Abstract

Educators recognize the need to provide users with context-appropriate challenges. Despite this belief, online learning games typically provide uninformed adaptive selection of learning tasks. Instead, we present our design of a quantified flow-channel metric, a ratio of user skill over problem difficulty. By mining and clustering historical play data from brainrush.com, a crowdsourced online learning platform, we weigh distractors for each learning objective. Our algorithm adaptively constructs questions based on our metric to maintain Flow. Our approach provides in-game adaptation to maintain the user's Flow experience and is applicable to a wide range of learning tasks.