Diving into the data generated during digital game-based learning: GraphoGame Flemish

Diving into the data generated during digital game-based learning: GraphoGame Flemish

First Author: Toivo Glatz -- Institut für Public Health - Charité Universitätsmedizin Berlin
Additional authors/chairs: 
Femke Vanden Bempt; Jolijn Vanderauwera; Maria Economou; Maaike Vandermosten; Jan Wouters; Pol Ghesquière
Keywords: Game-based learning, Early intervention, Early Literacy
Abstract / Summary: 

Purpose
Digital game-based learning (DGBL) has gained much attention for promoting early literacy skills and providing additional exposure to relevant reading materials in a fun and motivating environment. However, most research focuses on psychometric assessments of reading related skills and little is known about the factors that affect trial-by-trial performance within such games.
Method
Sixty-two kindergarteners (mean age 5.5 years) with an increased cognitive risk for developmental dyslexia received tablet-based literacy training with a Flemish adaptation of GraphoGame. The game trained skills such as visual and auditory discrimination, letter knowledge, phoneme counting, whole word reading and spelling. Children played 15 minutes per day for up to 12 consecutive weeks, leading to a total of 994 hours of intervention with over 273000 responses given inside the game. Using generalized additive mixed modeling we investigated intervention variables that explain in-game performance in these data.
Results
Many factors influence the accuracy and response times inside GraphoGame. Among others, variables such as the time of day, the accumulated exposure within each level, session and week, as well as level repetitions and content all explained unique variance.
Conclusions
DGBL yields very complex multidimensional in-game data and we found that comparing average accuracy and response times across studies is not meaningful. The present study helps to further improve DGBL interventions by suggesting optimal level, session and intervention durations, as well as guiding game design and content considerations. As a next step we can model individual learning trajectories to uncover potential risk factors hidden inside game data.