Serious Games and Artificial Intelligence for the Detection of Mathematical Difficulties at School
Loading...
Date
2024
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
Mathematical attainment at the beginning of primary school is the strongest predictor of later mathematical achievement. Mathematical difficulties are assessed objectively using screening tools based on cognitive assessment for numerical processing and calculation on an individual basis under professional supervision, which is why early detection/intervention in school is difficult to implement. Our main objective was to validate a tool that combines a serious game with machine learning (ML) algorithms to perform accurate prediction for cognitive assessment of numerical processing and calculation, facilitating early detection of unsupervised mathematical difficulties at school. Following an uncontrolled open trial with a small sample size (90 children in 2nd grade of primary school) we were able to train and compare different ML algorithms with the data generated with our serious game and traditional cognitive assessments. The best fitted models for each cognitive area offered promising results, showing accuracies between 65% and 96% that combined with other good performance metrics (high recall and F1 scores for some cases) appointed to a high fidelity on diagnose. Although the results are not totally conclusive, as this was an exploratory study and more research must be done, we were able to validate the system.
Description
@inproceedings{10.2312:ceig.20241141,
booktitle = {Spanish Computer Graphics Conference (CEIG)},
editor = {Marco, Julio and Patow, Gustavo},
title = {{Serious Games and Artificial Intelligence for the Detection of Mathematical Difficulties at School}},
author = {Hornos-Arias, Josep and Serra-Grabulosa, Josep Maria and Gómez-Berengueras, Jonathan and Grau-Carrion, Sergi},
year = {2024},
publisher = {The Eurographics Association},
ISBN = {978-3-03868-261-5},
DOI = {10.2312/ceig.20241141}
}