Challenge Balancing for a Kanji E-Tutoring System

In this paper, we investigate the potential of direct challenge balancing in e-tutoring, especially in domains where there are many skills to acquire. As a case study, we create an e-tutoring system for kanji. Our system estimates the perceived challenge level using both the correctness of the answers of the students and implicit feedback, and adapts accordingly. In order to make this estimation we train a classifier on labelled data collected via the same system. We show empirically that the perceived challenge can be estimated well using implicit feedback, and that the adaptive system based on challenge balancing is preferred over a system in which the student selects a difficulty setting, indicating that direct challenge balancing is a promising research direction for e-tutoring.