Dytective: Towards Detecting Dyslexia Across Languages Using an Online Game
Luz Rello, Kristin Williams, Abdullah Ali, Nancy Cushen White, Jeffrey P. Bigham · 2016 · Proceedings of the 13th International Web for All Conference (W4A) · doi:10.1145/2899475.2899491
Summary
This paper presents Dytective, a browser-based game designed to screen for risk of dyslexia across English and Spanish languages. At least 10% of the global population has dyslexia, and in Spain over 40% of school dropout is attributed to it. Current detection methods are language-specific, expensive, require professionals or specialised equipment, and do not scale well to classroom or home settings where early identification is most needed. A central challenge is that dyslexia manifests differently across languages due to orthographic depth — English has inconsistent grapheme-phoneme mappings (deep orthography) making reading and writing better predictors, while Spanish has more consistent mappings (shallow orthography) where reading speed and fluency predominate. Dytective addresses this by designing linguistically motivated exercises informed by analysis of common written errors made by people with dyslexia in both languages. The game targets 16 cognitive skills associated with dyslexia: orthographic processing, phonological awareness, reading speed, phonological memory, phoneme and syllable segmentation, word and non-word recognition, syntactic and semantic awareness, error detection and correction, written spelling, working memory, visual memory, and visual attention. Each stage presents timed linguistic tasks (25-second windows) where players solve problems using visual and auditory-visual exercises — for example, identifying different letters in a grid, or clicking on occurrences of a heard non-word. The game is built in HTML5, CSS, JavaScript with a PHP backend, designed for cross-platform browser access.
Key findings
A user study with 60 children aged 7-12 (30 English speakers, 30 Spanish speakers, half with confirmed dyslexia diagnoses in each group) showed that Dytective significantly differentiated children with and without dyslexia. Using Wilcoxon Signed Rank tests, significant differences were found for most dependent measures in both languages. For Spanish speakers, children with dyslexia had significantly fewer clicks (M=3 vs M=5, p<0.001), fewer hits (M=2 vs M=3, p<0.001), lower scores (M=2 vs M=3, p<0.001), lower accuracy (M=0.67 vs M=0.80, p=0.001), and higher miss rates (M=0.14 vs M=0.04, p=0.001). For English speakers, significant differences were found for hits (p=0.001), score (p=0.002), accuracy (p=0.040), and miss rate (p<0.001). An earlier Spanish-only version tested with 243 participants achieved 85.85% prediction accuracy using machine learning. The results demonstrate that a single game framework can detect dyslexia risk across languages with different orthographic depths.
Relevance
This research addresses a critical accessibility gap: the underdiagnosis of dyslexia, particularly in communities without access to expensive professional assessment. By using a free, browser-based game that works across languages, Dytective could enable universal screening in schools and homes worldwide. Early detection is crucial — providing reading support early in acquisition has been shown to reduce reading failure incidence from 18% to 5%. The cross-linguistic approach is particularly significant because most existing screeners are language-specific, limiting their global applicability. For accessibility practitioners, this work highlights that dyslexia affects how people interact with all text-based digital content, not just printed materials. Understanding dyslexia's prevalence and the importance of early detection reinforces the need for reading accessibility features (adjustable fonts, text spacing, text-to-speech) in web design. The game-based approach also demonstrates how engaging interface design can make assessment itself more accessible and less stressful for children.
Tags: dyslexia · screening · learning disabilities · game-based assessment · multilingual · reading accessibility · early detection · phonological awareness
Standards referenced: DSM-V