The Effects of Robotics and Active-Learning Training on STEM Interest, Curiosity, and Computational Reasoning

The intervention is a government-run educational robotics program implemented by the Ministry of Education of the Dominican Republic. Schools assigned to treatment received an age-appropriate robotics kit intended to support hands-on, active learning in science, technology, engineering, and mathematics. The kit provided depended on the grade levels served by the school: Early Simple Machines for lower primary, Dash or LEGO WeDo 1.0 for upper primary, and LEGO Mindstorms EV3 for lower secondary. The kits were designed to be used by students in small groups, with students taking roles such as organizer, builder, or programmer.

Approximately two teachers from the relevant grades and subject areas were expected to attend centralized training on use of the kits. Where available, the school’s ICT coordinator was also expected to participate. These staff members were then expected to introduce the kits to other teachers in the school. Teachers were encouraged to integrate the kits into regular mathematics and science lessons using activities that linked robotics builds to curricular topics.

Among schools assigned to receive a robotics kit, a second randomization assigned a subset to receive additional training and support in active-learning methods. This supplementary intervention included a virtual pedagogical training on integrating educational robotics into classroom instruction and an 80-hour online diploma in active-learning methodologies. The diploma combined asynchronous coursework with live virtual sessions, individualized tutor support, and a final classroom project applying an active-learning approach.

2023-11-01

2026-06-01

STEM interest and enjoyment
STEM anxiety
STEM aspirations
Behavioral curiosity: (Whether students choose to view the answer to each curiosity prompt and time spent viewing the explanations)
Computational reasoning

STEM interest and enjoyment. This outcome will be constructed from grade-appropriate survey items measuring students’ interest in and enjoyment of mathematics, science, technology, programming, and robotics, as well as their perceptions of the usefulness of STEM subjects. Items will be coded so that higher values indicate greater interest and enjoyment and combined into a standardized index within educational level.

STEM anxiety. This outcome will be constructed from grade-appropriate survey items measuring whether students feel nervous or afraid when studying mathematics or solving difficult mathematics problems. Responses will be coded so that higher values indicate greater anxiety and standardized within educational level.

STEM aspirations. This outcome will be constructed from survey items measuring students’ interest in pursuing further study or employment involving science, technology, engineering, mathematics, or programming. Items will be coded so that higher values indicate stronger STEM-related aspirations and combined into a standardized index within educational level.

Behavioral curiosity. Students respond to five STEM-related prompts and, after each prompt, choose whether to view an explanation. The principal curiosity measure will be the number of prompts, from zero to five, for which the student chooses to view the explanation. We will also examine the time spent viewing the explanation, conditional on choosing to reveal it.

Computational reasoning. This outcome will be constructed from students’ responses to a grade-appropriate assessment measuring skills such as pattern recognition, analysis of algorithms, and the application of algorithms to solve problems. Correct responses will be combined into an assessment score and standardized within educational level so that higher values indicate stronger computational reasoning. We will examine robustness to scoring assessments based on the count of correct responses and a two-parameter item response model.

Curriculum mathematics knowledge
Mindset
STEM self-efficacy
Gender-related beliefs about STEM ability

Curriculum mathematics knowledge. This outcome will be based on students’ performance on a grade-appropriate assessment of mathematics content drawn from the school curriculum. Correct responses will be combined into an assessment score and standardized within educational level so that higher values indicate stronger mathematics knowledge. We will examine robustness to scoring assessments based on the count of correct responses and a two-parameter item response model.

Mindset. This outcome will be constructed from grade-appropriate survey items measuring whether students believe they can learn difficult mathematics or science through effort. Responses will be coded so that higher values indicate a stronger growth-oriented mindset and standardized within educational level.

STEM self-efficacy. This outcome will be constructed from grade-appropriate survey items measuring students’ confidence in their ability to learn or perform well in mathematics and science. Responses will be coded so that higher values indicate greater self-efficacy and standardized within educational level.

Gender-related beliefs about STEM ability. This outcome will be constructed from survey items measuring whether students believe girls and boys can be equally capable in mathematics and other STEM subjects. Responses will be coded so that higher values indicate more gender-equitable beliefs and standardized within educational level.

Not available

Randomization was conducted in an office using a computer. Schools were assigned at the school level through stratified randomization, with strata defined by educational district and robotics kit type. A second computer-based randomization assigned a subset of schools receiving robotics kits to receive additional teacher training in active-learning methods.

Schools.

Yes

288 schools.

20000 students.

140 schools assigned to control; 148 schools assigned to receive a robotics kit. Among the 148 robotics-treatment schools, 73 were assigned to receive additional active-learning training and 74 were assigned not to receive it. One robotics-treatment school was not included in the second randomization.

The minimum detectable effect for the main pooled student outcomes is approximately 0.14 standard deviations, using a two-sided test at the 5 percent significance level, 80 percent power, 288 schools, and standard errors clustered at the school level.