Adaptive versus Non-adaptive EdTech: Evidence from a Large-Scale RCT in China

How to cite

Loyalka, Prashant. 2021. "Adaptive versus Non-adaptive EdTech: Evidence from a Large-Scale RCT in China." AEA RCT Registry. AEARCTR-0007888. https://doi.org/10.1257/rct.7888-1.2

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Prashant Loyalka*

* Primary investigator

Registry ID
AEARCTR-0007888
DOI
https://doi.org/10.1257/rct.7888-1.2
First registered
Last updated
Status
On going
Trial dates
2021-03-01 – 2025-08-31
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/7888

Abstract

The purpose of this study is to examine whether the adaptive learning component of one type of widely used EdTech – computer-assisted learning (CAL) – is effective at improving student achievement in developing countries. We have three specific objectives. First, we seek to determine whether adaptive CAL (versus regular non-adaptive CAL and versus a pure control group with no CAL) can increase average levels of student achievement. Second, we seek to determine whether adaptive CAL can increase student achievement among different subgroups (primarily for students at different by achievement levels—thus examining whether adaptive CAL reduces achievement gaps—but also by family wealth and gender). Third, we seek to determine whether adaptive CAL can improve students’ secondary outcomes such as attitudes towards learning.

Experimental design

We conduct a large-scale randomized experiment in 105 schools (315 school-grades or classes: grades 4-6, 1 class per school-grade) with 8647 students in rural China. The randomized experiment follows the following steps. First, we randomly sample rural primary schools in one region of China and randomly sample one fourth, fifth, and sixth grade class from each school. Second, we randomize the 315 classes (within grades) to one of three equally-sized treatment arms: Treatment arm #1 receives the adaptive learning software; Treatment arm #2 receives the non-adaptive learning software; Treatment arm #3 is a pure control with no learning software.

Design details

Randomization method
randomization done in office by computer
Randomization unit
classes or school-grades (one class per grade per school); randomized within strata (groups of 6 school-grades with closest average baseline math achievement scores)
Clusters
315 clusters
Observations
approximately 8500 students
Arms
approximately 2800 students and 105 classes (school-grades) per arm
Primary outcomes
math achievement at endline

Cited by

No citation data available for this registration yet.

Keywords