EEG- and VR- derived affective computing in excessive internet gamers with social anxiety

Excessive internet use may serve as an emotion regulation strategy. This retrospective cross-sectional study examined arousal and valence in students at high and low risk of internet gaming disorder (IGD) using a convolutional neural network (CNN)-based model applied to electroencephalographic (EEG) data. Sixty participants underwent virtual reality (VR)-based socioemotional stress and were classified into high-risk (HIGD) and low-risk (LIGD) groups based on IGD test scores. Depression, social anxiety, and emotion regulation were assessed using self-report questionnaires. Between-group differences were examined using parametric tests and linear mixed-effects models; associations between EEG-derived indices and self-report measures were evaluated using Spearman correlations with false discovery rate correction. The HIGD group reported significantly greater depression and social anxiety but comparable emotion regulation strategies. CNN-derived arousal and valence did not differ between groups across five conditions, including pre- and post-VR resting state and three scenarios (all p > .05). However, a significant group-by-time interaction in valence emerged during criticism of academic performance (p = .002). No correlations between EEG-derived indices and self-report measures remained significant across the five stages (all adjusted p ≥ .43). These findings suggest that IGD risk may be associated with a more negative valence trajectory during performance-related criticism rather than generalised differences in arousal or valence.