Using serious games and learning analytics for student profiling

Abstract

In recent years with the help of digital games there is an increasing interest in creating Serious Games for learning through play. With the help of machine learning algorithms, an educational serious game can be used, not only to assist the learner in his/her studies, but also help the teacher discover more about the students. In game-based learning we take into account that the student behaves differently according to his/her individual characteristics while learning by playing. The most used method to model a person’s personality is using self-report questionnaires. The drawback of this approach is that people may not assess themself correctly or their answers may be biased towards the more socially acceptable responses rather than being truthful. In this paper, we explore the idea of creating an educational serious game with the goals of helping the students to train in an introductory programming lesson and at the same time by capturing the students’ in-game actions-data with the utilization of machine learning techniques to predict their personality. A story-based game with gamified educational elements was created to help students to assess their knowledge in the programming language C. The students learn by evaluating code snippets and depending on their response the game would give constructive feedback. After the game’s end it is possible to model each student’s personality model. Particularly, for modeling the learner’s personality we used the Five-Factor Model (OCEAN), a taxonomy of five personality traits (Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism), each of which combines many personality characteristics. To evaluate the efficiency of the proposed serious game, we gathered data from 107 first year Computer Science students from the University of Macedonia. The students played the game and filled in the Big Five Inventory (BFI) questionnaire to capture their OCEAN traits. The BFI questionnaire was used as a ground truth regarding the personality of each student. After the data gathering, we used machine learning techniques and also classification algorithms to create our model. We used multiple metrics to assess the prediction of the created models. The results showed that it is effective to model both the extraversion and openness personality dimensions using serious games instead of questionnaires.