Probabilistic Graphical Models of Dyslexia by Yaïr LAKRETZ (‘Sagol’ school of neuroscience, Tel-Aviv University)
Reading is a complex cognitive faculty, errors in which assume diverse forms. To capture the complex structure of reading errors, we propose a novel way of analyzing these errors using probabilistic graphical models. Our study focuses on three inquiries. (a) We examine which graphical model best captures the hidden structure of reading errors. (b) We draw on the results of (a) to resolve a theoretical debate on whether dyslexia is a monolithic or heterogeneous disorder. (c) We examine whether a graphical model can diagnose dyslexia closely to how experts do. We explore three different models: an LDA-based model and two Naïve Bayes models which differ by their assumptions about the generation process of reading errors. The models are trained on a large corpus of reading errors. Our results show that the LDA-based model best captures patterns of reading errors and may therefore contribute to the understanding of dyslexia and to the diagnostic procedure. We also demonstrate that patterns of reading errors are best described by a model assuming multiple dyslexia subtypes, therefore supporting the heterogeneous approach to dyslexia. Finally, a Naïve Bayes model, which shares assumptions with diagnostic practice, best replicates labels given by clinicians and can be therefore used for automation of the diagnosis process.