AI Education Framework for Primary Schools Unveiled in Foshan Study
In a significant step toward integrating artificial intelligence (AI) into foundational education, researchers from Foshan University have developed a comprehensive teaching framework designed to bring AI literacy into primary school classrooms. The study, led by Zhang Jinyan, Lin Sufang, Ma Mengyi, and Li Xinhui of the College of Humanities and Education at Foshan University, presents a structured approach to teaching AI within the constraints of existing information technology curricula, limited instructional hours, and a shortage of specialized educators. Published in the July 2021 issue of Modern Information Technology, the research offers a practical model that balances conceptual understanding, hands-on technical experience, and creative application—three pillars essential for cultivating future-ready digital citizens.
As AI technologies such as facial recognition, voice assistants, and autonomous systems become increasingly embedded in daily life, the need for early exposure to these concepts in education has gained momentum globally. Governments and educational institutions are recognizing that AI literacy is no longer optional but a core component of 21st-century learning. In China, the push began with the 2017 release of the State Council’s New Generation Artificial Intelligence Development Plan, followed by the Ministry of Education’s Higher Education Artificial Intelligence Innovation Action Plan in 2018. These policy directives emphasized the importance of building a seamless AI education pipeline from primary schools through higher education.
Despite this top-down support, the implementation of AI education in primary and secondary schools has been uneven. Many schools struggle with fragmented content, lack of standardized curricula, and insufficient teacher training. In Foshan, a major city in Guangdong Province, these challenges are particularly evident. While some schools have introduced AI-related topics such as speech synthesis, image recognition, and augmented reality, the absence of a cohesive framework often results in isolated, disconnected lessons that fail to build a deep or sustained understanding.
To address these issues, the Foshan University team conducted an extensive field investigation across multiple primary schools in the region. They analyzed current teaching practices, evaluated the quality and consistency of instructional content, assessed teacher preparedness, and measured student learning outcomes. Their findings revealed a clear gap between policy aspirations and classroom realities. Most AI instruction was ad hoc, relying heavily on commercially available tools and platforms without a clear pedagogical roadmap. Teachers, many of whom were trained in general information technology rather than AI-specific domains, expressed uncertainty about how to teach complex concepts like machine learning or natural language processing in age-appropriate ways.
The researchers also conducted a comparative literature review, examining AI education initiatives in countries such as the United States, the United Kingdom, and Singapore. These international models often emphasize project-based learning, computational thinking, and interdisciplinary integration—approaches that align with broader educational reforms. However, the team noted that direct replication of foreign models is not feasible due to differences in curriculum structures, resource availability, and cultural context.
Drawing from both local insights and global best practices, the authors proposed a three-tiered AI curriculum framework tailored for Chinese primary schools. The framework is built on three interconnected dimensions: understanding AI concepts, experiencing AI technologies, and creating AI-powered projects. This tripartite structure ensures that students progress from passive consumers of AI to active creators and critical thinkers.
The first dimension, related concepts, focuses on building foundational knowledge. At the primary level, the goal is not to teach advanced algorithms but to cultivate an intuitive understanding of what AI is and how it differs from traditional programming. Students learn that AI systems can “learn” from data, make predictions, and improve over time. They explore key areas such as computer vision, natural language processing, and machine learning through age-appropriate analogies and real-world examples. For instance, they might compare facial recognition to how a teacher remembers students’ faces, or explain speech recognition as a machine “listening” and “understanding” human language.
The second dimension, technical experience and application, emphasizes experiential learning. Rather than diving into complex coding, students interact with AI through user-friendly platforms that allow them to test and observe AI behaviors. The study highlights the use of tools like the iFlytek Open Platform, where students can experiment with voice recognition, and block-based programming environments such as Scratch and programmingCat IDE, which integrate AI modules. These tools lower the technical barrier, enabling even young learners to engage with AI without needing deep programming expertise.
One of the most innovative aspects of the framework is its emphasis on intelligent work creation as the third dimension. Here, students move beyond passive use and begin designing their own AI applications. This phase is rooted in project-based learning (PBL), a pedagogical approach that encourages inquiry, collaboration, and problem-solving. Students work in small groups to identify real-world problems and develop AI-enhanced solutions. For example, they might design a “smart home” system that responds to voice commands, or create a language translation tool that converts spoken Chinese into English and vice versa.
The researchers illustrate this approach with a detailed lesson plan on “Speech Recognition,” drawn from the fifth-grade information technology curriculum in Foshan. The lesson begins with an engaging demonstration using iFlytek’s voice recognition platform, allowing students to see how their spoken words are instantly converted into text. This initial experience sparks curiosity and sets the stage for deeper exploration.
In the next phase, students transition to programmingCat IDE, a visual programming environment that includes built-in AI commands. The teacher models how to use voice recognition blocks, comparing them with standard input commands to highlight the differences between rule-based programming and AI-driven interaction. Students then work in heterogeneous groups of three, ensuring a mix of skill levels, to develop their own programs. They are given project choices such as creating a “Translation Assistant” for travelers or designing a “Smart Home” control system.
This hands-on phase is critical for developing computational thinking—the ability to break down problems, recognize patterns, and design algorithmic solutions. By using graphical programming blocks, students grasp abstract concepts like input processing, conditional logic, and output generation in a concrete and intuitive way. More importantly, they begin to see AI not as magic but as a set of tools that can be understood, manipulated, and applied creatively.
The final stage of the lesson involves peer evaluation and reflection. Each group presents their project to the class, explaining their design choices, demonstrating functionality, and discussing challenges they encountered. To ensure meaningful feedback, the researchers developed two evaluation rubrics: one for inter-group assessment and another for intra-group peer review.
The inter-group rubric evaluates projects based on criteria such as planning quality, innovation, completion level, and presentation clarity. Each dimension is scored on a five-point scale, encouraging students to assess both technical execution and creative merit. The intra-group rubric focuses on collaboration, measuring individual contributions in terms of cooperation, participation, and problem-solving. This dual assessment strategy promotes accountability, fosters teamwork, and provides teachers with valuable insights into both group dynamics and individual learning.
The impact of this structured approach extends beyond technical skills. By engaging with AI in a hands-on, project-based manner, students develop a broader set of competencies aligned with modern educational goals. They enhance their information literacy by critically evaluating AI applications, strengthen their problem-solving abilities through iterative design, and cultivate creativity by imagining new uses for technology. Moreover, the collaborative nature of the projects nurtures communication and social skills, preparing students for the interdisciplinary and team-oriented environments of the future.
One of the most pressing challenges in AI education is teacher preparedness. The study acknowledges that most primary school information technology teachers lack formal training in AI. To bridge this gap, the proposed framework is designed to be accessible and scaffolded. It does not require teachers to become AI experts but instead equips them with ready-to-use lesson plans, curated resources, and step-by-step guidance. The use of visual programming tools and pre-built AI modules reduces the cognitive load on educators, allowing them to focus on facilitating learning rather than mastering complex technical content.
The researchers also recommend a tiered course model—labeled beginner, intermediate, and advanced—to support progressive learning. In the beginner phase, AI concepts are introduced through everyday applications such as voice assistants and smart cameras. The intermediate phase integrates AI into programming activities, where students use block-based coding to build simple AI applications. The advanced phase incorporates physical computing devices like micro:bit or KOI boards, enabling students to create tangible AI projects that interact with the real world, such as automated plant watering systems or motion-sensing security alarms.
This staged progression ensures that students build knowledge incrementally, avoiding the overwhelm that can come from introducing too much complexity too soon. It also allows schools to implement AI education incrementally, starting with low-cost, software-based activities before investing in more advanced hardware.
The study’s implications are far-reaching. As AI continues to reshape industries, economies, and societies, the ability to understand and work with intelligent systems will become a fundamental literacy—on par with reading, writing, and arithmetic. By embedding AI education into the primary curriculum, educators can ensure that all students, regardless of background, have the opportunity to develop these essential skills.
Moreover, the Foshan model demonstrates that effective AI education does not require expensive equipment or highly specialized teachers. With thoughtful curriculum design, appropriate tools, and a focus on experiential learning, schools can deliver meaningful AI instruction within existing constraints. The framework’s emphasis on creativity and problem-solving also aligns with broader educational goals, such as fostering innovation, adaptability, and lifelong learning.
Looking ahead, the researchers call for greater investment in teacher professional development, the creation of standardized AI curricula, and the establishment of national benchmarks for AI literacy. They also advocate for stronger collaboration between schools, technology companies, and universities to ensure that educational content remains current and relevant.
In conclusion, the Foshan study offers a compelling blueprint for bringing AI into primary education. By combining conceptual understanding, technical experience, and creative application, the proposed framework empowers students to not only use AI but to understand it, shape it, and ultimately, lead in an AI-driven world. As the boundaries between humans and machines continue to blur, such educational initiatives will play a crucial role in shaping a future where technology serves humanity—not the other way around.
Zhang Jinyan, Lin Sufang, Ma Mengyi, Li Xinhui, College of Humanities and Education, Foshan University, Modern Information Technology, DOI:10.19850/j.cnki.2096-4706.2021.14.047