Reducing Frictions in Access to Childcare: A Pilot Evaluation of AI-Guided Navigation Agentic Support

Last registered on February 19, 2026

View Trial History

Pre-Trial

Trial Information

General Information

Title

Reducing Frictions in Access to Childcare: A Pilot Evaluation of AI-Guided Navigation Agentic Support

RCT ID

AEARCTR-0017883

Initial registration date

February 12, 2026

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

February 19, 2026, 7:14 AM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Primary Investigator

Name

Christopher Neilson

Affiliation

Yale University

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Seth Zimmerman

PI Affiliation

Yale University, School of Management

Contact Investigator

Additional Trial Information

Status

In development

Start date

2026-03-06

End date

2026-09-30

Keywords

Education

Additional Keywords

childcare, early childhood education, information frictions, subsidies, bilingual, AI, agent, outreach, randomized controlled trial, New Haven, Connecticut, Care 4 Kids, Head Start, Early Start CT

JEL code(s)

C93, I24, I28, J13

Secondary IDs

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

Each year, thousands of families who apply for New Haven Public Schools (NHPS) pre-Kindergarten placements are not assigned a seat through the District's centralized match process. These families must then independently navigate a fragmented landscape of alternative childcare options — including subsidy programs such as Care 4 Kids, Head Start, and Early Start CT — often under significant time pressure and with incomplete information about eligibility, application requirements, and provider availability.

This study evaluates an AI-based support system with two complementary components: (1) a bilingual (English/Spanish) childcare navigation guidance chatbot for families, and (2) a human-in-the-loop AI agent that supports outreach and case workflow. The chatbot provides personalized guidance on alternative childcare providers, subsidy eligibility, required documentation, and step-by-step application instructions. The agentic component is designed to execute workflow actions only with explicit human authorization (e.g., drafting and preparing outreach messages, generating checklists and next-step tasks, and logging follow-up actions). This distinction between guidance-only interaction and supervised agentic execution is central to the intervention.

We will randomize approximately 1,000 families that are not assigned in the NHPS school choice assignment system into two groups: a treatment group that receives access to the AI support system along with standard outreach information, and a control group that receives standard outreach information only. We compare the two groups on childcare application rates, confirmed enrollment, and self-reported barriers to access. Data sources include NHPS administrative records on applications and enrollment, anonymized usage logs, and baseline and endline surveys administered in English and Spanish.

External Link(s)

Registration Citation

Citation

Neilson, Christopher and Seth Zimmerman. 2026. "Reducing Frictions in Access to Childcare: A Pilot Evaluation of AI-Guided Navigation Agentic Support." AEA RCT Registry. February 19. https://doi.org/10.1257/rct.17883-1.0

Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Experimental Details

Interventions

Intervention(s)

The intervention is an AI-based support system provided to families on the New Haven Public Schools (NHPS) pre-K waitlist. It has two components:

(1) A bilingual (English/Spanish) childcare navigation guidance chatbot that provides tailored information on alternative childcare options, likely subsidy eligibility (Care 4 Kids, Head Start, Early Start CT), required documents, and step-by-step application guidance.

(2) A human-in-the-loop AI agent that can take supervised, parent-authorized actions on families’ behalf, including navigating program/provider websites, checking eligibility requirements, preparing application materials, submitting applications where feasible, and drafting/sending emails or required information on behalf of parents.

Control families receive standard outreach information only.

Intervention Start Date

2026-04-01

Intervention End Date

2026-05-31

Primary Outcomes

Primary Outcomes (end points)

1) Childcare applications (extensive margin): whether the family submitted ≥1 application to an alternative childcare option during the study window (binary).

2) Childcare applications (intensive margin): number of distinct childcare options/programs to which the family applied during the study window (count).

3) Subsidy program applications and access: whether the family applied to and/or obtained access to major childcare subsidy/program supports (e.g., Care 4 Kids, Head Start, Early Start CT) during the study window (binary; program-specific where feasible).

4) Confirmed childcare enrollment: whether the child was confirmed enrolled in an alternative childcare program by the primary endpoint date (binary).

Primary Outcomes (explanation)

Childcare applications (≥1) and confirmed childcare enrollment will be measured using NHPS administrative records where available and supplemented/validated with endline survey responses.

The “number of childcare options/programs applied to” outcome will be constructed as a count of distinct childcare applications reported in the endline survey and/or observed in administrative records (deduplicated by program/provider).

“Subsidy program applications and access” will be constructed as binary indicators (and, where feasible, program-specific indicators) for whether the family applied to and/or obtained access/approval for major supports (e.g., Care 4 Kids, Head Start, Early Start CT) during the study window, based on endline survey responses and any available partner/admin records.

Secondary Outcomes

Secondary Outcomes (end points)

NHPS enrollment: whether the child enrolls in a New Haven Public Schools (NHPS) school by end of September 2026 and by end of September 2027 (binary).

Secondary Outcomes (explanation)

Satisfaction with the childcare access process (scale/index) based on endline survey.

Experimental Design

This study is a two-arm randomized controlled trial among approximately 2,000 families on the New Haven Public Schools (NHPS) pre-K waitlist. Families are randomly assigned at the household level in a 1:1 ratio to treatment (standard outreach plus access to an AI-based support system) or control (standard outreach only).

The treatment combines (i) a bilingual AI guidance chatbot for childcare navigation and (ii) a human-in-the-loop AI agent that can take supervised, parent-authorized actions (e.g., navigating program/provider websites, checking eligibility requirements, preparing application materials, submitting applications where feasible, and drafting/sending required information on the parent’s behalf). Primary outcomes are measured using administrative data and surveys during the study period, with additional follow-up using NHPS administrative enrollment records through September 2027. The study is registered prospectively, and a pre-analysis plan will be uploaded prior to the start of randomized activities.

Experimental Design Details

Not available

Randomization Method

Family (household) from the NHPS pre-K waitlist.

Randomization Unit

Family (household)

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

1000 families

Sample size: planned number of observations

1000 families

Sample size (or number of clusters) by treatment arms

500 families are expected in each arm.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Planned design: individual (family/household) randomization, 1:1 allocation, no clustering. Final eligible sample size is uncertain ex ante; based on past years we expect N≈500–1,000 families total (≈250–500 per arm). For a binary primary outcome (“access” to childcare/subsidy support), the approximate minimum detectable ITT effect (difference in proportions) at 80% power and two-sided α=0.05 is: MDE ≈ (1.96+0.84)*sqrt(2 p(1-p)/n), where n is per-arm sample size and p is the control mean. Using plausible control means anchored in prior New Haven administrative outcomes: - If p≈0.54 (“any pre-k/childcare” control mean), MDE ≈ 12.5 percentage points with N=500 total (250/arm) and ≈ 8.8 pp with N=1,000 total (500/arm). - If p≈0.20 (Head Start/Care4Kids control means ≈0.20), MDE ≈ 10.1 pp with N=500 total and ≈ 7.1 pp with N=1,000 total. Unit: percentage-point change in the probability of the primary binary outcome (ITT). Standard deviation for a binary outcome is sqrt(p(1-p)) (e.g., SD≈0.50 when p≈0.5).

Supporting Documents and Materials

IRB

Institutional Review Boards (IRBs)

IRB Name

IRB Approval Date

IRB Approval Number

Analysis Plan