Gender Gaps in Math Anxiety and Achievement: Experimental Evidence from a Gender-Sensitive Pedagogy and Adaptive AI Intervention.

How to cite

Gomez-Ruiz, Marcela; Gonzalez, Mariana; Cancela, Valentina; Castelao, Gimena. 2026. "Gender Gaps in Math Anxiety and Achievement: Experimental Evidence from a Gender-Sensitive Pedagogy and Adaptive AI Intervention.." AEA RCT Registry. AEARCTR-0018898. https://doi.org/10.1257/rct.18898-1.1

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

Marcela Gomez-Ruiz* , Mariana Gonzalez , Valentina Cancela , Gimena Castelao

* Primary investigator

Registry ID
AEARCTR-0018898
DOI
https://doi.org/10.1257/rct.18898-1.1
First registered
Last updated
Status
In development
Trial dates
2026-05-20 – 2027-06-30
Countries
Private
JEL codes
I21, I24, J16, C93
Registry record
http://www.socialscienceregistry.org/trials/18898

Abstract

We evaluate the causal effect of an intervention that combines gender-sensitive teaching practices with the pedagogical use of Matific, a gamified mathematics platform powered by adaptive artificial intelligence, on reducing gender gaps in mathematics achievement through reductions in girls' mathematics anxiety among students in 3rd- and 6th-grade public schools in Uruguay.

Intervention

The intervention includes only two arms—treatment and control—and both 3rd and 6th grades within the same school receive the intervention. In treatment schools, teachers will receive training on gender-sensitive pedagogy and the use of Matific, a gamified mathematics platform with adaptive AI. Workshops will address gender stereotypes, inclusive strategies, and lesson integration. Matific’s AI adjusts task difficulty and feedback in real time, creating personalized learning paths. Teachers will incorporate Matific into regular math lessons, while platform data (usage, time on task, completion, and errors) will be automatically collected to assess engagement. The training consists of a 150-hour program, including 30 hours of synchronous workshops, covering data-based diagnostics, classroom climate improvement, the pedagogical use of Matific, and student engagement dynamics, designed to support the effective delivery of the intervention. In class, teachers are expected to work with the materials and lesson proposals designed by the mathematics pedagogical team at Ceibal, which are intended to address mathematics anxiety, the core focus of the intervention. Teachers in the control group will have access to the training materials once the post-implementation results have been measured. The intervention will last for 120 hours.

Experimental design

The intervention consists of two parallel cluster RCTs with 3rd- and 6th-grade students across 76 public schools. First, teachers are invited to participate in the project, then randomization will occur at the school level, ensuring both grades are included. Schools will be assigned to treatment (38) or control (38), stratified by socioeconomic context and school type, given gender disparities in mathematics across these dimensions. Uruguay’s one-teacher-per-class model eliminates within-school spillovers, as teachers do not rotate across grades. To avoid cross-school spillovers, if a teacher works in both a treatment and a control school, the control school will be replaced with a matched alternative.
Before the intervention begins, we will collect baseline data from both teachers and students. Data collection is expected to be completed by the mid of June 2026. Endline data will be collected for three weeks starting from the last week of October. Both fieldwork activities will be carried out by trained survey staff.
Because participating teachers agreed to take part in the study, compliance is expected to be relatively high. However, we acknowledge the possibility of dropout or non-compliance during the intervention period A dedicated Ceibal team will accompany teachers throughout the intervention period via WhatsApp groups organized by the treatment arm, with a specific role in monitoring the usage targets set by the pedagogical team and providing ongoing support. If any teacher decides to drop out of the intervention, we will assess whether dropout occurs at random.

Design details

Randomization method
Randomization will be performed on a computer at the offices of Ceibal.
Randomization unit
Randomization occurs at the school level, while outcomes are measured at the student level. Therefore, standard errors will be clustered at school level.
Clusters
76 schools
Observations
2,500 students in total, or about 1,250 per grade
Arms
38 schools in treatment and 38 schools in control group.
Primary outcomes
Math anxiety in the short term, and math achievement in the medium and long term.
Secondary outcomes
Self-efficacy, Self-confidence, error normalization

Cited by

No citation data available for this registration yet.

Keywords