SARA: AI-Assisted Self-Applied Tool for Reading Screening in Dominican Schools

• In Event: AERA Poster Session Sunday 1:45 pm
  In Poster Session: Systemic, Critical, and Applied Perspectives on Technology Integration

Abstract

SARA is a self-applied, AI-assisted reading assessment designed for students in the Dominican Republic. The tool integrates a culturally adapted test battery (ECLEC) that evaluates decoding, fluency, and comprehension. SARA was administered to 1,861 students in grades 2–6 across public and private schools. Psychometric analyses confirmed strong internal consistency and construct validity, supported by structural modeling and convergent correlations. AI models scored oral reading tasks with up to 86% agreement with human raters. Norms and screening cutoffs were established by grade and school type, enabling early identification of students at risk for reading delays. SARA demonstrates the feasibility of equity-oriented, scalable screening tools for reading development in low-resource, linguistically diverse educational contexts.

Authors

• Daniel Cubilla-Bonnetier, Universidad Iberoamericana

• Enzo Casamassima, Universidad Iberoamericana

• Laura Virginia Sánchez-Vincitore, Universidad Iberoamericana