Diagnosing developmental dyslexia using serious computer games
Jens van der Meer and Jennifer Spenader

Intervention for children with developmental dyslexia at an early stage is important since reading and spelling is an assumed skill in modern society. Currently, a diagnostic test is only ordered after parents or teachers suspect dyslexia. But by using serious games and clustering techniques we may be able to diagnose dyslexia earlier, and more efficiently.

In a large project involving hundreds of Dutch children between the ages of 7-12, we are collecting data related to reading and spelling using serious games. This project has two major innovations. First, our games are designed to identify fundamentally different types of dyslexia. Existing computational tests have not made any attempt to distinguish between different dyslexia types and we expect that this will improve our results. Second, in addition to a standard statistical analysis, we will also use unsupervised machine learning (e.g. clustering) to identify subgroups of children by their systematic error types. For this type of analysis to be successful we need to have access to a large number of participants, which has been organized.

Multiple specially designed games will be deployed on an online platform already used by well over a hundred primary schools. The scores and reaction times for each individual item (i.e. word) will be used for a within-subjects comparison. Specific features will be automatically extracted from each item. Preliminary results are expected to be available to present.