Chat2Learn: Comparing Static vs. AI-Enhanced Text-Based Conversational Prompts

How to cite

Bresciani, Daniela; Kalil, Ariel; Mayer, Susan. 2025. "Chat2Learn: Comparing Static vs. AI-Enhanced Text-Based Conversational Prompts." AEA RCT Registry. AEARCTR-0015280. https://doi.org/10.1257/rct.15280-1.0

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

Daniela Bresciani* , Ariel Kalil , Susan Mayer

* Primary investigator

Registry ID
AEARCTR-0015280
DOI
https://doi.org/10.1257/rct.15280-1.0
First registered
Last updated
Status
In development
Trial dates
2025-03-03 – 2025-12-19
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/15280

Abstract

Previous research has shown that Chat2Learn (C2L), a text-message based program delivering conversational prompts for parents to use with their children, can positively impact children's vocabulary. This study evaluates two different modalities of C2L, both delivered via text message. In the first modality, the static version, parents only receive predetermined prompts. In the second, the AI modality, parents receive the same prompts along with unlimited AI-generated responses. We will conduct a randomized controlled trial with approximately 1,000 parents of children aged 3 to 6 across the United States. Parents will be randomly assigned to either the static or AI modality of C2L. The study will assess which modality generates more engagement from parents, as measured by opt-out and response rates.

Intervention

Chat2Learn (C2L) is a text-message based program that delivers conversational prompts for parents to use with their children for 9 months. We will offer two versions of C2L to parents (parents will be randomly assigned to receive one of them):
- Static version: parents receive predetermined conversational prompts via text message.
- AI version: parents receive the same predetermined conversational prompts via text message. Additionally, parents can respond to the prompts or message the phone number and get unlimited AI-generated responses and prompts.

Experimental design

Randomized controlled trial: parents will be individually randomized, half to the static version of the program and the other half to the AI version of the program.

Design details

Randomization method
Randomization done in office by a computer
Randomization unit
Individual
Clusters
1000 parents
Observations
1000 parents
Arms
500 parents per arm
Primary outcomes
1) Our first outcome will be a binary variable indicating if the parent opted out of the program before it was over. 2) Our second outcome will be, out of the 121 days in which the parents will receive a pre-programed prompt from C2L, the number of days in which the parent sends a message to C2L. 3) The third outcome will be the total number of days in which the parent sends a message to C2L, counting both the days when they received and the days when they didn’t receive pre-programed messages. 4) Finally, we will analyze the messages that parents send to C2L, such as the word count and the complexity of the message (which will be measured with the help of AI).

Cited by

No citation data available for this registration yet.

Keywords