Resumen de Game learning analytics for understanding reading skills in transparent writing system
Marko Niemelä, Tommi Kärkkäinen, Sami Äyrämö, Miia Ronimus, Ulla Richardson, Heikki Lyytinen
Abstract Serious games are designed to improve learning instead of providing only entertainment. Serious games analytics can be used for understanding and enhancing the quality of learning with serious games. One challenge in developing the computerized support for learning is that learning of skills varies between players. Appropriate algorithms are needed for analyzing the performance of individual players. This paper presents a novel clustering-based profiling method for analyzing serious games learners. GraphoLearn, a game for training connections between the speech sounds and letters, serves as the game-based learning environment. The proposed clustering method was designed to group the learners into profiles based on game log data. The obtained profiles were statistically analyzed. For instance, the results revealed one profile consisting of 136 players who had difficulties with connecting most of the target sounds and letters, whereas learners in the other profiles typically had difficulties with specific sound-letter pairs. The results suggest that this profiling method can be useful for identifying children with a risk of reading disability and the proposed approach is a promising new method for analyzing serious game log data. Practitioner Notes What is already known about this topic Serious games are used to improve learning and to tailor learning environments for people with various difficulties in learning. Learning analytics and serious games analytics are growing research fields, applying and developing data analysis methods to analyze, profile and understand learning using serious games. GraphoLearn is a learning game for training reading skills. The game provides preventive support for learners with varying skill levels including individuals who are struggling with reading. What this paper adds The paper develops and presents a novel approach for serious games analytics to analyze GraphoLearn players. The proposed data analysis approach produces an interpretable set of error profiles, which characterize the learning difficulties in a unique way. The profiling method can be used for longitudinal studies and applied to analyzing logs of other serious games. Implications for practice and/or policy It is possible to reveal and understand profiles of serious game players. The proposed data analysis method can be used to identify players who have a potential risk for reading difficulties or disabilities. Even though the proposed method provides only limited information about players' future skills, it offers a good starting point for other studies in which players' development can be monitored more accurately.
