Temptation as a key driver between affective states and usage outcomes of problematic usage of the Internet: A 14-day...

Theoretical models of problematic usage of the Internet (PUI) assume that stress and negative mood can trigger cravings and desires as central drivers to use certain online applications that are expected to provide pleasure and relief, which increases the likelihood of use. Learning and reinforcement mechanisms can result in increased and uncontrolled Internet use, that manifests itself in the form of, for example, neglecting important areas of life like relationships or work. To test these associations in a natural environment, we conducted a 14-day ambulatory end-of-day assessment together with a diagnostic interview at baseline to identify different severity levels of PUI. The sample (N = 900) was part of a multi-center study (FOR2974) including participants who engaged in at least one of the following online activities: gaming, pornography use, social network use, or buying-shopping in either a non-problematic (n = 419), risky (n = 242), or pathological (n = 239) way. The adherence rate was about 88%. The data were analyzed using ANOVAs, generalized linear mixed modeling (GLMM), and multigroup multilevel structural equation modeling (MG-ML-SEM). In participants experiencing more severe PUI, ANOVAs revealed greater neglect of other areas of life (f = 0.531), longer use times (f = 0.442), and more stress (f = 0.220-0.248) as well as worse mood in daily life (f = 0.220-0.237). Overall, the variables and random effects in the GLMM explained R²conditional = 79.9% (R²marginal = 22.6%) of variance in the prediction of daily usage with temptation as primary within-person effect (OR=2.53) and PUI severity as between-person effect (OR = 0.41 to 11.33) being the main predictors. Results of the MG-ML-SEM indicate that pleasure and relief are reinforcing factors for continued use of the Internet with temptation as the central main driver (strongest effect sizes), as theoretically expected. With the extensive sample, the longitudinal approach, and the robust statistical methods, it shows that assumptions of common theoretical models for PUI can also be found in everyday life.