Testing the Effects of AI Access on Parent Knowledge of Child Development

How to cite

Royer, Rebecca; Pernaudet, Julie; Suskind, Dana; List, John; Levine, Dani. 2026. "Testing the Effects of AI Access on Parent Knowledge of Child Development." AEA RCT Registry. AEARCTR-0018452. https://doi.org/10.1257/rct.18452-1.0

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

Rebecca Royer* , Julie Pernaudet , Dana Suskind , John List , Dani Levine

* Primary investigator

Registry ID
AEARCTR-0018452
DOI
https://doi.org/10.1257/rct.18452-1.0
First registered
Last updated
Status
In development
Trial dates
2026-05-01 – 2026-05-22
Countries
Private
JEL codes
J13, C91, D83
Registry record
http://www.socialscienceregistry.org/trials/18452

Abstract

Caregiver knowledge of child development shapes early learning environments that drive long-term developmental outcomes. Artificial intelligence chatbots represent a potentially scalable tool for improving parent knowledge, though their effects remain untested. This pre-analysis plan describes a randomized experiment examining whether access to an AI chatbot improves parents' knowledge of early child development. We will recruit approximately 400 U.S. parents of children aged 0–6 from the online platform Prolific and randomly assign them to a treatment condition, in which they have access to an AI chatbot (Claude) while completing the Survey of Parent/Provider Expectations And Knowledge (SPEAK), or a control condition, in which they complete the SPEAK without access to the chatbot. Our primary outcome is SPEAK score. Secondary analyses will examine AI tool usage and query content as potential mediators, and will test for heterogeneity of treatment effects by parent socioeconomic status, race, gender, and AI perceptions.

Intervention

We will make public the intervention when the trial is complete.

Experimental design

We will make public the experimental design when the trial is complete.

Design details

Randomization method
We rely on the randomization feature in Qualtrics to assign participants to treatment and control groups, ensuring an approximately even split.
Randomization unit
Individual
Clusters
N/A
Observations
400 participants total.
Arms
Approximately 200 participants in treatment and 200 participants in control.
Primary outcomes
SPEAK score
Secondary outcomes
Heterogeneity by parent education, race, gender, age and AI perceptions. Mediation by quantity and quality of interactions with AI chatbot in the treatment group. See attached analysis plan for more details.

Cited by

No citation data available for this registration yet.

Keywords