To effectively grasp and apply key machine learning concepts, students benefit from learning environments that actively support structured problem-solving strategies. Prior research indicates that learners often rely on exploratory approaches when constructing decision trees, with limited use of systematic problem-solving strategies. This paper examines the design and implementation of collaborative features in a digital learning game focused on developing decision trees as classification models. Building on earlier findings, a multiplayer mode was developed to enhance social interaction through targeted game mechanics, with the aim of fostering the development and persistence of structured problem-solving. Results from an initial pilot study suggest that the multiplayer mode facilitates strategic peer discourse, promoting both collaborative engagement and more systematic problem-solving behavior. These findings highlight the potential of collaborative, game-based learning environments to support effective strategy use in machine learning education.