Building Games to Learn from Their Players: Generating Hints in a Serious Game

Andrew Hicks, Barry Peddycord III, Tiffany Barnes

This paper presents a method for generating hints based on observed world states in a serious game. BOTS is an educational puzzle game designed to teach programming fundamentals. To incorporate intelligent feedback in the form of personalized hints, we apply data-driven hint-generation methods. This is especially challenging for games like BOTS because of the open-ended nature of the problems. By using a modified representation of player data focused on outputs rather than actions, we are able to generate hints for players who are in similar (rather than identical) states, creating hints for multiple cases without requiring expert knowledge. Our contributions in this work are twofold. Firstly, we generalize techniques from the ITS community in hint generation to an educational game. Secondly, we introduce a novel approach to modeling student states for open-ended problems, like programming in BOTS. These techniques are potentially generalizable to programming tutors for mainstream languages.

The final publication is available at Springer via https://link.springer.com/chapter/10.1007/978-3-319-07221-0_39.