The use of machine learning to improve the identification and assessment of internet-related disorders

The Internet's growing significance has raised global concerns about Internet-related disorders. Organizations like the American Psychological Association (APA) and the World Health Organization (WHO) have already highlighted the potential negative effects of excessive Internet use on mental health. Since the inclusion of gaming disorder as a condition for further study in the DSM-5 and its recognition as
a mental disorder in ICD-11, research on the problematic use of the Internet (PUI) become a topic of even greater significance.
The present PhD thesis aims to address two key research priorities in the field of PUI, formulated by the European Network for PUI, related to (a) contributing to their conceptualization and (b) improving their assessment. In this regard, four different studies targeting gaming disorder and cyberchondria, a condition characterized by excessive and uncontrollable searching for health-related information on the
Internet, were deployed. This thesis centrally focuses on using machine learning (ML) and traditional statistics to reach these objectives.
In Study 1, the levels of cyberchondria during the pandemic were investigated and compared with the retrospectively assessed pre-pandemic levels. It also identified psychological factors that could predict the level of cyberchondria during the pandemic. In Study 2, different gamer groups based on their profiles
of passion for gaming were identified. It also observed how gaming disorder symptoms, assessed within the substance use disorder and gambling frameworks (e.g., tolerance, withdrawal, preoccupation, mood modification), are linked to harmonious and/or an obsessive passion for gaming. Study 3 used gaming
disorder criteria to predict depression and well-being levels. It also identified predictors of gaming disorder level and their importance in the prediction of each DSM-5 criterion proposed for Internet gaming disorder. Finally, Study 4 warns against the misuse of algorithm-generated data in ML analyses and its negative impact on the conceptualization and assessment of a PUI.
Results from the studies suggest that cyberchondria and gaming disorder can be understood within the same general framework. Nevertheless, additional models specific to each condition can enhance their understanding and provide important insights for their treatment and prevention interventions. Regarding their assessment, the thesis supports the idea of a possible transdiagnostic nature of the criteria proposed by the ICD-11 for the assessment of gaming disorder and their potential capacity to address the various forms of PUI. The thesis also demonstrates that ML methodologies offer a helpful and convenient instrument for psychological research topics such as the PUI.