Rapid Evaluation of the Impact of Gemini Guided Learning on Students’ Mathematics Scores in Junior Secondary Schools in Sierra Leone

Last registered on September 19, 2025

Pre-Trial

General Information

Title

RCT ID

AEARCTR-0016651

Initial registration date

September 18, 2025

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

September 19, 2025, 10:20 AM EDT

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

Locations

Country

Sierra Leone

Region

Africa

Primary Investigator

Name

Natalia Valdes Aspillaga

Affiliation

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

María José Ogando Portela

PI Affiliation

Fab Inc

Contact Investigator

PI Name

Usman Khawar

PI Affiliation

Fab Inc

Contact Investigator

Additional Trial Information

Status

In development

Start date

2025-09-22

End date

2026-01-31

Keywords

Education

Additional Keywords

Artificial Intelligence, EdTech, AI Tutoring

JEL code(s)

I24

Secondary IDs

Prior work

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

Abstract

This study evaluates the impact of Gemini Guided Learning, an AI-powered instructional experience within Gemini that provides personalised learning through guided questions, step-by-step scaffolding, and multimodal resources rather than supplying direct answers, on mathematics learning in Junior Secondary Schools (JSS1 and JSS2, ages 13–15). The trial randomly assigns classrooms either to continue with regular teaching or to use Gemini in approximately half of their mathematics lessons over an eight to nine (8-9) week period. The same teachers are responsible for both treatment and control classrooms; all will receive training on how to use the tool, though it will only be applied in treatment classes. Mathematics performance of students and teachers will be assessed at the beginning and end of the intervention. Teachers’ experiences with the tool will also be documented. The study will provide rigorous evidence on whether AI-supported guided learning can strengthen mathematics learning in junior secondary education.

External Link(s)

Introduction of Gemini Guided Learning

Registration Citation

Citation

Khawar, Usman, María José Ogando Portela and Natalia Valdes Aspillaga. 2025. "Rapid Evaluation of the Impact of Gemini Guided Learning on Students’ Mathematics Scores in Junior Secondary Schools in Sierra Leone." AEA RCT Registry. September 19. https://doi.org/10.1257/rct.16651-1.0

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)

Gemini Guided Learning is an AI-powered instructional tool within Gemini, designed to support students’ mathematics learning through guided questions, step-by-step scaffolding, adaptive feedback, and multimodal resources rather than providing direct answers. The intervention will be implemented in Junior Secondary School classrooms (JSS1 and JSS2, ages 13–15) over an eight to nine (8-9) week period, during which students in treatment classrooms will use Gemini in about half of their mathematics lessons. Teachers will receive training on how to use the tool, although it will only be applied in treatment classrooms.

Intervention (Hidden)

The study is a classroom-level cluster randomised controlled trial conducted in 12 government-supported schools in Sierra Leone. Within each participating school, JSS1 and JSS2 classrooms are randomly assigned to treatment or control. In treatment classrooms, students work in pairs on tablets or computers with Gemini in approximately two out of four weekly mathematics lessons for eight weeks. Gemini provides instructional scaffolds, prompts, and multimodal resources to guide reasoning and problem-solving, aligned with the national JSS curriculum. Teachers of both treatment and control groups receive training on the use of Gemini to ensure familiarity, but the tool is applied only in treatment classrooms. Classroom fidelity is monitored by field staff, who will be present in schools daily to support teachers with lesson delivery, ensure tablets are charged, and provide assistance with any technical or usage issues.

Intervention Start Date

2025-10-06

Intervention End Date

2025-12-12

Primary Outcomes

Primary Outcomes (end points)

The primary outcome of interest is student mathematics achievement, measured using standardised assessments aligned with the JSS1 and JSS2 curriculum and administered at baseline and endline. Assessments are designed to capture gains in core mathematical competencies such as numeracy, problem-solving, and application of concepts.

Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)

Secondary Outcomes (explanation)

Experimental Design

This study employs a cluster randomized controlled trial with classrooms as the unit of randomization. Randomization is stratified by school, grade level (JSS1 vs JSS2), and mathematics teacher using computer-generated random assignment with a pre-specified seed. This stratification approach ensures balanced allocation across key prognostic factors and prevents potential confounding between teacher effects and treatment effects.
Classrooms are randomly assigned to either: (1) Treatment arm receiving access to an AI-powered mathematics tutoring platform in addition to standard instruction, or (2) Control arm receiving standard mathematics instruction only.

Experimental Design Details

Randomization Method

Done in office by a computer using stata

Randomization Unit

Classroom

Was the treatment clustered?

Yes

Experiment Characteristics

Sample size: planned number of clusters

81 sections (44 JSS1, 37 JSS2) across 12 schools

Sample size: planned number of observations

Approximately 4,000 students

Sample size (or number of clusters) by treatment arms

40 control and 41 treated

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

Using a two-level classroom-cluster design with α=0.05 (two-tailed), power=0.80, ICC=0.15, 81 clusters, average cluster size=30, 50/50 allocation, and covariates explaining 50% of within-cluster variance (baseline maths and English) and 40% of between-cluster variance, the MDES is 0.203 SD on the student-level standardised mathematics score

Supporting Documents and Materials

IRB