Do artificial neural networks dream of understanding sentence comprehension?A preliminary study

Abstract

Artificial neural networks (ANNs) are an emerging field with a positive and encouraging outlook. In education, it is postulated that attention and academic performance could explain reading outcomes. The main goal of this research was to study the predictive capacity of an ANN with a backpropagation algorithm by analysing the relationship between sentence and text reading comprehension efficiency, attentional variables and academic performance in third-grade primary school students (N = 183). A non-experimental approach was adopted, using a cross-sectional and ex post facto design. Ten schools (70% public) located in southeastern Spain participated. Test of Reading Efficacy (TECLE), d2 attention test and TALE-2000 were administered. The results revealed that it is possible to design a network capable of learning by itself to predict sentence comprehension. Students who were good readers obtained better grades, concentrated better, scanned the stimulus more attentively, obtained more correct answers and made fewer omissions. The conclusions concerned the ethical implications of AI and the need to introduce ANNs in initial teacher training.