Chat2Learn-AI RCT

How to cite

Koechlin, Mauricio; Kalil, Ariel; Michelini, Michelle. 2026. "Chat2Learn-AI RCT." AEA RCT Registry. AEARCTR-0019133. https://doi.org/10.1257/rct.19133-1.0

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

Mauricio Koechlin* , Ariel Kalil , Michelle Michelini

* Primary investigator

Registry ID
AEARCTR-0019133
DOI
https://doi.org/10.1257/rct.19133-1.0
First registered
Last updated
Status
In development
Trial dates
2026-08-10 – 2027-12-31
Countries
Private
Secondary IDs
J-PAL: GR-14173
Registry record
http://www.socialscienceregistry.org/trials/19133

Abstract

This study is a randomized controlled trial (RCT) evaluating Chat2Learn-AI, an AI-enhanced, text-based parenting intervention that delivers personalized, illustrated conversation prompts to help low-income parents foster their children’s language skills through back-and-forth, open-ended conversation. The study will randomize a sample of N=1,000 low-income preschool and kindergarten families (children ages 3–6) to a treatment (Chat2Learn-AI) or business-as-usual control condition for a six-month intervention period. Families will be recruited through multiple partner school networks across the U.S. The primary outcome measure is child receptive vocabulary skill, and secondary outcome measures are broader child language skill (listening comprehension), child curiosity, and parental beliefs about investing in their children's language skill development. Child outcomes will be collected through direct assessment and parental beliefs will be collected through online surveys administered pre- and post-intervention period.

Intervention

Chat2Learn-AI is a text-based parenting program supplemented by large language models. Treatment families receive 3–5 researcher-designed, illustrated conversation prompts ("Chatty") per week via text, plus access to an on-demand AI chatbot ("Chatty-AI") that generates personalized, dynamic prompts based on the child's interests. The six-month intervention aims to increase back-and-forth, open-ended parent-child conversation. The business-as-usual control group receives no text-based program, only study-administration texts (e.g., opt-out procedures and survey reminders).

Experimental design

Families are randomly assigned 1:1 to the Chat2Learn-AI treatment or a business-as-usual control condition. Outcomes are assessed via direct child assessments and parent surveys at baseline and endline.

Design details

Randomization method
Randomization done in office by a computer.
Randomization unit
Individual randomization.
Clusters
N/A
Observations
1000 families
Arms
500 control, 500 treatment (C2L-AI)
Primary outcomes
Children's receptive vocabulary, measured by the Peabody Picture Vocabulary Test (PPVT-5).
Secondary outcomes
Children's listening comprehension (Woodcock-Johnson IV listening comprehension composite). Exploratory: children's expressive language skill and curiosity; parents' beliefs (growth mindset, agency, relational closeness); parental program engagement (text-back frequency).

Cited by

No citation data available for this registration yet.

Keywords