Date
AI Overview
Why it matters
Learning to read is critical for long-term success, and tutoring provides effective phonics instruction but is resource and time intensive. Leveraging technology, including artificial-intelligence reading tutors, offers one approach to expand access.
Study at a glance
- Question: How does the amount of AI reading tutor usage affect early reading growth?
- Setting and data: Four elementary schools in one Indiana district, 838 kindergarten through Grade 2 students across 43 classrooms.
- Method: Propensity weighting within schools and teacher fixed effects regression to estimate dose-response.
- AI tools: Project Read Tutor
Key findings
- High-dose users averaging at least 15 minutes weekly outperformed near-zero users on end-of-year DIBELS reading composite scores by a medium effect size.
- Each additional 10 minutes of weekly use predicted 7.1 composite percentile points, with larger gains for students with disabilities and weaker starting skills.
- For students starting below benchmark, each additional 10 minutes weekly was associated with 2.40 times the odds of reaching benchmark by year-end.
Summary drafted with AI from the paper’s abstract and may contain errors. See the full paper for details.
Learning to read is among the most consequential skills students learn and is a critical skill for long-term success. Research suggests that direct and explicit instruction in phonics skills in earlygrades increases the likelihood that students develop the foundational literacy skills they need to become proficient readers. One approach to increasing access to explicit phonics instruction is through tutoring. However, tutoring is resource and time intensive. One approach to address this limitation is to leverage technology, including artificial-intelligence (AI) reading tutors. In this study, we explored the effect of one such tutor, the Project Read Tutor. Four elementary schools in one Indiana district deployed the Project Read Tutor, and nearly all of the 838 students in kindergarten through Grade 2 were exposed to it (97%) but used it sparingly (median = 6.7 min per week). First, we compared high-dose users (15 min per week or more) with near-zero users (fewer than 2 min per week) within schools, achieving baseline equivalence through propensity weighting; this contrast showed a medium effect on the end-of-year DIBELS composite (Hedges's g = 0.47). Second, we estimated a continuous dose-response within classrooms using teacher fixed effects: each additional 10 min per week predicted 7.1 composite percentile points (g = 0.24, p = .004) across 43 classroom clusters, and the association was statistically significantly larger for students who began the year with weaker reading skills and for students with disabilities. Dose effects extended across DIBELS subtests with a near-transfer gradient: strongest for the code-focused skills the tutor teaches and roughly half the size for connected-textfluency. Among the 520 students who began the year below benchmark, each additional 10 min per week was associated with 2.40 times the odds of ending the year at or above benchmark, with the predicted probability of reaching benchmark rising from .31 at near-zero use to .68 at 30min per week. Within the schools that adopted the tutor, dose, not access, predicted reading growth
Research Categories
What is the application?
Who is the user?
Who age?
Why use AI?
Study design
AI tools
Datasets
Publication Source
Publisher
EdArXiv
