Automating the Stereotype student model

Abstract

The stereotype student model classifies students into groups according to their frequent characteristics [7]. Although the approach of stereotyp- ing has matured and many interesting possibilities have been explored, a number of problems with this model have remained unsolved. The difficulties lie in implementing a particular stereotype model. The most apparent and widely used approach is having system designers build the stereotypes for a particular learning environment, an error-prone and time-consuming task [17][33]. Moreover, such an approach does not pro- duce an ontology or abstract data structure which can subsequently be used for other learning environments. To tackle these interconnected problems, this paper presents a software design containing two critical components. Firstly, we use the ‘Decision Tree Regression’ machine-learning algorithm to automate creating and updating stereotypes. We use this algorithm because its output resembles stereotypes and contains all necessary data to describe them. Secondly, our design structures the data flow from a learning environment to our algorithm. In this paper, the learning environment is the ‘BattleQuiz’ game platform with content about construction safety. Our algorithm outputs semantically meaningful stereotypes, which we send to a dashboard. The dashboard and the information it contains is evaluated by educational psychologists. By using a rather basic learning environment, we have limited the complexity and expressiveness of our data and subsequently of our stereo- types. Applying our design to a richer learning environment would be an interesting continuation of research on the approach introduced in this paper.