Diagnosing risk of Dyslexia Based on the analysis of mouse tracking data and human computer interaction measures

Abstract

Our study presents the 'Direction' game as a tool for assessing the risk of reading difficulties in young children. This game evaluates directional confusion in three aspects: up-down, left-right, and cardinal directions, which is advantageous for pre-literate children as it does not rely on language skills. In this research, we conducted tests with 102 children, both boys and girls, aged between 7 and 10 years, studying at Ban Pong Sanuk School in Lampang Province. The children were divided into two groups: 59 children in the at-risk group and 43 children without identified risks. In the experiment, the experiment was divided into 2 periods: 1) pilot study t with a sample size of 24 people by collecting a questionnaire to bring the paper-based game to children to try and ask about their satisfaction in playing the game and 2) the main study test. The game was tested on a sample size of 102 individuals and the results were analyzed using a component classification model. The performance of this proposed model will be compared with other models including Naive Bayes, Support Vector Machine (SVM), AdaBoost, and Random Trees (RT) to assess the performance of the new model. The results of the experiment revealed that children were able to play our game without needing someone to teach them. They felt more like they were playing a game rather than being tested or evaluated and did not feel isolated from their peers. 85% of the children who were tested reported this experience. Additionally, the children enjoyed playing our game, with 90% of the tested children expressing a liking for it. Regarding the evaluation of reading disability risk among the sample group of 102 children, the implemented framework and procedures proved to be effective in improving the accuracy of risk assessment. In particular, the combined ensemble model of Random Trees (RT) and Support Vector Machine (SVM) demonstrated a screening efficiency of up to 86% and an area under the ROC curve of 82%. These results indicate that the model is proficient in assessing the risk of reading disabilities and enhances the screening performance. Additionally, incorporating mouse movement variables, such as fixation duration and the number of duplicate points, increased the accuracy of the assessment to over 70%. These variables are crucial in detecting deficiencies in organizational skills and attention in children at risk of reading difficulties. In analyzing the behavior of children in at-risk and normal groups while playing the 'Direction' game, it was found that children in the at-risk group were more prone to making mistakes due to hesitation and difficulties in following instructions. However, their mouse movement speed was similar to that of the normal group, indicating that their overall game-playing abilities were not significantly different. Despite this, children in the normal group generally scored better.