Self-efficacy, which is defined as an individual’s belief in their ability to successfully complete tasks, is a well-established predictor of academic achievement. In the field of computing, however, gender differences in self-efficacy are observed despite inconclusive findings regarding actual performance. This study investigates how generative learning strategies, specifically peer teaching, influence students’ computing self-efficacy and achievement, with a particular focus on gender differences. The quasi-experimental design was conducted in two phases and involved a total of 377 secondary school students in Year 9 (Scratch programming) and Year 10 (Robotics) classrooms. In both phases, the experimental groups received peer teaching, while the control groups received traditional instruction. Domain-specific instruments were developed and validated to measure computing self-efficacy and achievement.

The first phase took part in a Scratch programming course. A gender gap favouring males was observed in the pre- and post-test self-efficacy scores, but no significant difference in achievement emerged. The generative learning intervention enhanced self-efficacy over time, but did not fully eliminate the gender differences. In Phase 2, a similar gender difference was present at the beginning of a robotics classroom, but it had significantly diminished by the end of the semester. The robotics context, combined with extended peer collaboration and hands-on activities, appeared to create a more engaging environment.

Self-efficacy scores consistently correlated positively with achievement outcomes across both phases, aligning with prior literature. The discrepancy between attitudinal and performance-based findings further highlights the importance of using both types of measure in gender-related educational research. Both Bayesian and frequentist analyses supported the robustness of the findings, with generative strategies producing moderate-to-large effects on self-efficacy.

The study highlights the potential of generative learning strategies to enhance self-efficacy, although outcomes may vary depending on instructional content, age group and learning context. Although Scratch instruction produced modest gains in self-efficacy, the robotics course, with its interactive, project-based format, facilitated greater engagement and reduced gender disparities. These results suggest that interventions aiming to reduce perceived gender disparities in computing should be tailored to the specific context, taking into account factors such as age, content, and socio-cultural learning environment.

Keywords: Computing education, Self-efficacy, Gender differences, Generative learning, Peer teaching, Scratch, Robotics.