Detecting Gambling Addiction: How Text Mining of Customer–Operator Online Chats Reveals Behavioral Risks

Gambling disorder is a growing public health concern, exacerbated by the rapid expansion of online gambling. Prevention is key to supporting individuals from developing or exacerbating risky conducts. This study presents a multidisciplinary project across psychology and computational linguistics. The project aimed at developing a predictive system for detecting linguistic markers of problematic gambling behaviours in customer–operator chats from Sisal, a major Italian gambling company. The objective is to support staff in promptly recognizing at-risk individuals and providing timely assistance. The research was conducted in two phases. First, a qualitative analysis of Italian conversations on the online support centre chat identified linguistic and discursive markers of risk. These insights informed the training of a supervised binary classification model. In the second phase, the system was refined through dataset expansion, manual coding carried out by psychologists, and the introduction of a four-level risk classification. Markers were grouped into macro-categories and weighted according to emotional intensity and frequency before being integrated into the predictive model. Findings showed that combining linguistic analysis with psychological expertise can effectively enhance the early detection of problematic gambling behaviours. The research highlighted the potential of text mining as a tool for prevention and person-centred interventions in gambling disorder.