Grant: CIHR Fellowship

Some people who gamble experience an addiction similar people who use alcohol or drugs. In order for us to provide the right tools to prevent gambling problems from becoming an addiction, we need a method for predicting when people are at-risk. Using machine learning models, we previously developed a successful method for identifying at-risk gamblers using their betting patterns on several gambling websites. Although these "Risk Detection Models" (RDMs) did a great job overall, we noticed that they struggled to make accurate predictions about several groups of people who have been historically disadvantaged in Canadian society: women, younger adults, and people experiencing unemployment or poverty. We aim to improve the fairness of RDM predictions so that our harm prevention efforts are not biased against people who have been marginalized in Canada. This effort will fit inside a larger research programme where our RDMs are being developed, statistically evaluated, and implemented in the real-world. Our first project will migrate our existing RDMs to a live online gambling website, revalidating and updating them with new participant samples. During this study, we will also be testing several different ways of preventing addiction in at-risk people. Crucially, we will look at how the models perform with people of different ages, genders, ethnicities, and economic backgrounds to develop methods for achieving equitable prediction accuracy and harm prevention within these groups. Our second project will involve interviewing online gamblers to understand how they feel about machine learning and our RDMs. We want to make sure that our final system is socially acceptable and does not make people from different backgrounds feel like their freedom, dignity, or privacy have been hurt. When the research is over, we will deliver a system that helps prevent gambling addictions. We will then write about our methods, so that other provinces and countries can recreate our work.