Generation of Virtual Children for testing a Recommendation System for Interventions with Children with Dyslexia.

Abstract

The LEEDUCA project has developed a recommendation system to generate intervention sessions tailored to children with dyslexia. Due to the limitations in obtaining real data for preliminary testing, the generation of in silico data, referred to as ”virtual children,” has been implemented. This approach allows for the simulation of a wide range of profiles and response patterns, enabling comprehensive testing of the system before its implementation with real users. The behavior of virtual readers is modeled using logistic curves, which reflect the natural evolution of users in a system that suggests words ordered by difficulty over time. By introducing variations to the model based on the coefficients that define the logistic curve, response sequences with different difficulty levels and learning rates can be simulated. To evaluate the stability of the system, multiple variations are generated from a given virtual child, creating a shadow of possible sequences. The generation of virtual children using logistic curves and the controlled introduction of variations in their responses provide a robust framework for testing the recommendation system, ensuring its reliability and adaptability to the individual needs of children with dyslexia.