Early Childhood Coding Classrooms: Evidence from a Randomized Controlled Trial in Uruguay

• In Event: Highlighted Session: Technology, AI, and Policy in Education

Proposal

Research shows that early exposure to computational thinking (CT) or solving problems and designing systems using core computer science (CS) principles (Wing, 2006), is valuable at any age, but especially during the early years of education, to support traditional academic growth while also fostering creativity, problem-solving, collaboration, and adaptability (Bers, 2018; Sarama & Clements, 2009; Friedman-Krauss et al., 2012). Despite these benefits, developmentally appropriate CS programs for young children remain scarce, limiting children’s access to foundational computing knowledge by the end of elementary school (Yang et al., 2025).
Building on this evidence, the current study responds to the increasing need for applied research on early CS interventions by implementing a program in Uruguay, which provides a particularly relevant case because public school students have had access to devices and the internet since 2007. Additionally, the 2020–2024 National Curricular Framework established CT as a K-12 cross-disciplinary competency (ANEP, 2022). However, despite these progressions, early CS implementation remains limited due to a lack of training opportunities, developmentally appropriate technologies, and pedagogical approaches.
To address these challenges, the Coding as Another Language (CAL) project is a school-based intervention aimed at improving early childhood educators' ability to teach CS while developing children’s CT knowledge and coding skills. Based on the Positive Technological Development (PTD) framework (Bers, 2006), CAL views coding as a form of literacy that supports both technical learning and social-emotional growth, introducing CS through creativity and personal expression, using developmentally appropriate technologies, such as ScratchJr, that enable young learners to design meaningful and shareable digital artifacts (Brennan & Resnick, 2012).
After an adaptation process in partnership with local educators and researchers, the CAL intervention comprised a 15-lesson curriculum for second-grade classrooms that introduced core concepts such as algorithms, sequencing, and debugging through storytelling and writing . Each lesson included unplugged activities, structured challenges, and open-ended explorations to promote expressive and collaborative learning (Bers et al., 2023). To support the implementation, professional development (PD) opportunities aimed at addressing the limited training available to elementary teachers in these areas, thereby enhancing their self-efficacy, preparedness, and confidence in integrating new technologies and concepts into their practice.
Following a randomized controlled trial with a delayed-treatment design (Figure 1) stratified by school-level characteristics like sociocultural quintiles and teachers’ prior exposure to CT programs, the study explored how early interventions can improve CS education by boosting teachers’ self-efficacy and confidence, while also supporting children’s skill development. Student outcomes were assessed using two validated instruments: the Automated Coding Stages Assessment (CSA-A), which measures programming skills in the ScratchJr environment (de Ruiter & Bers, 2021), and TechCheck, which evaluates basic CT skills (Relkin et al., 2020). Teachers also completed the CSA-A and pre- and post-training surveys measuring self-efficacy. Following Kapoor et al. (2023), a self-efficacy index was constructed from seven survey items assessing teachers’ perceived ability to teach programming.
Figure 1. Study Design
A total of 37 teachers were recruited and randomly assigned to either the treatment group (N=17), which received PD and implemented the curriculum from May to August 2024, or to the control group (N=20), whose training was postponed until October 2024. At the same time, 501 children completed pre- and post-intervention CSA-A (260 in the treatment group and 241 in the control group), and 550 students participated in pre- and post TechCheck (257 in the treatment group and 275 in the control group) (Appendix 1, Table 1).
To analyze these data, a multivariate linear regression with school-clustered standard error was employed, including control variables such as gender, location, class size, teacher experience, prior CT exposure, and school sociocultural level, obtaining that the CAL intervention significantly improved students’ programming skills, as measured by CSA-A, but did not produce measurable gains in CT according to TechCheck (Appendix 2, Figures 2 and 3).
Treatment students outperformed controls by 4.5 points on the CSA-A posttest, with more students advancing from “Emergent” to “Coding and Decoding” and “Fluency” (Appendix 2, Graph 1). Subgroup analysis revealed no gender-based differences (Appendix 3, Table 2), but students in higher sociocultural quintiles (3–5) gained an additional 2.8 points compared to peers in lower quintiles (Appendix 3, Table 3). At the teacher level, treatment participants reported a 29% increase in self-efficacy scores (22→29), versus 3% in the control group, and their CSA-A scores nearly doubled (8.4→16.2) after the PD (Appendix 4, Graph 2). However, implementation fidelity emerged as a key moderator: students with highly engaged teachers scored 5.4 points higher, versus 1.9 with low-engagement teachers (Appendix 4, Table 4), pointing to the importance of sustained teacher commitment.
These results provide new causal evidence from a middle-income country, while also raising key questions about equity, sustainability, and long-term impact regarding the potential of introducing programming in early education to enhance both student learning and teacher instructional strategies. Teachers reported greater self-efficacy, confidence, and competence in integrating technology and programming into classroom practice, emphasizing the value of early exposure not only for cognitive development but also for fostering pedagogical innovation.
However, benefits were not evenly distributed, with outcomes influenced by sociocultural context, suggesting that without targeted support, programming initiatives risk reinforcing existing inequalities. Additionally, short-term effects on CT, as abstractly measured by TechCheck, were not significant, indicating that either more prolonged exposure or more sensitive assessments are needed. Yet, teachers informally reported that students demonstrated gains in foundational CT skills such as sequencing, problem-solving, and logical reasoning.

Author

• Francisca Carocca, Boston College, Lynch School of Edu