Integration of Artificial Intelligence in K-12: Analysis of a Three-Year Pilot Study

Abstract

A three-year pilot study investigated the effectiveness of artificial intelligence (AI) as a motivational tool for teaching programming concepts within the Croatian Informatics curriculum. The study was conducted in schools through the extracurricular activity EDIT CodeSchool with the Development of Intelligent Web Applications (RIWA) module. Twelve schools in Split-Dalmatia County in the Republic of Croatia participated, resulting in 112 successfully completed student projects. The program consisted of two phases: (1) theoretical instruction with examples and exercises, and (2) project-based learning, where students developed final projects using JavaScript and the ml5.js library. The study employed project analysis and semi-structured student interviews to assess learning outcomes. Findings suggest that AI-enhanced learning can effectively support programming education without increasing instructional hours, providing insights for integrating AI concepts into existing curricula.

1. Introduction

In an era where artificial intelligence (AI) is changing industries and automating numerous jobs, AI literacy has become essential for students. AI education is no longer limited to graduate-level computer science students; it has now expanded to elementary and high school students, with varying levels of AI integration [1]. In many countries, it has taken decades to incorporate computational thinking and programming into school curricula. This raises the question of how AI can be integrated into education, given the constraints of limited resources and experience [2]. Teachers are actively seeking meaningful ways to incorporate AI topics into the K-12 curriculum [3]. Existing curricula designed to develop AI literacy for primary and secondary school students focus on computational thinking and AI concepts [4]. However, the curricula require further adaptation to reflect the growing emphasis on machine learning and artificial intelligence [5]. The lack of integrated AI curricula poses a barrier to exposing students to artificial intelligence [6]. Beyond informal programs, AI competencies can be strengthened by embedding AI education within school curricula. This can be a crucial element in implementing national AI strategies, as seen in some countries [7]. Although interest in AI education is substantial, there is a shortage of empirical research supporting the development of high-quality curricula, highlighting the need for further studies [8].

Despite the surge in research on AI in education over the past years, long-term studies spanning multiple years remain scarce. We conducted a three-year study, allowing for intervention and modification once a year at the beginning of each cycle. While most research focuses on tools and generative methods, our study examines how AI can be integrated into existing content rather than being introduced as an entirely new subject. This raises an important consideration: increasing the volume of instructional content may not necessarily yield the desired learning outcomes and should be carefully evaluated. Given that programming concepts can be challenging for students, we decided to shift the context rather than increase the workload. Computational thinking and programming are already part of the existing Informatics curriculum. Therefore, we leveraged students’ interest in artificial intelligence as a means to teach programming and demonstrate the real-world applications of computers.

This study proposes the following research question: Can artificial intelligence be successfully integrated into the primary school Informatics curriculum within the existing content and instructional hours?

2. Background

2.1. AI Literacy in Education

AI literacy has become a prerequisite for full participation in today’s society [9]. An AI-literate citizen should have the ability to use artificial intelligence for the benefit of humanity [10]. In a world full of fake news, such as AI-generated images and videos, understanding the principles and concepts behind them will help people develop critical reasoning skills [11]. Therefore, it is essential that children learn to develop an informed understanding of artificial intelligence, which is one of the goals of the AI literacy frameworks [12]. Understanding the basics of how artificial intelligence works is one of the key aspects of AI literacy [13]. Recent advancements in AI research, such as few-shot learning, demonstrate how machine learning models can classify new information with minimal training data, improving generalization in real-world applications [14]. These techniques highlight the importance of teaching students both the limitations and capabilities of AI models within educational settings.

To develop AI literacy, it is important to introduce children to basic AI concepts as early as possible and to help them discover the connection between AI applications and fundamental principles of artificial intelligence [15]. AI literacy encourages students to engage with new technologies, reducing the fear of an AI-driven world [16]. Successful introduction of AI literacy to diverse students requires that AI concepts be relevant and relatable, using pedagogical methods that align with the students’ cultural context. This approach connects AI literacy, technologies, and tools to students’ culture and everyday life, thus engaging them and empowering their interest in learning about artificial intelligence [17]. As AI research continues to evolve, advancements in machine learning and representation learning contribute to the development of intelligent systems with enhanced capabilities [18,19,20]. AI will radically transform the labor market, which will accordingly make it important to develop students’ AI literacy and skills for working with AI systems [21]. Researchers and curriculum developers have begun to create learning opportunities to address gaps in students’ AI literacy [22].

Research on the implementation of AI literacy in education is challenging because the mechanisms and potential of artificial intelligence are unknown to most people outside of computer science. Artificial intelligence topics are quite absent from the curriculum, which causes limited time to teach AI content [23], for example, through enrichment of existing content through extracurricular activities. Research on children’s initial perception of artificial intelligence is at an early stage, from which questions arise about the development of curriculum, methods, and materials for students’ AI literacy [12]. If the education system aims to prepare students for the demands of an AI-driven future, then it must integrate AI technologies as a curriculum subject to equip students with fundamental knowledge about artificial intelligence. Recent advancements in AI, including applications in areas such as human pose estimation and behavioral biometrics, highlight the growing impact of AI across various industries, further emphasizing the need for AI literacy in education [24,25,26]. Motivation and attitudes about artificial intelligence have been shown to be important factors influencing AI literacy and should be taught before other elements, such as concepts and teaching content [27]. It is important to constantly align the educational curriculum with modern technological requirements [28], thereby achieving comprehensive AI literacy [29]. A standardized curriculum and specialized teacher training will be essential for developing students’ skills in the future.

2.2. The Role of Teachers in Teaching AI

There is a need for the professional development of teachers to enhance AI literacy in students. This will prepare teachers to make AI curricula accessible and inclusive for their students across different subject areas [30]. The success of AI education is closely related to the readiness of teachers, as the teaching and learning process will hardly succeed without their central role. The lack of teacher expertise and well-defined educational programs is often highlighted as the main drawback, with professional teacher training being suggested as a possible solution [31].

Along with the need to develop students’ AI literacy, it is important to inspire and empower teachers to create teaching interventions related to AI topics [32]. Teachers with higher AI literacy are better able to recognize and evaluate AI tools and foster self-efficacy and creative intent, leading to improved AI performance in the classroom [33]. This partly includes the creation of technological resources that support learning and teaching [34]. However, many teachers feel they lack sufficient preparation to teach AI and find it challenging to incorporate additional content into the curriculum due to their existing responsibilities [35].

2.3. Pedagogical Approaches in Teaching AI

In the educational process, teachers use different learning methods to help students develop AI competencies. The most common approaches are constructivism and constructionism. Combined with constructionism, design-oriented pedagogy emphasizes open-ended, real-life tasks [36]. These tasks do not have a single solution or a “correct” answer but allow students to generate different solutions that they consider meaningful to certain problems [37]. A systematic review of empirical studies on AI in primary education identified key pedagogical approaches, including project-based learning, AI-assisted instruction, and experiential learning, highlighting their impact on cognitive, affective, and psychomotor domains [38].

Project-based learning (PBL) is an active, learner-centered approach characterized by learner autonomy, constructive inquiry, goal setting, collaboration, communication, and reflection within real-world practice [39]. It is often accompanied by collaborative learning and the use of technology. Through working in pairs or groups, students explore the meaning of the content, with the teacher serving as a guide. The exchange of ideas, suggestions, and experiences helps students evaluate their own ideas and those of others with the goal of finding the optimal solution. Each student’s unique skills contribute to the group in this process. By working together, students can reflect critically on their learning and recognize that learning is a process that presents them with challenges they must overcome [32].

Collaborative learning and project-based learning have been highlighted as effective methods for teaching artificial intelligence, with special emphasis on the use of devices and software tools that allow students to actively acquire knowledge about AI [40]. Recent research further confirms that AI literacy education frequently incorporates project-based, human–computer collaborative learning and play- and game-based approaches, utilizing tools such as Google’s Teachable Machine 2.0, PopBots, and Scratch 3.0 to enhance engagement and conceptual understanding [41]. Both approaches include problem-based learning, which is frequently used in classrooms when teaching artificial intelligence. Students grasp AI concepts more quickly when they are connected to practical examples and real projects. Therefore, curricula should emphasize the application of theoretical knowledge through practical tasks to improve students’ understanding of complex technological concepts [42]. The systematic introduction of AI literacy through practical, problem-oriented approaches is recommended to foster deeper student engagement, thereby ensuring comprehensive AI literacy [43].

3. Methodology

3.1. Research Design

A three-year pilot study was conducted to examine programming concepts and AI literacy among upper-grade primary school students through the extracurricular activity EDIT CodeSchool. This extracurricular activity is part of a larger seven-year project, Digital Dalmatia, with artificial intelligence being taught over the last three years using a specially designed module, Development of Intelligent Web Applications (RIWA), created specifically for this study.

During this three-year period, the extracurricular activity was implemented in 12 primary schools across Split-Dalmatia County in Croatia, resulting in a total of 112 successfully completed student projects. The activity focused on programming and the use of AI tools, utilizing JavaScript and the ml5.js library. The final outcomes were intelligent web projects for object recognition.

During the pilot phase, four in-person workshops were held to familiarize teachers with the RIWA module content and structure. All supplementary materials (including brief video lessons, step-by-step programming tasks with explained solutions, and slide presentations) were made available on a shared Moodle course. Teachers also received ongoing support through an online chat group for real-time troubleshooting and monthly feedback meetings.

The RIWA module was carried out over three separate years. It covered educational content related to artificial intelligence, computational thinking, and programming and was delivered through theoretical lessons, problem-solving tasks, unplugged activities, and programming using various devices and AI tools. The primary focus of the module was on programming in JavaScript, with additional use of HTML and CSS, while the ml5.js library enabled students to work with machine learning models. This ready-made library was used following the method of “didactic hiding” [44], as a tool to help concretely present the learning material for students without requiring detailed abstraction.

Students developed applications that recognize objects in images using machine learning, and their final projects demonstrated their acquired skills. They independently developed their projects using a template provided to them, with support from teachers.

The module was designed for students with no prior knowledge of artificial intelligence. It is broad and flexible enough to emphasize inclusivity and adaptability for individualized educational programs for students with special needs. With different pacing and depth of knowledge, it also offers gifted students the opportunity to engage more deeply and creatively, aligning with Papert’s principle of “low floor, high ceiling tasks” [45]. The evaluation was based on the analysis of student projects, semi-structured interviews, and final knowledge tests.

The EDIT CodeSchool module was developed by university teachers from the Faculty of Science at the University of Split and was legally implemented in schools with the approval of school boards and parental consent for student participation and data processing. Principal consent facilitated the organization of extracurricular activities in schools. All raw research data are available upon request.

3.2. Population

The study included 112 upper-grade primary school students from 12 schools in Split-Dalmatia County, Republic of Croatia. In addition, interviews were conducted with a convenience sample of students to obtain additional insights. Groups were organized according to schools, with the possibility of including students from schools that did not participate in the implementation of this activity. Groups were matched based on student achievement, ensuring a mix of students with different levels of prior knowledge. It is worth noting that the activity was designed to be inclusive for all students, not just the gifted.

Participation in the extracurricular activity did not result in school grades for students. Attendance was voluntary, with parental consent required for participation and data processing. Teachers played an important role, providing feedback on student progress throughout the process.

3.3. Data Collection and Instruments

Data collection and analysis were carried out using the triangulation method:

• Evaluation of student projects: Projects were reviewed based on the following criteria: functionality, use of programming concepts (such as variables, conditional statements, loops, and functions), code clarity, quality of input data for training and testing, and student creativity in problem-solving.
• Semi-structured interview: Individual students participated in interviews to gain deeper insights into their knowledge, perceptions, challenges, and motivation. The questions explored their understanding of the module content, their perception of AI and related activities, and their overall experience. A special review of the projects and the tools used was also conducted.
• Teacher’s notes: During the activities, teachers took notes on students’ progress, observations made during sessions, and challenges that arose. These observations served as additional data for a comprehensive analysis and a better understanding of how effectively the module was implemented.

3.4. The RIWA Module and the Informatics Curriculum

3.4.1. Description of the RIWA Module

The Development of Intelligent Web Applications (RIWA) module is structured as an extracurricular activity for students in the upper grades of primary school, aiming to familiarize them with the essential aspects of digital transformation. The module focuses on practical exercises and project-based learning, with an emphasis on problem-solving. Through a series of activities, students are gradually introduced to more complex topics such as artificial intelligence, neural networks, data processing, and the tools used to create a web application. The final goal for students is to develop a simple yet functional web application that uses artificial intelligence to recognize objects in an image or via a camera.

The module follows a “from concrete to abstract” approach, meaning that students first engage with the practical applications of these technologies through hands-on activities. From there, they move on to abstract concepts, such as the algorithms used in image recognition. With carefully selected tools and a programming language appropriate for their age, the module guides students through each stage of application development, from displaying data on a computer to using the necessary tools to bring an idea to life. The focus is on the machine learning process, where students train the application to identify objects—such as fruits and vegetables—by collecting data and examples.

To facilitate this learning process, the ml5.js library was selected as a beginner-friendly wrapper for TensorFlow.js, providing extensive documentation, project examples for various network types, and online video tutorials. These built-in resources allowed teachers to familiarize themselves with the concepts and guide students effectively, even without prior AI expertise. The structured support within ml5.js lowers the barrier to entry for educators while also enabling a smooth transition to more advanced AI tools and problems in higher education, making it a scalable and adaptable choice for AI literacy programs.

One of the key aspects of the RIWA module was teaching students the importance of data quality in machine learning. The concept of “garbage in, garbage out (GIGO)” was visually depicted in the learning materials, emphasizing how poor-quality training data lead to incorrect AI predictions. This was further reinforced in video lessons through concrete examples, such as demonstrating how a model trained only on green apples might misclassify a red apple as a different fruit. By integrating these concepts into practical programming tasks, students developed a clearer understanding of how data selection impacts AI performance.

The module also helps students understand the operation of intelligent systems and how such systems can digitize everyday tasks. The key topics covered include data display on a computer, image processing, the concepts underlying artificial intelligence and machine learning algorithms, and the application of these skills in developing an application capable of recognizing objects in an image. After completing practical tasks and acquiring the necessary skills, students begin developing an application that is supported on different platforms, from desktops to mobile devices. Upon completion, students present their projects, showcasing their solutions and demonstrating an understanding of artificial intelligence and computational thinking concepts.

Working through practical examples not only increases students’ interest in computer science but also equips them with skills essential for both the present and future digital landscape. This approach enables students to gain a deeper understanding of how technology is integrated into everyday life.

3.4.2. Comparison and Mapping

Considering that the goal of this study is to assess whether this approach can be incorporated into regular classes without changing the existing curriculum or the number of instructional hours, we compared the RIWA module with the Informatics curriculum for primary schools in the Republic of Croatia. Informatics includes 70 instructional hours per year, while RIWA is an extracurricular activity scheduled for 35 h per year.

The Informatics curriculum consists of four main domains: Information and Digital Technology, Computational Thinking and Programming, Digital Literacy and Communication, and E-Society. This framework provides guidelines for achieving the required learning outcomes, offering teachers flexibility in adapting content and teaching methods while allowing for the introduction of contemporary topics. The adaptive approach helps teachers align lessons with student interests, integrate new technologies, and encourage creativity in learning and problem-solving.

Both the Informatics curriculum and the RIWA module focus on developing digital skills and computational thinking. The Informatics curriculum places particular emphasis on digital skills, covering digital tools, online resources, and multimedia content. The RIWA module also includes the use of technology and tools to build applications, with a specific focus on developing digital skills through web applications for image object recognition.

The development of computational thinking and problem-solving is essential in both programs. While the Informatics curriculum encourages students to understand and create algorithms, the RIWA module takes a hands-on approach, such as using object recognition. Both programs utilize project-based learning. In the Informatics curriculum, project activities emphasize teamwork and problem-solving, while the RIWA module includes projects in which students apply their acquired skills and knowledge independently or in teams and present their solutions.

The RIWA module emphasizes modern technologies by introducing innovative approaches for primary school students, such as object recognition in intelligent applications. Although the Informatics curriculum includes programming and digital skills, it is less specific about the integration of advanced technologies such as machine learning and computer vision, which are core components of the RIWA module. While the RIWA module is focused on intelligent applications, the Informatics curriculum covers a broader domain of computational thinking and programming. It is more general in scope but remains open to incorporating various technologies and tools.

To illustrate the alignment between the RIWA module and the Informatics curriculum,

Table 1. Mapping of the RIWA module and the Informatics curriculum.

RIWA Module (Outcomes and Content)	Informatics Curriculum (Outcomes)
Explain how data are represented in a computer.	Collects and inputs data to analyze a problem using an appropriate program; discovers relationships between data using different program tools and options for data representation.
Describe what computer intelligence is.	Recognizes and studies interdisciplinary tasks that have been enhanced by the development of informatics and information and communication technology.
Discuss the concept of learning from examples.	Uses simulation to solve a problem that may not necessarily be computer-related.
Apply tools to create a web application for solving object recognition problems.	Develops algorithms to solve various problems using a programming language, employing appropriate data structures and types.
List examples of intelligent systems for digitalization and digital transformation of society.	Recognizes and studies interdisciplinary tasks that have been enhanced by the development of informatics and information and communication technology.
Variables	Uses programming tools to create a program that includes input and output values and repetition.
Functions	Designs and creates modular programs that include subprograms in a programming language.
Arrays	Develops algorithms to solve various problems using a programming language, employing appropriate data structures and types.
Decisions	Creates, monitors, and restructures programs containing branching and conditional repetition structures and predicts the behavior of simple algorithms that can be represented by diagrams, natural language descriptions, or programming code.
Loops	Uses programming tools to create a program that includes input and output values and repetition.

presents a mapping of the components of the RIWA module to the corresponding domains within the Informatics curriculum. This mapping provides a structured overview of how the outcomes and activities of the RIWA module align with the general objectives of the Informatics curriculum.

Both programs aim to develop digital skills and computational thinking. The mapping demonstrates how students engage with key concepts of digital literacy, computational thinking, and programming through both modules, while the RIWA module further expands their skills in artificial intelligence—one of the essential aspects of the digital age. This approach fosters a deeper understanding of intelligent systems, which is particularly valuable for preparing students for the world they are living in.

4. Findings

The projects were analyzed based on the programming concepts students applied in their work. The focus was on identifying the most commonly used concepts necessary for the functional operation of the applications. Additionally, an analysis of results obtained through semi-structured interviews with students about their experiences during the creation of final projects within the RIWA module revealed several key themes related to their understanding of project work, sense of achievement, pride, and perception of success.

4.1. Overview of Student Projects

After three years of activity, a total of 112 student projects were successfully completed. Considering the students’ maturity and prior knowledge, most were not capable of independently programming an entire project from scratch. For this reason, the project utilized the ready-made JavaScript ml5.js library for training and testing the network, along with a teacher-provided project proposal (Figure 1) that students expanded by incorporating programming concepts and adding their own examples for training the model.

These project proposals were carefully developed by the course authors, who designed them to facilitate learning by incorporating pre-configured parameters, optimized neural network architectures, and well-tested configurations. Each template was handcrafted to ensure that students could focus on applying AI and programming concepts without being overwhelmed by the complexities of setting up a machine learning model from scratch. The structured design allowed students to modify key aspects of the models while still following a guided approach, making the learning process more intuitive and efficient.

This approach required students to apply previously acquired knowledge and skills in a new context, in alignment with Bloom’s taxonomy. A code-based approach was used in programming, enabling students to explore AI integration in a structured and manageable way.

The projects demonstrated the students’ significant progress in applying programming concepts with artificial intelligence, emphasizing creativity alongside their programming skills. A detailed analysis of student projects reveals how students successfully applied programming concepts in developing their applications. The most commonly used concepts were variables, conditional branching (if-else), and repetition statements (for and while loops).

Variables were primarily used to store input values, conditional branching enabled decision-making within programs based on conditions, and loop statements were used to automatically repeat tasks. Additionally, students used functions to organize their code, leading to more modular and efficient programming structures. This application of programming concepts demonstrates that students have built a strong foundation in computational thinking, understanding how to structure their code logically and efficiently. Each year of programming experience influenced the quality of their projects, their self-confidence, and their willingness to use different programming concepts.

Through the analysis of the final projects, the challenges students encountered during application development were also evident, primarily related to the accuracy of object recognition. A common issue was that, due to a lack of training examples, the program produced poor results during testing. For example, distinguishing between dog breeds such as the Husky and the Alaskan Malamute was problematic. Because the neural network was trained with an insufficient number of diverse examples, the program often made incorrect classifications—primarily because most of the selected images for both breeds had snow as a background. However, some students recognized the importance of a well-structured dataset and made a considerable effort to collect diverse and balanced training data. Their projects, which included a significant number of images with varied backgrounds, lighting conditions, and object orientations, demonstrated noticeably better classification accuracy.

From the analysis of projects, it can be concluded that the minimum essential elements for creating functional projects include variables, branching, loops, functions, fetching HTML elements, and managing images as input values. More advanced learners incorporated additional features, such as using a webcam as an input source and implementing a chatbot with voice input for conversation.

Despite all teachers having access to the same materials in the extracurricular activity, differences in teaching approaches led to varying student outcomes. Some teachers did not emphasize certain programming concepts that turned out to be essential, which affected the quality of student projects. It was also found that the teacher–student relationship played a significant role. Students who already knew the teacher from regular classes put in more effort, participated more regularly, and demonstrated greater knowledge through their project assignments. Conversely, teachers who conducted extracurricular activities with students they had not previously taught faced challenges in motivating students and establishing a clear hierarchy and authority. Additionally, the projects of students who worked online were of lower quality than those of students who were under the constant supervision of a teacher.

After completing their projects (an example of one project is shown in Figure 2), students presented their work in video format. It was observed that preparing for the presentation phase helped students gain additional understanding and consolidate their knowledge. It is important to emphasize that before developing their project applications, students created a plan outlining how the application should look and function. For most students, the initial documentation of their plan was modified during project development. This indicates that while planning is one thing, actual implementation—along with its unexpected challenges—is quite another, especially for students encountering this type of project-based application development for the first time. Furthermore, students’ initial expectations regarding artificial intelligence were notably high.

4.2. Thematic Analysis

4.2.1. Theme 1: Projects

When students discussed how their projects recognized objects in an image, most demonstrated a basic understanding of how their applications worked. They explained that the application takes an image as input and then processes it to identify what is in the image. However, the level of detail in students’ explanations varied depending on their programming experience. Those with more experience provided detailed responses about how their application functioned, while students with less experience gave more general explanations.

The results show that students are more motivated to work on projects involving artificial intelligence. In interviews, most students expressed satisfaction with the module, emphasizing how the practical tasks helped them understand the fundamental concepts of artificial intelligence and programming. Most students were proud of their projects, explaining that the process of creating them allowed them to meaningfully apply what they had previously learned, which gave them a strong sense of achievement.

This sense of pride was reinforced by the fact that many students shared their projects with family and friends, demonstrating their satisfaction with the results. Some students acknowledged that they could have put in more effort. This feeling was more common among those who had high expectations at the beginning or encountered technical difficulties while developing their applications. Their self-criticism reflects an awareness of the importance of continuous learning.

In interviews, students conveyed their enjoyment of the creative process involved in developing their own projects. They emphasized that working on projects gave them the freedom to explore different technical possibilities, which they found enjoyable. Many felt that working on the project not only improved their technical skills but also boosted their confidence in programming and problem-solving.

4.2.2. Theme 2: Assessing the Success of Learning Programming Concepts

After analyzing the students’ projects, their responses regarding their understanding of key programming concepts were considered. Some insights were gained regarding their knowledge of programming terms.

Most students demonstrated a basic understanding of variables, recognizing them as a “tool for storing values within a program”. The level of detail in their responses varied. Some students provided precise definitions, while others gave more general statements such as “to store something”. This indicates that while most students have a solid foundational understanding of variables, there is still room for a deeper grasp of how they are used in different programming contexts.

Conditional statements (if–else) were generally well understood by students, and they used them correctly in their projects. However, the depth of understanding varied. For example, some students clearly explained the difference between “if” and “else if”, while others showed uncertainty in explaining how the program executes based on different conditions.

Questions about loop statements (for and while loops) revealed similar differences. Some students made a clear distinction between their execution, explaining when each should be used in the repetition part of a program. Others, despite recognizing that both loops serve the purpose of repetition, were unsure when it was appropriate to use each one. This suggests that additional practice with examples is needed for a clearer understanding of these concepts.

Most students understood the purpose of using functions in a program, although the clarity of their responses varied. Students who demonstrated more knowledge through their projects provided more precise answers about when to use functions, emphasizing the importance of code organization and the reusability of program components. Students with less demonstrated knowledge, although they used functions, were uncertain about their advantages.

When asked about the roles of HTML, CSS, and JavaScript, most students correctly identified their functions: HTML structures content, CSS changes styles, and JavaScript controls application behavior. However, some students confused the roles of these languages, suggesting the need for more examples to help them clearly distinguish their purposes.

4.2.3. Theme 3: Students’ Perception of Artificial Intelligence

Conversations with students revealed important findings regarding their understanding of artificial intelligence and its application in education. Most students had a basic understanding of artificial intelligence, though the depth of their understanding varied. Some described artificial intelligence as a trained network designed to perform certain tasks, while others referred to robots and automated systems that make decisions on their own.

Students’ understanding of neural networks was more limited compared to their understanding of artificial intelligence. A few described a neural network as “a network that works like the human brain”, while others were unsure of its meaning. These responses suggest that neural networks are a rather complex topic for most students and that further explanations are needed to clarify how they work. Students who provided more specific descriptions often related neural networks to examples from their projects, such as image recognition. This suggests that practical examples helped students grasp a more complex topic.

When asked whether artificial intelligence should be introduced into the school curriculum, most students expressed a positive attitude. They believed that learning about artificial intelligence should be part of their education because it would prepare them for the future, especially now that artificial intelligence has become part of everyday life.

4.2.4. Theme 4: Perception of Content and Teaching

With the video lessons that were on the EDIT portal, specially prepared for the students by the university teachers, the students showed their satisfaction. Many stated that the lessons were clearly explained and easy to follow and understand. Video lessons played an important role in independent learning, which allowed students to watch them from home at any time. Some students particularly liked the flexibility that was provided. Several students mentioned that they would like more supplementary video content that would go into more detail, which would perhaps expand their deeper interest in a particular topic.

Programming tasks, especially those related to JavaScript, were well received, although students had different views on their difficulty. While some found them easy and wanted even more challenging tasks, others considered them quite difficult. The variation in student responses primarily depended on their prior programming experience. Students found the assignments useful, as they helped them work more effectively on their projects, and some expressed interest in additional assignments to reinforce what they had learned.

The overall feedback on the program was very positive. Students believed that the module met their expectations. They were satisfied with the content, teaching style, and practical tasks that gradually introduced them to the key steps involved in creating an application. Although some students felt overwhelmed at times, they later recognized the value of such assignments. The students’ results and comments indicate that the module was well designed and aligned with their needs and interests. This feedback highlights the need for additional support at certain stages but also demonstrates a high level of satisfaction and student engagement.

4.2.5. Theme 5: The Impact of the Module on Students

Increased enthusiasm for programming among students after attending this activity was evident. Many students stated that they enjoyed programming while working on practical assignments and the final project. By tackling real-world challenges, they had the opportunity to apply their theoretical knowledge in a practical way, helping them better understand how algorithms and computer programs work. They mentioned that bringing theory to life through application development sparked additional interest in programming. The practical aspect of the module deepened their curiosity and motivation for further learning.

Many students expressed an interest in programming as a valuable skill for the future. Some stated that they were already engaging in programming in their spare time, further strengthening the connection between what they learned in the module and their personal interests.

Most students agreed that they gained a significant amount of programming knowledge through this module. They felt that the practical tasks enhanced their understanding of programming concepts and provided a solid foundation for expanding their knowledge with more complex programming structures.

Students also reported that they learned much more than they had expected. In addition to programming, they gained insight into how technologies function and solve problems. They believed that this module would make learning programming easier for them in the future. This sense of confidence is particularly significant, as it suggests that the module not only succeeded in imparting knowledge but also motivated them for further learning.

4.2.6. Theme 6: Environment and Influence

Most students reported strong family support for their participation in the module. Parents encouraged them to attend regularly, creating a sense of security and motivation to engage in such educational activities. This support was crucial for many students, especially those who felt insecure about their technical skills. Parental involvement was often evident through their interest in the module’s content, which was passed on to the students, giving them a sense of a safe and positive environment where they could share their experiences.

Interviews revealed that students communicated with family members and friends about what they had learned through the module. Many noticed that their friends became interested in participating in this activity after hearing about their experiences. This type of communication had a dual effect: in addition to strengthening the connection between students and their environment, it also sparked greater interest in computer activities among peers. Sharing their experiences with friends and family contributed to their sense of achievement and pride.

Student activities and projects were often covered by the media. Students stated that this contributed to their sense of pride, particularly in front of family and friends. Presenting their projects to the public not only gave them a sense of achievement but also strengthened their motivation to engage in similar activities in the future. This experience also reinforced their perception that their efforts were recognized outside the classroom. This external factor illustrates how media presence can positively impact self-confidence and students’ perception of their own abilities.

4.2.7. Theme 7: Motivation to Participate

One of the main motivating factors was students’ personal interest in programming applications and learning modern technologies. Many students expressed a desire to participate because they were curious about how to program apps and how to use artificial intelligence to recognize objects in the real world. Some mentioned the influence of their parents on their participation, but this was mostly an additional incentive rather than the primary motivation. While attending the module, they reported increased motivation due to the interesting and enjoyable content, as well as a friendly learning environment.

Although it may seem that friends played a significant role in students’ decision to attend the module, most indicated that they would have participated regardless of their friends’ involvement. This suggests that students were primarily motivated by their own interest and desire to learn, demonstrating a strong level of intrinsic motivation. Since students did not receive grades for the subject, this further supports the idea that their motivation was driven by personal interest. Additionally, students viewed the activities as an opportunity to acquire skills that would benefit them in the long term, especially in the future.

Almost all students stated that they would like to participate in a similar activity next year, reinforcing the positive experience of this module. Many mentioned that they do not have the opportunity to learn such useful skills in regular classes. This positive feedback is particularly important, as it enhances the module’s appeal in attracting students and strengthening their engagement through hands-on activities. Many students also said they would recommend the activity to their friends, further reflecting their level of satisfaction. This word-of-mouth recommendation is extremely valuable for the future of such activities, as it confirms their quality and attractiveness from the students’ perspective.

Although students acquired technical skills through this activity, they also demonstrated the development of responsible and independent learning. This approach, which combines entertainment and education, can serve as an excellent model for future initiatives aimed at increasing students’ interest in IT fields.

4.2.8. Theme 8: Challenges, Difficulties, and Student Suggestions

Most students indicated that they had no significant difficulties during the module, although they found some parts of the teaching content more challenging. Many cited starting to work in JavaScript as the most difficult part, as they were encountering the language’s syntax and new concepts for the first time. Given that they were also experiencing project-based programming for the first time, the initial stages of creating a project—with all that it entailed—proved to be challenging as well. This is further evident from the fact that students’ initial project plans often differed from their final projects due to encountering unexpected outcomes during the process. Some students noted that greater patience was required for image recognition tasks, especially when recognizing images from a video camera. Despite these challenges, students stated that they successfully solved all problems with the help of their teachers and additional clarification.

Although most students were extremely satisfied with the overall program, some mentioned certain aspects they enjoyed less. For example, some found unplugged activities, especially paper-based tasks, less engaging than tasks on the computer. Others felt that some lessons were too long or too detailed, which occasionally caused them to lose interest.

Additionally, some students were not enthusiastic about the video presentation of their project because they felt uncomfortable, although they acknowledged that such an activity helped improve their presentation skills and allowed them to showcase their project more effectively.

Students also proposed several suggestions for improving the module. Some suggested adding more fun and interactive tasks to increase engagement, while others preferred lessons that focused exclusively on practical tasks. However, most students did not want major changes, emphasizing that they enjoyed the entire program. Their suggestions primarily reflected small adjustments that would make the program a more dynamic and enjoyable experience.

5. Discussion

The research showed that artificial intelligence concepts can be successfully integrated into primary schools through project-oriented activities. By using artificial intelligence, students adopted key programming concepts through practical projects, providing momentum for the broader application of AI in education. However, challenges were noted in understanding more complex concepts, such as neural networks, suggesting the need for additional educational materials.

Similarly, research conducted in Finland identified students’ misconceptions about artificial intelligence, particularly the belief that AI is entirely pre-programmed. This highlights the need for a more specific approach to teaching AI at an early age to prevent such misconceptions [12]. The results can be compared with research in other countries.

In Sri Lanka, an artificial intelligence curriculum implemented in secondary schools has been shown to improve students’ motivation and their ability to apply AI in problem-solving [46]. A transdisciplinary approach to artificial intelligence education through an international program in Saudi Arabia has demonstrated the connection between AI and other disciplines, enabling a better understanding and broader application of AI concepts across various fields [47].

Students who have undergone intensive AI programs, such as boot camps, have shown a significantly better understanding of AI concepts and improved programming skills, emphasizing the importance of informal educational programs that complement traditional classes [48]. In Ukraine, AI literacy is still in its early stages, and teachers emphasize the need for a flexible learning approach based on a strong foundation in basic concepts to bridge gaps in the use of AI tools [49].

Interviews revealed that students were satisfied with all aspects of the RIWA module. The activity allows for a combination of technical and creative work, making the module more accessible to students with different interests and levels of prior knowledge. Other studies also highlight the positive effects of similar programs based on problem-solving, awareness of AI’s societal impact, and skills that prepare students for the technological world [1].

During the module, students engaged in various activities in different ways, including theoretical lessons, problem-solving tasks, unplugged activities, and programming using different devices and software. The results emphasize the importance of combining theoretical concepts with practical tasks to achieve the highest level of student engagement. Although students showed good engagement during unplugged activities through lighter and more enjoyable tasks, interest in these activities tended to fade more quickly compared to tasks involving digital tools. The final impression is that students prefer programming on the computer over theoretical lessons and unplugged activities.

5.1. Projects

Project-based AI literacy programs demonstrate improvements in problem-solving using artificial intelligence, thus preparing students for the practical challenges of the future [50]. Through projects, students apply key concepts of computational thinking, such as variables, decision-making, branching, and repetition, which are integrated into programming and emphasized in the Computational Thinking and Programming domain of the Informatics curriculum. By incorporating artificial intelligence into projects, such as the development of object recognition applications, students enhance their understanding of more complex concepts and are motivated to use more creative solutions. This approach exposes students to complex and authentic problems without a predefined “correct” answer, encouraging collaboration, knowledge application, and the development of self-regulated learning skills [51].

Practical tasks deepen students’ understanding of artificial intelligence concepts [52]. The efficiency of project applications largely depends on the quality of the input dataset used to train the neural network. Students become more aware that the accuracy of the output depends on the number and quality of training examples. A useful analogy is found in mathematics—the more problems we practice with, the more likely we are to solve them correctly. Similarly, students develop an understanding of how algorithms function in the context of intelligent systems, encouraging them to explore computing and artificial intelligence more deeply.

Students are often more interested in projects with real-world applications, such as those utilizing artificial intelligence, because they interact with such technologies daily, for example, through facial recognition on mobile devices. These types of projects can help students develop an awareness of the underlying algorithms behind such systems. As artificial intelligence becomes an increasingly prominent part of technology and society, introducing these topics equips students with the skills needed to participate in the digital world, aligning with the E-Society domain in the Informatics curriculum.

Additionally, as part of the Computational Thinking and Programming domain, problem-solving skills are further strengthened through the application of AI-related tasks, empowering students with competencies essential to the educational system. Through this process, students not only learn basic programming structures but also gain an understanding of how these structures form the foundation for the complex intelligent systems that surround them.

5.2. Pedagogical Approach and Influence of Teachers

The students’ feedback on the program was very positive, and they noted that the module met their expectations. Students expressed satisfaction with the teaching methods, content, and practical tasks that gradually guided them through the application development process. Research indicates that students with more developed AI literacy tend to have a more positive attitude toward artificial intelligence [40].

Due to the positive impact of artificial intelligence on students’ AI and digital literacy and their interest in technology, such flexible methods and resources are recommended for more effective implementation in various school subjects [1]. Some students stated that they occasionally felt overwhelmed, but they enjoyed the experience and recognized the benefits that such challenges brought them. This feedback highlights the need for additional support during certain phases of the program while also demonstrating a high level of student engagement and satisfaction.

The results suggest that the module is well designed, with a balanced combination of theoretical content and practical tasks adapted to the diverse needs of students. AI curriculum designers should consider students’ intrinsic motivation before they even begin learning about AI, as well as how experiences and career aspirations shape their long-term interest in continuing to explore AI after the program ends [53].

An effective curriculum should motivate all students to learn AI, as their intention to engage with AI influences the effectiveness of programming components, including logical thinking, algorithms, and debugging [54].

The successful introduction of AI literacy into the existing curriculum depends primarily on the quality of prepared materials and the knowledge and readiness of teachers responsible for teaching students. However, caution is needed when integrating such a complex topic into regular classes. AI literacy education must take into account students’ preconceptions about AI, as initial conceptions provide insight into student understanding and can help curriculum designers address misconceptions that could hinder learning [12].

To introduce AI literacy, we incorporated it into an extracurricular activity following the “from concrete to abstract” principle. By starting with something familiar to them—such as object recognition—and working on a project, students gained an impression of how this initially abstract process functions in the background. Some components, such as the ml5.js library, remained a “black box” to students, but in this case, it served as a motivational tool to engage them with artificial intelligence and help them acquire programming and digital skills. Beyond Informatics, AI literacy can be integrated into other subjects, regardless of format. The national curriculum for primary schools in the Republic of Croatia includes cross-curricular topics that span multiple subjects, making AI literacy adaptable across different disciplines.

AI literacy should not be viewed in isolation but as a concept applicable across all subjects, considering that artificial intelligence has become embedded in nearly every aspect of society. For this reason, teacher education is crucial—not only for computer science teachers but for all educators. The success of AI literacy adoption, as well as the development of computational thinking and programming, depends on a teacher’s readiness, motivation, and knowledge [55]. Although all teachers participating in the EDIT extracurricular activity used the same materials provided through the RIWA module, different teaching approaches resulted in varying student outcomes. Some teachers placed less emphasis on programming concepts, which proved to be significant, as it led to lower-quality student projects. This reinforces the idea that teacher readiness and expertise are key factors influencing student success.

A study conducted in Kenya found that teachers’ attitudes toward AI, self-confidence, and ethical considerations significantly affect their willingness to teach AI. However, the lack of specific training remains a major obstacle, highlighting the need for professional development in AI education [56]. It is crucial to support teachers in integrating AI tools as a teaching resource to boost their motivation and self-confidence [57]. Professional development and teacher support enhance their ability to adapt lessons to technological advancements [58]. The development of technological–pedagogical knowledge among teachers is essential for effectively using AI tools in teaching, defining pedagogical contexts for student activities, and integrating AI learning into the existing curriculum [32].

The teacher–student relationship also plays a significant role. Students who already knew the teacher demonstrated greater knowledge in project tasks, a broader scope of work, and more consistent attendance. In contrast, teachers conducting extracurricular activities for students they had not previously taught faced challenges in motivation, establishing hierarchy, and maintaining authority. Online classes also proved to be an additional obstacle for some students. Issues such as technological limitations at home, unstable internet connections, and the teacher’s inability to monitor student progress effectively resulted in lower engagement and poorer outcomes for students in these groups.

No approach is perfect on its own, but with high-quality materials, well-prepared teachers, and strong motivation, a solid foundation can be established for integrating AI literacy into regular classes. If an analogy is made with something familiar to students, the learning outcomes are generally positive.

5.3. Implications

The findings of this study provide several important implications for teachers, curriculum designers, and educational policymakers. The integration of artificial intelligence concepts into the primary and secondary school curriculum is crucial for preparing students for a future shaped by these technologies. By incorporating artificial intelligence into problem-based tasks that allow students to express themselves creatively, they can more naturally acquire the necessary skills for an evolving digital landscape.

Additionally, our experience with the RIWA module suggests that AI-focused lessons can be incorporated into the existing Informatics curriculum without adding to the total number of instructional hours. To mitigate potential issues with teacher workload and time management, it is essential to provide high-quality, ready-made teaching materials, offer professional development tailored to AI concepts, and adopt a modular structure that aligns AI projects with existing programming or digital literacy outcomes. For broader scalability, these resources can be packaged into a standardized training kit—encompassing video tutorials, ready-made lesson plans, and a moderated online discussion forum—allowing schools in different regions to adapt the program with minimal additional overhead. This combination of concise instructional materials, periodic workshops, and continuous support fosters a sustainable model for introducing AI concepts across diverse educational settings.

The positive perception of the RIWA module by students indicates that project-based learning has a significant impact on engagement and motivation, especially when learning topics such as artificial intelligence and programming. Therefore, schools and educational systems should invest in continuous teacher training and provide educators with comprehensive resources relevant to their subject areas.

Although the RIWA module demonstrated strong engagement and skill development, certain challenges were observed. Students found hands-on AI projects engaging, but some struggled with abstract AI concepts such as neural networks. Teacher preparedness also played a key role, as additional training was required for effective AI instruction. Furthermore, project outcomes varied due to dataset limitations, sometimes reinforcing biases in AI models. Future iterations should focus on refining instructional materials, incorporating unplugged AI activities, and strengthening teacher training to address these challenges while maintaining the benefits of project-based AI learning.

Parental involvement also played an important role in student motivation. Many parents attended the final ceremony, where students received certificates, and the best projects were awarded. Additionally, all finalized projects were made available online, and students created video presentations showcasing their work. Parents expressed enthusiasm and pride in their children’s achievements, highlighting the value of early AI education. Expanding media coverage of such activities could further increase awareness and support for AI education, potentially encouraging broader adoption beyond the initial study region.

This study highlights the importance of collaboration between educational institutions and the IT industry. By connecting policymakers, teachers, and industry experts, students can gain an improved understanding of current technological developments and be better prepared for the demands of the modern workforce.

5.4. Limitations

This study provides valuable insights but also has its limitations.

The research was conducted as an extracurricular activity in 12 schools within Split-Dalmatia County, meaning that participation was voluntary and subject to self-selection bias. Since Informatics is an elective subject in the seventh and eighth grades, students who participated may have already had a higher interest in programming and technology. This limits the ability to generalize the findings to all students and may have influenced motivation and engagement levels. Additionally, because the activity was outside the regular curriculum, it was not feasible to randomly assign students to a control group, making direct comparisons between AI-enhanced and traditional programming instruction difficult. Future research should explore ways to conduct controlled experiments within the formal education system where structured comparisons may be more feasible.

Another limitation is that this study focused primarily on motivation and engagement rather than in-depth assessments of learning outcomes. While some additional pre- and post-test data were collected, as well as comparisons with students not enrolled in the module, these analyses fall outside the scope of this paper and will be presented in separate studies. Given the extracurricular nature of the intervention, we prioritized qualitative insights, project analysis, and student reflections over extensive standardized testing. However, future studies may incorporate additional quantitative measures such as knowledge retention assessments and creativity metrics.

Additionally, while this study captured immediate learning outcomes over three years, it did not track participants’ long-term academic or career trajectories. Given the constraints of an extracurricular setting, a full longitudinal analysis was beyond the scope of this study. However, student performance in regular Informatics courses may provide an indirect indicator of retention, which future research may explore further.

Finally, this study was conducted in a specific educational and cultural context. While the methodology and findings may be applicable to other regions, variations in curriculum structure, teacher training, and technology access may require localized adaptations. Future work should explore how AI-based instructional methods can be effectively scaled and adapted for different educational settings.

6. Conclusions

The research was conducted through the EDIT extracurricular activity with the Development of Intelligent Web Applications (RIWA) module. The results demonstrate the feasibility of successfully integrating artificial intelligence into primary school education. Through practical examples and project-based learning with the use of advanced technology for object recognition, students acquired basic computer thinking and programming skills, which is an integral part of the Informatics curriculum. The students applied the acquired knowledge through the projects, so they most often used programming concepts such as variables, conditional statements, loops, and functions. Although the students encountered different challenges in the accuracy of object recognition, such challenges provided them with the value of how important quality data are for training models, one of the key elements in working with artificial intelligence.

The interviews revealed a high level of student satisfaction with the module. Most students expressed pride in their projects and enthusiasm for further work in programming. This positive feedback underscores the value of a module that aligns with student interests. The importance of hands-on tasks and real-world applications was repeatedly emphasized by students, who reported that these aspects increased their motivation.

However, integrating artificial intelligence into the regular curriculum requires careful planning, additional teacher training, and adaptation of teaching materials. While students demonstrated interest and capability in learning AI concepts, their success depends on teacher preparedness and the availability of high-quality educational resources. The RIWA module effectively introduced students to AI concepts through project-based learning, but interviews highlighted varying levels of understanding, particularly regarding neural networks. To enhance comprehension, future iterations of the module could incorporate unplugged activities to illustrate key machine learning principles intuitively. Linking AI concepts to everyday applications may also help students better grasp their practical relevance. Additionally, integrating interactive AI simulations—where students modify model parameters in real time—would provide a more structured, engaging, and hands-on approach, reinforcing conceptual understanding.

Beyond technical skills, future iterations of the module should include a dedicated component on ethical AI considerations, such as data privacy, bias in AI models, and responsible data collection. Since the Croatian National Framework Curriculum (NOK) [59] already covers digital citizenship and responsible technology use, integrating these topics into the module would further align with global AI education standards while fostering students’ awareness of ethical challenges in AI applications.

To ensure broader implementation, future efforts should focus on expanding teacher training, refining curriculum integration, and developing scalable support materials. Standardized training kits, including ready-made lesson plans and interactive simulations, could help educators integrate AI concepts more effectively. Additionally, adapting AI education to different educational systems will require flexible teaching approaches, ensuring accessibility in diverse contexts. The findings of this study highlight the potential for incorporating AI literacy into national curricula, which aligns with digital education policies and fosters collaboration between educational institutions and industry. Future initiatives should explore partnerships with policymakers to promote AI education as an essential component of modern informatics curricula.

In conclusion, the RIWA module presents a viable approach for integrating artificial intelligence into the Informatics curriculum for primary schools. With additional adaptations, AI can become an integral part of school education, preparing students for the challenges and opportunities of the digital age.