Engineering Adaptive Serious Games Using Machine Learning

Engineering Adaptive Serious Games Using Machine Learning

by Michael A. Miljanovic, Jeremy S. Bradbury

Abstract: The vast majority of serious games (SGs) do not feature any form of
machine learning (ML), however, there is a recent trend of developing SGs that
leverage ML to assess learners and to make automated adaptations during game play. This trend allows serious games to be personalized to the learning needs of the player and can be used to reduce frustration and increase engagement. In this chapter we will discuss the development of new ML-based SGs and present a generalized model for evolving existing SGs to use ML without needing to rebuild the game from scratch. In addition, to describing how to engineer ML-based SGs, we also highlight five common challenges encountered during our own development experiences, along with advice on how to address these challenges. Challenges discussed include: selection data for use in an ML model for SGs, choosing game elements to adapt, solving the cold start problem, determining the frequency of adaptation, and testing that an adaptive game benefits from learning.

Bibliography: Michael A. Miljanovic, Jeremy S. Bradbury. “Engineering Adaptive Serious Games Using Machine Learning.” in Software Engineering for Games in Serious Contexts – Theories, Methods, Tools, and Experiences, 2023, 17 pages.

Chapter: [PDF]