Final Report: CLOUD BASED LEARNING PLATFORM Milestone 9

The platform, Math Whiz, utilized adaptive learning technology to provide personalized content and increased practice opportunities for students. Implementation of the platform began in 2014, with the rollout of Math Whiz to two schools, one using the CBSE curriculum standard and the other using the Punjab Board standard. The key activities included the development of content mapped to grades 3-5 of the CBSE curriculum, development of the cloud-based learning platform, and development of the Math Whiz adaptive learning applications. Challenges encountered during implementation included developing content for grades 1-6, utilizing advertising as a source of revenue, and experiencing delays in rolling out the intervention at the National Public School due to unavailability of internet access. However, these challenges were addressed through experimentation with portable routers and cellular connectivity, and by focusing on content development for grades 3-5. The response from the two pilot schools was overwhelmingly positive, with students, teachers, and administrators reporting significant improvements in math skills and practice opportunities. The platform also received critical feedback from the schools, which allowed for enhancements to be made and the range of content to be extended. The final analysis of the results of the first year utilized a difference-in-differences approach, combining data from the pilot schools and the current and previous year. The sample comprised a balanced panel of students in grades 2-7 at the two schools, with the identifying assumption being that the treated students would have experienced the same trend as the control students from last year to this year in the absence of the intervention. The estimated overall impact of the treatment was 0.22 standard deviations, which is relatively large in the education literature. The effect was found to be resilient to controlling for school and grade, and the positive and economically meaningful coefficient was maintained. Additionally, the analysis found that girls may disproportionately benefit from the intervention, with the point estimate of the impact being roughly 1.5 times larger for females than males. The dynamically adaptive nature of the software was designed to promote achievement across the entire distribution of initial skill. However, it is possible that those individuals starting at the lowest levels of achievement may differentially benefit from content better suited to their current level of comprehension. The analysis found that the impact of the intervention varied by initial achievement, with the estimated effect being largest for students in the bottom tercile.