Stavropoulos, Vasileios ORCID: 0000-0001-6964-4662, Zarate, Daniel ORCID: 0000-0002-1508-8637, Prokofieva, Maria ORCID: 0000-0003-1974-3827, Van de Berg, Noirin, Karimi, Leila ORCID: 0000-0003-2364-504X, Alesi, Angela Gorman, Richards, Michaella, Bennet, Soula and Griffiths, Mark ORCID: 0000-0001-8880-6524 (2023) Deep learning(s) in gaming disorder through the user-avatar bond: a longitudinal study using machine learning. Journal of Behavioral Addictions. ISSN 2062-5871

Abstract

Background and aims: Gaming disorder [GD] risk has been associated with the way gamers bond with their visual representation (i.e., avatar) in the game-world. More specifically, a gamer’s relationship with their avatar has been shown to provide reliable mental health information about the user in their offline life, such as their current and prospective GD risk, if appropriately decoded. Methods: To contribute to the paucity of knowledge in this area, 565 gamers (Mage 5 29.3 years; SD 510.6) were assessed twice, six months apart, using the User-Avatar-Bond Scale (UABS) and the Gaming Disorder Test. A series of tuned and untuned artificial intelligence [AI] classifiers analysed concurrently and prospectively their responses. Results: Findings showed that AI models learned to accurately and automatically identify GD risk cases, based on gamers’ reported UABS score, age, and length of gaming involvement, both concurrently and longitudinally (i.e., six months later). Random forests outperformed all other AIs, while avatar immersion was shown to be the strongest training predictor. Conclusion: Study outcomes demonstrated that the user-avatar bond can be translated into accurate, concurrent and future GD risk predictions using trained AI classifiers. Assessment, prevention, and practice implications are discussed in the light of these findings.

Search Google Scholar

Repository staff login