Machine learning is an area of artificial intelligence (AI) that focuses on how models can learn through experience with data. These models show promise for promoting safer gambling in online gambling. However, whether these models can treat people fairly remains unclear. The goal of this study was to explore the performance and fairness of three models for identifying harmful gambling.
The data came from surveys of people gambling online as well as account data from a provincial-owned gambling website in Canada. The researchers found that, if the goal was to identify as many people with high-risk gambling as possible, then the ‘classification parity model’ performed the best. Yet, none of the models could be considered truly fair across all the performance metrics that were measured. The researchers suggest that models tested on-site with large samples could reach a higher degree of fairness.
