Reading and writing are considered basic skills that are necessary for people?s success in education, training and work, and vital for their overall quality of life. However, a large number of children struggle with learning how to read. It is well known that early intervention is crucial to help pupils at risk of developing reading and writing difficulties. However, there are currently no research-based, precise tools for teachers to help these pupils at the beginning of first grade. In addition, many of today?s tests can cause negative experiences and a lack of mastery in struggling readers.
The main goal of Gameplay is to develop a method for early identification of first graders at risk of developing reading and writing difficulties, while at the same time providing a positive experience for the children. The pupils will play a learning to read-game, while algorithms will look for hidden clues in the game that may indicate risk. Gameplay is probably the world?s first project to use machine learning in detecting early reading and writing difficulties.
The pupils will play the reading game daily for five weeks during the first months of first grade. Data from the game will provide detailed information about, among other things, response time, choices they make in the game, and their number of correct or incorrect answers.
In training the machine learning algorithms a large dataset is needed. Over 1500 first-graders in Oslo will participates to this end. In its last phase, Gameplay will integrate machine learning with a gaming platform that Norwegian teachers will experience as both a positive and practical addition to their instruction.
Research has established the importance of early literacy interventions to help struggling readers to overcome their reading difficulties. Early intervention for children with reading difficulties (RD) is therefore a societal priority. Early reading intervention requires early identification. However, identifying school starters who risk facing RD has remained an error-prone process.
In Norway, the national quality monitoring system has no validated tool available for the identification of struggling readers at school start. Our aim in GAMEPLAY is to develop a non-intrusive method for accurate early detection of risk for developing reading difficulties in first-grade school children. We do this through innovative use of gameplay data obtained from the child's interaction with a digital reading game. The children will play the reading game daily during a five-week period during the first months of school. The data contain detailed recordings of each game session in terms of, e.g., response times, item clicked, number of correct responses and number of incorrect responses.
Hence, through playing an enjoyable digital reading game, we obtain rich data which carries information about the child's developmental trajectories of perceptual, cognitive and linguistic skills. This data will be subjected to state-of-the-art AI/machine learning algorithms in order to detect whether the child may be at risk for developing reading difficulties.
To train and evaluate various machine learning algorithms, we will initially use high-quality data already obtained as part of the On Track research project. In order to train the more powerful AI algorithms, this dataset is too small. We therefore will collect a larger dataset, involving many more participants.
Our project will also specify how the predictive model may be integrated with the game platform for the benefit of teachers.