Toward a recommender system for gamification element based on user-centered design

Abstract

The use of gamification has increased significantly in recent years, it is typically applied in a generalized manner. Not all users are the same, and gamification elements cannot be used arbitrarily. First, user preferences must be identified to select the gamification elements that best suit each individual and thereby achieve better outcomes. Therefore, this investigation proposes a proof of concept for a gamification element recommender system based on User-Centered Design, using Rule-Based Filtering and Reinforcement Learning through the Contextual MultiArmed Bandit problem, which, unlike the traditional MultiArmed Bandit, specifically incorporates each user’s player profile and feedback as contextual information. A preliminary feasibility study of the system is presented using a spreadsheet, and its evaluation was conducted using prediction accuracy, ranking, predictive error, diversity, and personalization metrics. The simulation results demonstrate that, after an initial exploration period, the system converges from the third iteration onward, significantly reducing predictive error and generating highly personalized recommendations (P=0.91) that are consistent with the evolution of the hybrid profile and user preferences.