Low-cost robots revolutionize how AI is taught in secondary classrooms

The rapid expansion of AI and machine learning into everyday life has made it critical for students to gain foundational literacy in these technologies. However, traditional methods of teaching AI, often reliant on lectures, slides, and software simulations, have struggled to hold student attention or foster meaningful understanding, particularly among younger learners with no prior exposure to programming or computational theory.

The use of low-cost, 3D-printed robots to teach artificial intelligence and machine learning can dramatically enhance student engagement and understanding in secondary schools, according to a new study published in Applied Sciences. The study titled “Teaching Artificial Intelligence and Machine Learning in Secondary Education: A Robotics-Based Approach” outlines a two-part lesson plan using physical robots to explain complex AI and ML concepts without requiring prior programming knowledge or expensive infrastructure.

The robotics-based approach addresses the growing need to make AI education more accessible, intuitive, and motivating for students aged 14 to 15. Unlike conventional computer-based or theoretical instruction, this method leverages the physical presence and interactive potential of robots to teach core ideas such as rule-based automation, supervised learning, and reinforcement learning. The intervention was tested in a Greek lower secondary school and compared with a prior software-only lesson. Results showed a marked increase in student interest, retention, and comprehension when robots were introduced.

What challenges does traditional AI education face in the classroom?

The rapid expansion of AI and machine learning into everyday life has made it critical for students to gain foundational literacy in these technologies. However, traditional methods of teaching AI, often reliant on lectures, slides, and software simulations, have struggled to hold student attention or foster meaningful understanding, particularly among younger learners with no prior exposure to programming or computational theory.

In a previous implementation without robots, students were introduced to AI concepts using Scratch and ML4Kids. While initially engaging, interest waned over time. Classroom observations and post-lesson questionnaires indicated moderate enthusiasm and uneven comprehension. Several students disengaged during sessions, and many struggled to apply AI concepts to real-world scenarios. The challenge was clear: without tangible, relatable applications, AI can seem abstract and disconnected from students' experiences.

This is where robotics enters the equation. By allowing students to see and manipulate machines that embody the principles they’re learning, robots make AI visible and tactile. The study’s authors identified cost, technical complexity, and teacher preparedness as key barriers to robotics integration. Their solution was to design two simple, inexpensive robots using open-source software and hardware, allowing educators with limited resources and technical background to implement the lesson with minimal preparation.

How do the robots bring AI and ML concepts to life?

The lesson introduced students to two physical robots, each designed to highlight different aspects of intelligent behavior. The first robot, the Arduino Mini Manipulator, is a robotic arm that operates in two modes. In the first mode, students learn about automation by manually setting the robot’s positions using a joystick and programming it to repeat them - a form of rule-based, pre-programmed behavior. This lays the foundation for understanding that not all machine behavior is intelligent or adaptive.

In the second mode, students use Google's Teachable Machine platform to train an image classification model that allows the robot to respond to human body poses. This transition from manual control to AI-guided interaction illustrates the leap from automation to machine learning. Students see firsthand how a machine can be trained to recognize inputs and act accordingly - not just follow fixed scripts.

The second robot, called the SelfLearn robot, teaches reinforcement learning. It moves along a track and uses an ultrasonic sensor to measure its distance from a fixed object. By executing random joint movements and recording the outcomes, the robot gradually learns which actions increase its distance from the obstacle. Over time, it optimizes its behavior through trial and error - a core principle of reinforcement learning. Students observe how the robot improves its performance without direct programming, internalizing the concept of self-directed learning in machines.

What impact did the robotics-based lesson have on student learning?

The robotics-enhanced lessons were implemented in two 45-minute sessions. The first session introduced theoretical concepts and demonstrations using the robots, while the second took place in a computer lab where students trained machine learning models and integrated them with the robots. The instructional design required no prior coding experience, and all materials were freely available online, including software, robot design files, and scripts.

Results were evaluated through a combination of questionnaires and classroom observations. Compared to the earlier software-only version, the robotics lesson produced significantly higher engagement. None of the 20 students lost interest during the sessions, and many displayed heightened enthusiasm, with several asking to continue working with the robots during break time. When surveyed, students reported a mean interest level of 4.85 out of 5 for the lecture and 4.20 for the topic overall, both higher than in the non-robotic version.

Equally important, comprehension improved. Students more accurately distinguished between rule-based behavior and machine learning, showed a better understanding of the Teachable Machine platform, and grasped the basics of reinforcement learning. All 20 students gave thoughtful, relevant examples of how AI could be used in real-world applications - ranging from coaching and fashion advising to inventory management and personalized tutoring. In the previous lesson, nearly a quarter of the students gave incorrect or irrelevant responses to the same question.

The physical presence of the robots and the interactivity they offered appeared to be the deciding factors. Students were more motivated when they could see their code executed by a physical object. The process of training models, testing their accuracy, and witnessing real-time feedback created a powerful loop of learning and reinforcement that digital-only tools could not replicate.

Moreover, the hardware proved reliable and the lesson manageable for both students and instructors. The Arduino Mini Manipulator cost approximately €60, while the SelfLearn robot cost around €30 - substantially less than most commercial educational robotics kits. The robots did not require dedicated classrooms or specialized labs, making the setup adaptable to a wide range of school environments.

Notably, time was the only noted limitation. A single teaching hour was often insufficient for in-depth exploration or extended student-teacher dialogue. Nevertheless, students’ interest in continuing the activity, paired with their increased conceptual clarity, confirmed the added pedagogical value of this approach.

As a result of the overwhelmingly positive feedback, the methodology has been extended to more classes and will be used in teacher training programs. National curriculum updates are also making room for AI instruction, which will further support the deployment of hands-on robotics-based teaching strategies in the years to come.