Using Machine-Learning Algorithms to Predict Self-Reported Problem Gambling Among a Sample of Online Gamblers

The Problem Gambling Severity Index (PGSI) has been extensively used to identify the severity of problematic gambling behaviors in the general population; however, presently there is a lack of research that ensures that the PGSI measures the same latent construct in a consistent way across different socio-demographic groups (age, gender, income, education, and race) and gambling modalities (online, sports, and casino gamblers). This concept, known as measurement invariance (MI), is important as it reinforces the validity of the scale as well as survey research conclusions in the field of problem gambling. A sample of nationally representative respondents in the United States was used to test the measurement invariance of the PGSI (n = 2,972). Measurement invariance was tested using multiple group confirmatory factor analysis across the various comparison groups. Analysis supported the measurement invariance of the PGSI across demographic groups (sex, age, race, income, and education) as well as gambling modalities (online gambling, sports wagering, and casino gambling). Differences in latent means demonstrated that younger adults, sports wagerers, and online gamblers reported higher problem gambling severity. As the global gambling industry continues to grow and expand into new jurisdictions, this study has implications for both scientific and clinical use of the PGSI as an instrument to diagnose gambling disorder.

In recent years researchers have emphasized the importance of artificial intelligence (AI) algorithms as a tool to detect problem gambling online. AI algorithms require a training dataset to learn the patterns of a prespecified group. Problem gambling screens are one method for the collection of the necessary input data to train AI algorithms. The present study’s main aim was to identify the most significant behavioral patterns which predict self-reported problem gambling. In order to fulfil the aim, the study analyzed data from a sample of real-world online casino players and matched their self-report (subjective) responses concerning problem gambling with the participants’ actual (objective) gambling behavior. More specifically, the authors were given access to the raw data of 1,287 players from a European online gambling casino who answered questions on the Problem Gambling Severity Index (PGSI) between September 2021 and February 2022. Random forest and gradient boost machine algorithms were trained to predict self-reported problem gambling based on the independent variables (e.g., wagering, depositing, gambling frequency). The random forest model predicted self-reported problem gambling better than gradient boost. Moreover, problem gamblers showed a distinct pattern with respect to their gambling based on the player tracking data. More specifically, problem gamblers lost more money per gambling day, lost more money per gambling session, and deposited money more frequently per gambling session. Problem gamblers also tended to deplete their gambling accounts more frequently compared to non-problem gamblers. A subgroup of problem gamblers identified as being at greater harm (based on their response to PGSI items) showed even higher values with respect to the aforementioned gambling behaviors. The study showed that self-reported problem gambling can be predicted by AI algorithms with high accuracy based on player tracking data.

BackgroundGambling has quickly emerged as an important global public health issue. With new technologies used to develop high intensity gambling products and promotions aimed at driving consumption, public health organisations and researchers, community groups, and politicians have argued for a range of regulatory and education measures aimed at reducing gambling harm. However, there has been limited research seeking to understand community perceptions of the harms associated with gambling products and environments, and the level of community support for strategies designed to prevent and reduce gambling harm.MethodsAn online study of 500 adolescents and adults (aged 16 and over) was conducted with a representative sample (by age and gender) of individuals who were current residents in the state of Victoria, Australia. Participants were asked a range of questions about their own gambling behaviours, with the Problem Gambling Severity Index (PGSI) used as a measure of problem gambling. Participants were asked about their perceptions of harms associated with electronic gambling machines (EGMs), sports betting, horse betting, and casino gambling. They were also asked about the extent to which they agreed or disagreed with gambling harm reduction strategies related to marketing and promotions, restrictions on gambling products and venues, and public education campaigns. Quantitative data were analysed using descriptive statistics and paired t tests, with thematic analysis used to interpret qualitative responses to open-ended questionnaire items.ResultsMore than one third (n = 201, 40.2%) of participants were at risk of experiencing some level of harm from gambling (PGSI ≥ 1), with 83 participants (16.6%) recording scores that indicated problem gambling (PGSI ≥ 8). One in five participants gambled on EGMs at least monthly (n = 100, 20.0%). Those who gambled on sports did so frequently, with nearly 1 in 5 gambling on sport at least once a month (n = 87, 17.4%). Over half of the sample rated casino gambling and EGMs as very harmful, while one third rated these forms of gambling as extremely harmful. Over one third of the sample rated horse and sports betting as very harmful, with one in five rating these products as extremely harmful. There was strong agreement with the need to ban gambling advertising during children’s viewing hours, during sporting matches and at sporting venues. The majority of participants agreed with reducing and restricting EGMs, and providing more public education for both adults and children about the negative consequences from gambling.ConclusionsThe findings suggest a strong perception in the Victorian community that gambling products are harmful. While governments have been reluctant to implement a comprehensive approach to reducing gambling harm, this study reveals significant community support for a range of harm reduction and prevention measures associated with gambling products. Public health practitioners can use this evidence in advocating for a comprehensive public health approach to reducing the harms associated with gambling promotions and products.

BackgroundSports betting has become increasingly popular among Nigerian undergraduates, particularly males. Once a recreational activity, sports betting has now become an economic pursuit for many young students to augment income. This trend raises concerns about its potential impact on academic performance and mental health. This study examined the patterns, correlates, academic implications, and predictors of sports betting and problem gambling among University of Calabar undergraduate students.MethodsA cross-sectional analytical study was conducted among undergraduate students at the University of Calabar using a self-administered questionnaire. Data on gambling behaviour, problem gambling severity, reasons for gambling, and demographic information were collected using the Reasons for Gambling Questionnaire (RGQ) and the Problem Gambling Severity Index (PGSI). The impact of sports betting on participants’ academic performance was also assessed. Data was analysed using SPSS Statistics v25, employing descriptive statistics, bivariate analysis, and logistic regression.ResultsAmong 406 respondents, 36.7% had engaged in sports betting, with 30.3% participating in the past year. One underage bettor was identified. Males (83.2%) were more likely to bet. Two-thirds of respondents preferred online sports betting. The primary motivation behind sports betting was monetary gain (87.8%), followed by recreational purposes (47.1%) and thrill-seeking (45.1%). Problem gambling prevalence was 14.3%, increasing to 47.2% among those who had gambled in the previous year. Negative academic impacts were reported by 64.2% of respondents who gambled in the year prior. Strong evidence of an association between male gender and higher income with sports betting and problem gambling was identified (p < 0.05). Gambling frequency had strong evidence of association with adverse academic effects (p < 0.01). Problem gambling was associated with spending study time on sports betting (60.7%) or following sports betting games (75%) (p < 0.001). Logistic regression showed strong evidence of an association between advertisement influence and problem gambling, with those influenced by exposure to gambling adverts being four times more likely to have problem gambling (OR 4.36, 95% CI 1.93–9.86).ConclusionThe study revealed a high prevalence of sports betting and problem gambling among Nigerian undergraduates. Academic impacts were also evident, highlighting the need for targeted interventions. We recommend developing and implementing educational campaigns, financial literacy programs, and support services for at-risk individuals. Additionally, regulatory measures to restrict the advertising and marketing of sports betting to minors and vulnerable populations should be considered.Supplementary InformationThe online version contains supplementary material available at 10.1007/s44192-025-00214-5.

Technology-based gambling prevalence is not well understood since relevant questions are not included in health and disease surveillance studies. The current study sought to estimate the prevalence of internet-based and smartphone app-based gambling, along with casino gambling, in a sample of U.S. young adults and determine if gambling modality was associated with problem gambling symptoms or substance use. The 2022 Rhode Island Young Adult Survey included N = 1,022 students between the ages of 18 to 25 years old who had lived in Rhode Island, with n = 414 lifetime gamblers (40.5%) included in this study. Odds of gambling via a smartphone app and on the internet, respectively, were greater in heterosexual cis-males compared to heterosexual cis-females (OR[95%CI] = 3.14 [1.25,7.91]; OR[95%CI] = 6.30 [2.05,19.3]). Internet gambling amongst employed students was less common than among those who were not a student and not employed (OR[95%CI] = 0.25 [0.06,1.00]). Odds of problem gambling symptoms were higher among those who gambled via a smartphone app (OR[95%CI] = 3.23 [1.21,8.60]). All forms of gambling were associated with alcohol consumption, although the strength of the association was stronger in app and internet gamblers. Casino gamblers were more likely to be high risk marijuana and illicit drug users. The rising availability of app gambling coupled with its social, psychological, and cultural context may presents an alternative pathway to problem gambling. Bans on internet gambling and/or strict guidelines on the frequency of wagers and cash placed per wager, should be considered as viable methods to mitigate associated consequences.

Participating in online gambling is associated with an increased risk for experiencing gambling-related harms, driving calls for more effective, personalized harm prevention initiatives. Such initiatives depend on the development of models capable of detecting at-risk online gamblers. We aimed to determine whether machine learning algorithms can use site data to detect retrospectively at-risk online gamblers indicated by the Problem Gambling Severity Index (PGSI). Exploratory comparison of six prominent supervised machine learning methods (decision trees, random forests, K-nearest neighbours, logistic regressions, artificial neural networks and support vector machines) to predict problem gambling risk levels reported on the PGSI. Lotoquebec.com (formerly espacejeux.com), an online gambling platform operated by Loto-Québec (a provincial Crown Corporation) in Quebec, Canada. N = 9145 adults (18+) who completed the survey measure and placed at least one bet using real money on the site. Participants completed the PGSI, a self-report questionnaire with validated cut-offs denoting a moderate-to-high-risk (PGSI 5+) or high-risk (PGSI 8+) for experiencing past-year gambling-related problems. Participants agreed to release additional data about the preceding 12 months from their user accounts. Predictor variables (144) were derived from users' transactions, apparent betting behaviours, listed demographics and use of responsible gambling tools on the platform. Our best classification models (random forests) for the PGSI 5+ and 8+ outcome variables accounted for 84.33% (95% CI = 82.24-86.41) and 82.52% (95% CI = 79.96-85.08) of the total area under their receiver operating characteristic curves, respectively. The most important factors in these models included the frequency and variability of participants' betting behaviour and repeat engagement on the site. Machine learning algorithms appear to be able to classify at-risk online gamblers using data generated from their use of online gambling platforms. They may enable personalized harm prevention initiatives, but are constrained by trade-offs between their sensitivity and precision.

The prevalence of online gambling and the potential for related harm necessitate predictive models for early detection of problem gambling. The present study expands upon prior research by incorporating a cross-country approach to predict self-reported problem gambling using player-tracking data in an online casino setting. Utilizing a secondary dataset comprising 1743 British, Canadian, and Spanish online casino gamblers (39% female; mean age = 42.4 years; 27.4% scoring 8 + on the Problem Gambling Severity Index), the present study examined the association between demographic, behavioral, and monetary intensity variables with self-reported problem gambling, employing a hierarchical logistic regression model. The study also tested the efficacy of five different machine learning models to predict self-reported problem gambling among online casino gamblers from different countries. The findings indicated that behavioral variables, such as taking self-exclusions, frequent in-session monetary depositing, and account depletion, were paramount in predicting self-reported problem gambling over monetary intensity variables. The study also demonstrated that while machine learning models can effectively predict problem gambling across different countries without country-specific training data, incorporating such data improved the overall model performance. This suggests that specific behavioral patterns are universal, yet nuanced differences across countries exist that can improve prediction models.

Gambling activities are rapidly migrating online. Algorithms that effectively detect at-risk users could improve the prevention of online gambling-related harms. We sought to identify machine learning algorithms capable of detecting self-reported gambling problems using demographic and behavioral data. Online gamblers were recruited from all licensed online gambling platforms in France by the French Online Gambling Regulatory Authority (ARJEL). Participants completed the Problem Gambling Severity Index (PGSI), and these data were merged and synchronized with past-year online gambling behaviors recorded on the operators’ websites. Among all participants (N = 9,306), some users reported betting exclusively on sports (N = 1,183), horseracing (N = 1,711), or poker (N = 2,442) activities. In terms of Area Under the Receiver Operating Characteristic Curve (AUC), our algorithms showed excellent performance in classifying individuals at a moderate-to-high (PGSI 5+; AUC = 83.20%), or high (PGSI 8+; AUC = 87.70%) risk for experiencing gambling-related harms. Further, these models identified novel behavioral markers of harmful online gambling for future research. We conclude that machine learning can be used to detect online gamblers at-risk for experiencing gambling problems. Using algorithms like these, operators and regulators can develop targeted harm prevention and referral-to-treatment initiatives for at-risk users.

Establishing the temporal stability of machine learning models that detect online gambling-related harms

ABSTRACTThe aim was to examine the association of socio-demographic and lifestyle characteristics with gambling severity level. The study was a part of the National Survey on Lifestyles in Serbia: Substance Abuse and Gambling, in 2014. The sample consisted of 5385 individuals. Based on the Problem Gambling Severity Index (PGSI) score participants were divided into non-problem gambling, low-/moderate-risk gambling and problem gambling. Prevalence of problem gambling was 0.5%, which was associated with having poor self-perceived financial status, having moderate or high risk for psychological distress, playing sports betting, casino games and slot machines. Low/moderate gambling was associated with having poor self-perceived financial status, number of drinking days per year, using any illicit drugs in the last 30 days, playing sports betting, slot machines, and online betting. Programmes of early detection of problem gambling should be developed, and regulation of availability of slot machines and sports betting.

Risk factors for the development of problem gambling in individuals with ADHD symptoms: The mediating roles of gambling engagement and ADHD characteristics.

Online gambling has demonstrated a significant and growing impact on overall gambling involvement. To examine online gambling practices and associated problems, we conducted surveys with national samples of Internet users in several European countries. The objective of this paper is to provide a portrait of online gambling practices and problems in five European countries (i.e., France, Italy, Germany, Switzerland, and Poland) with different online gambling regulatory systems. This paper is the first presenting national comparisons of online gambling practices. Participants reported on their gameplay patterns, demographic characteristics, gambling-related problems, gambling frequency, average spending on different online gambling activities, and offline gambling participation over the past 12months. The Problem Gambling Severity Index was used to assess the severity of gambling problems in relation to online gambling (Ferris & Wynne, 2001 in The Canadian Problem Gambling Index: Final report, Canadian Centre on Substance Abuse, 2001). Pairwise logistic regressions examined the association between problem gambling and sociodemographic variables as well as gambling patterns and problem gambling. In all countries except Italy, the most popular online gambling activity was lotteries with nearly three quarters of participants reporting participation. The prevalence of at-risk and problem gambling was high in all countries with Poland reporting the highest prevalence and Switzerland the lowest. Gamblers who participated in gambling activities other than lottery were more likely to experience gambling-related problems. The rates in the present study are well above the levels usually found in surveys. High online gambling frequency, high spending, and participation in multiple online gambling activities were significantly associated with experiencing gambling-related problems.

Online gambling and suicide: Gambling with lives

BackgroundVoluntary self-exclusion from gambling is a common harm reduction tool in individuals with a gambling disorder. Previous data have demonstrated that many gamblers breach their own self-exclusion, typically through other online services outside the jurisdiction in which they are self-excluded. The present study aimed to carry out a new follow-up measure—similar to previous studies in the same setting—of self-exclusion and its breaching in Sweden, in order to allow for the follow-up assessment of a nationwide, multi-operator self-exclusion system introduced in Sweden in 2019.MethodsA web survey to the web panel of a market survey company addressed 1505 past-year gamblers, who responded to a number of questions about gambling habits, including screening for gambling problems using the Problem Gambling Severity Index and self-exclusion-related items corresponding to previous studies.ResultsNine percent of past-year gamblers had self-excluded using the Spelpaus service. In logistic regression, self-exclusion was significantly associated with gambling problems, past-year online casino gambling, and absence of online poker gambling. Among self-excluders, 49 percent had ever gambled despite being self-excluded. Among those breaching their self-exclusion, the most common gambling types during self-exclusion were online casino (82 percent), sports betting (47 percent) and lotteries (43 percent).DiscussionSelf-exclusion remains a popular harm reduction tool against problem gambling, more common than in previous studies, mostly in individuals with recent gambling problems and in online casino gamblers. However, breaching self-exclusion is somewhat more common than in previous research. Online casino represents the most common means of self-exclusion breaching. Policy-making in the area needs to further address the risk of breaching one’s self-exclusion and may further address the risk of overseas gambling.

Growth of Internet gambling has fuelled concerns about its contribution to gambling problems. However, most online gamblers also gamble on land-based forms, which may be the source of problems for some. Studies therefore need to identify the problematic mode of gambling (online or offline) to identify those with an online gambling problem. Identifying most problematic form of online gambling (e.g., EGMs, race betting, sports betting) would also enable a more accurate examination of gambling problems attributable to a specific online gambling form. This study pursued this approach, aiming to: (1) determine demographic, behavioral and psychological risk factors for gambling problems on online EGMs, online sports betting and online race betting; (2) compare the characteristics of problematic online gamblers on each of these online forms. An online survey of 4,594 Australian gamblers measured gambling behavior, most problematic mode and form of gambling, gambling attitudes, psychological distress, substance use, help-seeking, demographics and problem gambling status. Problem/moderate risk gamblers nominating an online mode of gambling as their most problematic, and identifying EGMs (n = 98), race betting (n = 291) or sports betting (n = 181) as their most problematic gambling form, were compared to non-problem/low risk gamblers who had gambled online on these forms in the previous 12 months (n = 64, 1145 and 1213 respectively), using bivariate analyses and then logistic regressions. Problem/moderate risk gamblers on each of these online forms were then compared. Risk factors for online EGM gambling were: more frequent play on online EGMs, substance use when gambling, and higher psychological distress. Risk factors for online sports betting were being male, younger, lower income, born outside of Australia, speaking a language other than English, more frequent sports betting, higher psychological distress, and more negative attitudes toward gambling. Risk factors for online race betting comprised being male, younger, speaking a language other than English, more frequent race betting, engaging in more gambling forms, self-reporting as semi-professional/professional gambler, illicit drug use whilst gambling, and more negative attitude toward gambling. These findings can inform improved interventions tailored to the specific characteristics of high risk gamblers on each of these online activities.