DGOJ prepares risk-detection algorithm as earlier study examined family and emotion factors

Spain's DGOJ is preparing a mandatory common algorithm to detect at-risk behaviour among online gambling users, according to a report. The regulator's planned mechanism is to run during the first seven days of each month and examine a window of 183 calendar days of account activity. The system is structured in four successive phases; in the later stages more than 60 variables linked to behaviour recorded on the platform are calculated, and XGBoost, a machine-learning model based on decision trees, is used to carry out the classification.

The documentation accompanying the draft resolution sets out limits on the data operators may use. The system will draw on information generated by gambling activity on the platform itself and excludes external sources, clinical data, questionnaires, personal communications and other information the regulator considers unnecessary for the purpose. The resolution states that classification as a person with at-risk behaviour does not constitute a clinical diagnosis; it activates the protection measures established by the rules.

The report revisited an earlier publicly funded Spanish study. The project, file number 2020I007, came from the 2020 call of the Delegación del Gobierno para el Plan Nacional sobre Drogas and was led by Paula Jáuregui Bilbao of the Universidad de Deusto under the title on psychological, family and social vulnerability factors associated with gambling problems in women and the impact of COVID-19. The study by Laura Macía, Ana Estévez and Paula Jáuregui was published in the Journal of Gambling Studies.

The sample consisted of 351 women aged between 12 and 26. Of these, 312 were placed in the group without gambling risk and 39 in the group with gambling risk or problems, based on the tool used by the researchers. Participants came from educational centres and from treatment centres linked to FEJAR.

The researchers examined gambling alongside drug and alcohol consumption, spending and video games, as well as motivations for playing and emotional and family variables. Motivations included playing to cope with certain emotions or situations, playing as a social activity and seeking particular sensations or stimuli. Family attachment and the presence of positive and negative affect were also studied. Results found higher scores on drugs, spending, maternal attachment and the various motivations for playing among participants in the problem gambling group, and the weight of predictor variables changed depending on whether gambling, drugs, spending, alcohol or video games were analysed.

The study does not establish that family, alcohol, drugs or video games cause a gambling problem. Its authors refer to vulnerability factors and associations, and suggest that identifying different combinations of factors could help design prevention and treatment. The report said there is no direct technical line allowing the earlier research to be described as the origin of the DGOJ algorithm, as the projects have different objectives, data and methodologies.

The report also noted that, following a report by the Agencia Española de Protección de Datos, the DGOJ has reinforced data minimisation, the prohibition on later using the data for commercial purposes and guarantees of review and complaint for the person classified. The 2020 Plan Nacional sobre Drogas call also included another gambling-specific project at the Universitat Jaume I, led by Juana María Bretón López, on technological innovations for addressing pathological gambling, using the internet as a treatment tool in a randomised controlled study.