Gen AI Platform for Teachers

How to cite

Goulas, Sofoklis; Megalokonomou, Rigissa; Sotirakopoulos, Panagiotis. 2025. "Gen AI Platform for Teachers." AEA RCT Registry. AEARCTR-0017487. https://doi.org/10.1257/rct.17487-1.0

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

Sofoklis Goulas* , Rigissa Megalokonomou , Panagiotis Sotirakopoulos

* Primary investigator

Registry ID
AEARCTR-0017487
DOI
https://doi.org/10.1257/rct.17487-1.0
First registered
Last updated
Status
In development
Trial dates
2025-12-19 – 2026-06-15
Countries
Private
JEL codes
I21, I28, C93, O33
Registry record
http://www.socialscienceregistry.org/trials/17487

Abstract

This study evaluates how access to and training in generative artificial intelligence (gen-AI) tools affect teachers’ instructional practices, beliefs, and task allocation. All participating teachers are provided access to a gen-AI platform designed to support core teaching tasks such as lesson planning, assessment design, feedback generation, and administrative work. Teachers are then randomly assigned to one of two groups:
1. The treatment group receives additional informational and behavioral nudges designed to promote effective and responsible gen-AI use, including guidance on prompt formulation, examples of high-quality use cases, and reminders emphasizing human oversight and pedagogical alignment.
2. The control group receives access to the platform but does not receive these nudges.

The experimental design combines randomized variation in training nudges with detailed usage data captured directly through the gen-AI platform. Primary outcomes include measures of gen-AI adoption and intensity of use, task-specific usage patterns, and characteristics of generated content. Secondary outcomes include teachers’ beliefs about gen-AI, perceived usefulness and risks, confidence in using AI-assisted tools, and self-reported changes in instructional practices. These outcomes are measured using baseline, intermediate, and endline questionnaires administered over the course of the intervention, and are linked to high-frequency platform-level data.

By integrating randomized nudges with behavioral usage data, this study provides causal evidence on how targeted training interventions shape teachers’ engagement with gen-AI technologies and the mechanisms through which AI tools are incorporated into educational work.

Intervention

The intervention introduces teachers to the use of generative artificial intelligence (gen-AI) for core instructional and administrative tasks. All participating teachers receive access to a gen-AI platform that can assist with lesson planning, assessment design, individualized feedback to students, and routine paperwork.

Teachers are individually randomly assigned to treatment and control conditions. Teachers in the control group receive platform access only. Teachers in the treatment group receive platform access plus targeted informational and behavioral supports intended to foster effective and responsible use of gen-AI. These supports include: (i) guidance on constructing effective prompts, (ii) curated examples of high-quality classroom use cases, and (iii) periodic reminders emphasizing appropriate human oversight, data privacy, and alignment with pedagogical goals.

The intervention is delivered remotely and can be accessed asynchronously by teachers over the study period.

Experimental design

This study is a randomized controlled trial designed to evaluate how access to and training in generative artificial intelligence (gen-AI) tools affect teachers’ use of AI in instructional tasks and related beliefs. All participating teachers are granted access to a gen-AI platform that supports a range of teaching activities, including lesson planning, assessment design, feedback generation, and administrative tasks.

Following enrollment and completion of a baseline questionnaire, teachers are randomly assigned at the individual level to either a control group or one of the treatment groups.

Control group: receives access to the gen-AI platform only, with no additional training or nudges beyond basic onboarding.

Treatment group(s): receive platform access plus informational and behavioral nudges that vary in type and intensity. These nudges are designed to promote effective and responsible use of gen-AI tools and may include guidance on prompt formulation, examples of high-quality use cases, reminders emphasizing human oversight, and recommendations for aligning AI-generated content with pedagogical objectives.

The intervention period is followed by intermediate and endline surveys to measure changes in beliefs, confidence, and self-reported instructional practices. Throughout the study, detailed platform-level data are collected, allowing observation of teachers’ AI usage behavior at high frequency. Survey data are linked to platform usage data using unique, anonymized identifiers. The experimental design enables estimation of the causal effects of gen-AI training nudges on adoption, intensity, and patterns of AI use, as well as on teachers’ perceptions of and engagement with AI tools.

Design details

Randomization method
Random assignment is implemented server-side within the gen-AI platform at the time of first login (or immediately after baseline completion). Teachers are assigned at the individual teacher level to one of the experimental arms using a pre-generated random assignment list (or an automated randomization algorithm) with equal allocation probabilities (or pre-specified unequal probabilities). The platform stores the assigned arm and delivers the corresponding nudges/training content. Survey responses (baseline/intermediate/endline) are linked to platform assignment via a unique anonymized participant identifier.
Randomization unit
Individual teacher (participant-level)
Clusters
Not applicable
Observations
1,200 teachers (approx.)
Arms
2 arms (1:1): ~600 per arm
Primary outcomes
The primary outcomes focus on teachers’ adoption and use of generative AI tools, measured using objective platform-level data. These outcomes include: Gen-AI adoption, defined as whether a teacher uses the gen-AI platform at least once during the intervention period. Intensity of gen-AI use, measured by the frequency of platform interactions (e.g., number of prompts submitted, sessions initiated) over the study period. Task-specific usage patterns, capturing how gen-AI is used across different teaching tasks, such as lesson planning, assessment design, feedback generation, and administrative activities. Characteristics of AI-generated content, including indicators of prompt complexity, degree of revision or iteration, and use of guidance features emphasized in the training nudges. These outcomes are pre-specified to capture both extensive-margin (whether teachers use gen-AI) and intensive-margin (how and for what tasks it is used) responses to the experimental interventions.

Cited by

No citation data available for this registration yet.

Keywords