all licensees. Meta should adopt platform-level changes to reduce compliance friction. Prohibiting gambling adverts from targeting vulnerable demographics using technology is a practicable policy. Legally mandated data access can facilitate public scrutiny of policy implementation, enhancing accountability. Other countries may benefit from emulating these Dutch gambling advertising restrictions and EU advertising transparency regulations, better protecting vulnerable consumer demographics and enabling local research.
Online gambling operators collect detailed behavioural data that can identify customers at risk of harmful gambling. However, there is limited clarity on how to optimally achieve this in practice, including which variables are most useful and whether short-term data windows are sufficient for risk detection. These details are increasingly important as regulatory frameworks emphasise timely intervention. We examined the value of machine learning in this context by comparing models trained on 30 days versus six months of behavioural data and exploring whether incorporating survey responses enhanced performance. Customers from two Australian sports and race betting sites (N = 1,470) completed a survey including the Problem Gambling Severity Index (PGSI) and measures of employment, income, gambling satisfaction, and number of gambling accounts. We built machine learning models to classify participants into risk groups (PGSI 1-7 [no-to-moderate-risk] vs. PGSI≥8 [high-risk]), comparing performance across data windows (30 days vs. six months), and with or without survey variables. Models using only behavioural data achieved adequate classification accuracy (AUROC = 0.74-0.75), with similar performance across 30-day and six-month windows. The most predictive account-based variables were age, deposits per active day, average stake, and days since betting. Combining behavioural data with self-reported variables enhanced performance (AUROC = 0.76-0.85). Two self-reported variables-number of gambling accounts held and gambling satisfaction-were primarily responsible for these improvements. Machine learning models can detect at-risk customers on online sports and race betting sites using only 30 days of behavioural data. Performance can be improved by adding minimal, non-intrusive self-report measures.
Conversational artificial intelligence (AI) chatbots are increasingly used for emotional support, companionship, and psychological reflection. Their capacity to simulate attentiveness, empathy, and conversational continuity may shift technology use from instrumental interaction toward relational engagement. Emerging clinical observations suggest that intensive chatbot use may, in vulnerable individuals, contribute to psychiatric destabilization, including reinforcement of maladaptive beliefs and, in rare cases, psychotic symptoms. This viewpoint proposes that conversational AI engagement can be conceptualized within a behavioral addiction framework capable of interacting with psychosis vulnerability without requiring biological intoxication. An illustrative clinical case is presented involving a young adult woman who developed paranoid psychosis following escalating emotional reliance on a chatbot, including a fixed delusional belief that her husband was covertly communicating through the system. Potential mechanisms include persistent algorithmic attention, sleep disruption, social withdrawal, cognitive reinforcement loops, and displacement of attachment needs. Although causal relationships cannot be inferred from current evidence, systematic assessment of intensive AI engagement may become clinically relevant in behavioral addiction psychiatry and early psychosis evaluation. Further empirical research is needed to clarify prevalence, mechanisms, and clinical implications of AI-mediated relational dependence.
Gambling disorder (GD) is a growing public health concern, with emerging evidence suggesting potential sex differences in clinical presentation, gambling preferences, and treatment-seeking patterns. Furthermore, non-Western countries/cultures and women's clinical populations are understudied. Therefore, this study aimed to examine sex-related sociodemographic characteristics, gambling behaviors, and psychiatric symptoms among treatment-seeking individuals with GD in Türkiye. Data from 12,505 treatment-seeking individuals who registered for Green Crescent Counselling Center between 2021 and 2024 were analyzed. A total of 330 women participants who met the inclusion criteria were identified and compared with a sample of 990 men participants (matched on age, marital status, and education). Sociodemographic and clinical features, gambling preferences, and psychiatric symptoms were compared between groups. A notable increase in the number and proportion of women seeking-treatment for GD was observed between 2021 and 2024. Women reported beginning gambling at a later age and engaging in it for a shorter duration compared to men. They showed a distinct gambling pattern, with a strong preference for online casino games, whereas men primarily engaged in sports betting. Women also exhibited higher rates of co-occurring psychiatric concerns, including depression, anxiety, PTSD, along with greater impulsivity. Our findings highlight meaningful sex-related differences in the clinical presentation and gambling behaviors of treatment-seeking individuals with GD. The sizable representation of women suggests the need for sex-sensitive prevention and intervention strategies. Tailored approaches that address co-occurring psychiatric concerns, gambling preferences, and speculative motivations, may enhance treatment outcomes and inform future service provision in the treatment of GD, particularly for women.
Literature suggests that individual differences in tendencies toward gaming disorder (GD) may be associated with characteristics of the video games people prefer. We examined how game genre, multiplayer capability, and platform relate to GD tendencies assessed using both APA (DSM-5) and WHO (ICD-11) frameworks in a large international sample of gamers. We analyzed cross-sectional online data from 116,047 gamers. Participants completed validated measures of GD symptoms aligned with DSM-5 and ICD-11 and reported the genre, multiplayer capability, and platform of their currently preferred game. Associations were examined using bivariate and multivariate analyses. We additionally applied a Random Forest model to evaluate the predictive contribution of game-related variables to GD scores. GD levels were highest among participants preferring games with multiplayer capabilities, while positive associations with shooter and casual games were particularly evident among console users. In contrast, puzzle, platformer, and board games showed negative associations with GD scores. Players preferring desktop or laptop computers reported higher GD scores than those favoring consoles or small devices. In Random Forest models, game genre and multiplayer capability were modestly predictive of GD scores (R ≈ .08-.15) with predictive accuracy being higher for participants who used console and desktop computers and lowest among small-device users. Multiplayer capability and preferred genre were consistently related to GD tendencies across DSM-5 and ICD-11 measures. However, these variables alone offered limited predictive power, indicating that GD risk is only partly explained by game characteristics and likely depends on broader individual and contextual factors.