3

Curriculum Development and Policy Implementation

The first panel focused on curriculum development and policy implementation and how different countries integrate data science, computational thinking, and artificial intelligence (AI) into K-12 mathematics education. It has become a global imperative, although curriculum design approaches vary widely across international contexts. For example, centralized systems employ processes that create coherent learning progressions, align instructional materials, and support consistent implementation. Decentralized systems with local decision making can foster innovation and responsiveness but can lead to uneven adoption or varied interpretations of national guidance.

International frameworks vary in their emphasis on algorithmic thinking, data interpretation, modeling, or AI ethics. However, regardless of specific focus, systems that explicitly connect mathematics learning to workforce readiness, civic decision making, technological fluency, and real-world problem solving often generate stronger support. When policy makers articulate how new mathematical competencies contribute to national economic goals, innovation strategies, or societal resilience, reforms tend to attract broader coalitions of stakeholders.

Adoption strategies range from rapid nationwide mandates to more incremental, evidence-driven approaches. The latter typically involves structured pilots, targeted professional learning, and curriculum refinement based on classroom-level evidence. While many countries aspire to this research-informed approach, maintaining the discipline required for iterative implementation remains a common challenge. Short political cycles, shifting priorities, and resource constraints frequently disrupt the continuity needed to build and sustain large-scale change. By examining both the conceptual rationales and the practical processes behind international reforms, leaders can craft strategies that are not only ambitious but also feasible, coherent, and responsive to the realities of classrooms.

This session was moderated by Padhu Seshaiyer (George Mason University) and Hollylynne Lee (North Carolina State University). Participants came from education systems in Scotland, Singapore, South Korea, and Australia. In brief presentations and discussions, they were asked to address

such questions as the policies essential to support curriculum design, barriers related to teacher capacity or technological access, accommodating schools with varying levels of resources, and how they foresee curriculum revisions in their countries in the future.

SCOTLAND: COMPUTING, DATA, AND AI EDUCATION

Kate Farrell (University of Edinburgh) briefly described the structure of Scotland’s education system before explaining how the system is addressing computing, data, and AI. It encompasses early education (ages 3 to 5), primary (P1 to P7, ages 5 to 12), and secondary (S1 to S6, ages 12 to 18). At age 16, students make subject choices in which to specialize, and this is also when mandatory education ends. Free college and university is available for most Scottish residents. Education is devolved in the United Kingdom, so Scotland makes its own decisions around education.

The curriculum rests on four fundamental pillars to enable young people to become successful learners, confident individuals, responsible citizens, and effective contributors.¹ There are eight curricular areas, including technologies, which is where computing lies. Computer science began as a secondary-school elective; about 10 years ago, the National Curriculum outcomes were updated to include it for all learners age 3 and up. A spiral curriculum was introduced in which younger students begin with loop-like playground games, then go on to understand the concepts behind the games and to create their own ideas using computers.

Since 2019, the University of Edinburgh has been involved in a data literacy project in southeast Scotland. Recognizing that a new subject could not just be added to the curriculum, they instead looked at where data literacy already resided in the existing curriculum, such as math, technology, and social studies, and found opportunities to apply data literacy skills across the curriculum. A guide was developed for primary teachers.² National qualifications in data science were developed for secondary learners, with support guides and materials for learners and educators.³ The goal, Farrell said, is that any teacher in any subject can teach these skills, not just computing or math teachers. For example, students may enhance

___________________

¹ The Scottish Government. June 2008. Building the Curriculum 3: A Framework for Learning and Teaching. https://education.gov.scot/media/0cvddrgh/btc3.pdf

² Teaching Data Literacy: A Guide for Primary Teachers. Available at https://dataschools.education/teachdata/

³ For example, Data Science: Educators Guide to the National Progression Award (https://dataschools.education/wp-content/uploads/2020/11/educators-guide_Oct2020.pdf) and Data Science: Learners Guide to the National Progression Award (https://dataschools.education/resource/guide-for-learners-studying-the-npa-data-science/).

their data literacy skills through projects about energy use, health, or weather.

Curriculum outcomes for AI literacy were created based on existing National Curriculum outcomes. The first step was to solicit feedback from learners. Their comments included that AI should be in the curriculum since it will be in their lives, but that they need to know how to keep themselves safe.⁴ Another important step was to address how AI in education fits in with the United Nations Convention on the Rights of the Child. Children’s rights are at the heart of the AI framework being developed, Farrell stressed.

Learners need core building blocks before they can use AI responsibly. Resources include a guide to teach AI literacy. AI is used differently in different subject areas, such as science versus religion, and teachers have developed exemplar lessons in different subjects.⁵ Alongside this, she and her team developed core lessons in how learners can use different AI systems and understand how they work, as well as critical thinking skills. Issues related to sustainability and bias are also covered. A board game was developed that challenges learners to run a data center without running out of power and water. It was recently announced that the University of Edinburgh will house Scotland’s Center for Teaching Excellence AI Hub. Farrell expressed the hope that the Center can develop more resources and work with more teachers in different areas.

SINGAPORE: AI IN EDUCATION

Kang Hao Cheong (Nanyang Technological University) underscored that AI has moved from a technical tool to something that is part of learning, and thus governance is needed. Across education systems, AI tools are being introduced in Singapore in calibrated ways for student practice, feedback, and explanations, used by both teachers and students. AI is not just a new tool or software, he stressed. It changes the learning environment. The key question is not what AI can do but how an education system can introduce AI in a way that supports learning while preserving educational responsibility. “Governance frameworks provide an important foundation for responsible implementation and professional development,” he said.

Cheong distinguished between general AI and educational AI. A sufficiently powerful generative AI model is not automatically suitable for education. For example, general-purpose generative AI is optimized to generate answers, but educational AI should prioritize reasoning, understanding,

___________________

⁴ Colton Botta et al. 2025. Secondary students as co-researchers on generative AI in learning: Empowering youth to shape national educational policy. UKICER ‘25, 1:1–7. https://dl.acm.org/doi/10.1145/3754508.3754519

⁵ Resources are available at TRAILS.scot.resource

and metacognition. A correct answer can be a poor outcome if it bypasses reasoning, he pointed out. In his view, educational AI calls for different criteria than general AI: Does it promote thinking? Does it support understanding? Does it build learner independence? It is important to set very clear task boundaries and teacher-guided workflows, and not just better AI models per se.

In Singapore, educational AI is positioned as part of the broader teaching and learning system, not just a standalone tool. An assessment structure already exists. When AI enters the classroom, the key question is what role it plays to support learning. Instruction remains teacher led with AI intended to support professional judgment. The Ministry of Education has developed an “AI-in-Education (AIEd) Framework” to offer practical guidance to principals and teachers.⁶ It helps them evaluate use cases and anticipate and assess risk in order to use AI in ways that align with learning quality and student development. It has four core principles to define the appropriate role of AI in education: agency, inclusivity, safety, and fairness.

For example, focusing on the principle of safety, the idea is not to eliminate AI use but to ensure that the education system can manage and take responsibility for it through governance mechanisms and usage boundaries. The first and foremost assumption is that AI is not always correct and should not be used without human judgment for high-stakes educational decisions. There will always be risk, but governance mechanisms can ensure data protection, visibility, and institutional accountability.

Singapore’s approach suggests that successful educational AI depends not only on technical capability but also on how AI is integrated into teaching practices, curriculum design, and school processes, he underscored. A system design approach calls for establishing the governance first, then deployment within that structure. Implementation is not a one-time rollout, he stressed, because technology changes quickly. Guidance and professional development must be sustained.

Looking to the future, the focus in Singapore will likely shift from generating content to supporting the learning process. AI may increasingly operate as part of the instructional system aligned with curriculum, teachers, and assessment, he predicted. However, pedagogy must come first, with age-appropriate boundaries and clear accountability. Cheong concluded with two warnings. First is to guard against cognitive offloading by designing tasks that require reasoning, as well as keeping teachers in the loop. Second, it is important to protect students’ well-being with appropriate boundaries, clear pathways to trusted adults, and support channels for social-emotional learning, counseling, and other services. Educational AI

___________________

⁶ For more information on the framework, see https://www.moe.gov.sg/education-in-sg/educational-technology-journey/edtech-masterplan/artificial-intelligence-in-education

success is not just about models, model generation, or model capability. It is about governance and pedagogy to support learning in a reliable manner.

SOUTH KOREA: CURRICULUM DEVELOPMENT AND POLICY IMPLEMENTATION FOR COMPUTATIONAL THINKING, DATA SCIENCE, AND AI

Kyungwon Lee (Seoul National University; Dankook University Middle School) explained South Korea’s educational system is organized around a national curriculum that is revised about every 6 years. The most recent revision, in 2022, responded to the era of digital transformation and aligned with the main directions of the Organisation for Economic Co-operation and Development (OECD) 2030 report about the future of education.⁷ The main themes of that report included ecological transition and democratic citizenship education, as well as digital literacy. South Korea’s mathematics curriculum incorporates these directions. Digital literacy is positioned as a basic literacy for future generations. The framework for mathematics is built around cultivating key general competencies (such as knowledge and understanding), mathematical content domains (such as probability and statistics), mathematical competencies (such as problem solving and communication), and instructional and environmental features (such as utilizing digital learning environments and digital tools). Data science, computational thinking, and AI are not explicitly mentioned in the mathematics curriculum, but several components create meaningful connections to these areas.

Elective high school mathematics courses, such as practical statistics, mathematics for economics, and AI mathematics, serve as entry points for integrating data-related inquiry and AI explorations. Practical statistics introduces data science, while AI mathematics focuses on mathematical concepts related to AI, for example matrices, vectors, and functions and derivatives. Students understand AI through these underlying mathematical ideas and not just as a technological tool. Mathematics tasks related to computational thinking incorporate core processes, including decomposition, abstraction, and algorithmic thinking.

Infrastructure plays a key role in implementation of the curriculum. One-to-one tablet initiatives and a range of locally developed mathematics software tools are widely used, including AlgeoMath (graphing and block-based coding tools), Easy Statistics, Tongrami (statistical tools), and Tok Tok Math (AI-based personalized and game-based learning). All help translate curriculum intentions into mathematics classroom practices.

___________________

⁷ OECD. 2022. Future of Education and Skills: Education 2030. https://www.oecd.org/content/dam/oecd/en/publications/reports/2018/06/the-future-of-education-and-skills_5424dd26/54ac7020-en.pdf

Another initiative, AI Digital Mathematics Textbooks, supports personalized learning through diagnostic assessment and adaptive text recommendations. The shift from textbooks to supplementary materials, due to cost, technical constraints, and concern about student exposure to AI-mediated environments, illustrates the complexity of policy implementation. Generative AI has introduced a new layer of transformation. Teachers have been provided with training and resources. At the same time, policy guidelines have been developed at the metropolitan, provincial, and national levels to address ethical considerations in the implementation of AI in school settings.

Lee looked ahead with two summarizing points. First, the South Korean national curriculum has created multiple opportunities to connect mathematics with data science, computational thinking, and AI through curriculum documents and textbooks in response to evolving societal demands. Future curriculum revisions should move beyond the inclusion of new content toward the redesign of mathematical tasks and teacher professional learning in response to these developments. Second, because the 2022 revision was developed before generative AI became widespread, future revisions will need to include more explicit national guidance. In the meantime, greater efforts are required to disseminate mathematics lessons, teacher professional development programs, and research and practice-based evidence, which can serve as a foundation for subsequent curriculum revisions.

AUSTRALIA: THE JOURNEY TO INTEGRATE COMPUTATIONAL THINKING, DATA SCIENCE, AND AI IN MATHEMATICS

Justine Sakurai (University of Melbourne) shared several iterations of policy sequencing by the Australian government. Before 2020, she noted, teachers began introducing computational thinking into their classrooms. They helped inform curriculum (standards) development, and the professional learning infrastructure was developed from the ground up by teachers. During a 2020–2022 curriculum redesign, interested teachers had the chance to be deeply involved, along with industry and academia. All states and territories in Australia agreed to use the new national curriculum with full implementation in 2026. Computational thinking was integrated into mathematics across the curriculum from K to 10. In other words, before 2020, it was teacher led, but from 2022, it was curriculum led. In addition to embedded computational thinking, statistics and probability include explicitly stated data science concepts where possible. Years 9 and 10 offer a data science elective, although student uptake has been low. Supporting resources include cross-curricular links, lesson plans, and a mathematics hub.

Two states, South Australia and Victoria, illustrate two different ways that states have implemented the national curriculum, she explained.

South Australia opted for what she termed a pragmatic use. The South Australia Department for Education and Microsoft developed an EdChat

trial, a custom educational chatbot in 2023. It is a closed system with data stored in Australia and not shared with ChatGPT servers. It is not as accurate or effective as open AI but safer, she noted. In the trial, 93 percent of prompts were curriculum related, but fewer than 10 percent were in mathematics. Some students used it as a tutor or teacher, for example to probe alternative explanations of complex problems or practice questions with adjustable difficulty.

Victoria adopted a more cautious approach and chose not to use a large language model. Instead, computational thinking was embedded in the curriculum at all grade levels, and this has been effectively delivered in the curriculum since 2010. A generative AI policy was developed in 2024. Victoria also made the first attempt to explicitly integrate systems thinking and data science into “systematics,” which has been implemented in a vocational mathematics course. It is a nonacademic course in the final 2 years of schooling aimed at using technology to solve real-world mathematics problems. It embeds computational thinking into mathematics while preparing students for the automated systems they will encounter in the workplace.

Sakurai identified several challenges related to policy, curriculum, and teachers’ capacity. Australia is a big country geographically with a small population and there is policy dissonance between the states and territories. As elsewhere, teachers express concern that they have too much to teach with insufficient time. It is important to build capacity to use AI responsibly. Policy priorities in the next mathematics curriculum review in 2027 include a national framework and consistency of approaches, especially related to AI; professional learning for teachers; equitable technology access; and a refocus on understanding the mathematics. Future curriculum priorities include dynamic simulations using real-world data, emphasizing understanding over calculation, explicit inclusion of data science and critical data literacy, and an integration of systems thinking. Sakurai closed by offering three resources of interest: the Australian Framework for Generative AI in Schools,⁸ Australian Curriculum Assessment and Reporting Authority,⁹ and Computational Thinking in the Australian Mathematics Curriculum.¹⁰

DISCUSSION

Launching the discussion, Seshaiyer asked the presenters about guidance in their countries provided to school districts, schools, and teachers

___________________

⁸ See https://www.education.gov.au/schooling/resources/australian-framework-generative-artificial-intelligence-ai-schools

⁹ See https://www.australiancurriculum.edu.au/

¹⁰ See https://www.mathematicshub.edu.au/understanding-maths/the-curriculum/australian-curriculum-mathematics/

on how to integrate computational thinking, data science, and AI into the classroom. In Scotland, Farrell said, a crucial point is to “sell” these concepts with a playful approach. Teachers were feeling scared but they are assured that they can just teach a few lessons in the subject. They can engage students in activities such as escape rooms, board games, card games, and live lessons. Once they use an engaging activity and see the relevance, they may explore other activities. Playfulness as a selling point is working well in Scotland and beyond. She noted that people from 120 countries have used their materials.

In Singapore, the Ministry of Education offers practical implementation of AI in education, Cheong said, with the framework described above and with case examples, situations, and other guidance. The guidance and implementation are broad, and principals set the tone for their schools. Teachers have a list of tools they can use and the flexibility to try them.

Lee said that launching new courses in South Korea’s curriculum document is a powerful way to introduce new concepts. Textbooks are developed for these new courses. Professional development is, however, challenging. Teachers may not know the content or pedagogical knowledge about AI or statistics, and they have to learn new instructional approaches for the new courses. Another challenge is how to address generative AI. It is not reflected in the national curriculum, but students and teachers are interested in it, so schools are trying many approaches in the math classroom, such as error analysis probing.

In Australia, teachers need to know how to use the technology, not just when to use it, Sakurai said. She noted another challenge is balancing curriculum and policy. Some policies are pushing for students to use less technology in the classroom. In primary school, it is mandated to have not more than 90 minutes of technology a day. She also identified a push-pull in which there is a push toward procedural mathematics, but the curriculum still leans toward conceptual understanding. In addition, the textbook market is open to its own market forces. But, she said, the textbook model will not always work with rapid advances in computational thinking and AI. So, she posed, where will teachers get their resources and how will resource development be funded?

Seshaiyer asked about incentives for teachers taking on extra work to develop resources. In Scotland, Farrell replied, a school must be funded to release a teacher for professional development. In the last 6 months or so, demand for the training has grown because of the addition of AI. Teachers are looking for guidance. In Australia, teachers must have a master’s degree so they have 2 years to build new skills. Sakurai expressed hope that the emerging workforce will help more established teachers with new technology. In Korea, Lee said, math teachers receive a salary increase if

they complete 60 hours of professional development in a year. An “AI in education” master’s degree is offered, as well as many individual courses.

Hollylynne Lee asked Cheong how teachers and students use AI in secondary classrooms in Singapore. He responded that there is a student learning space within a portal provided by the Ministry. Teachers can infuse the content and questions for students to work on, with autogenerated feedback. Each student has an iPad, so there is equitable access to the portal.

Markku Hannula praised Scotland’s efforts to base curriculum on the United Nations Convention on the Rights of the Child. He noted that policies are often looked at to ensure they are lawful and ethical, but this is a focus based on rights. In practice, Farrell said, prioritizing a rights-based approach can mean, for example, that young people have the right not to use AI because of their concerns about sustainability, privacy, or for other reasons. The curriculum should not insist that young people use these tools and should model how they can navigate their preference with other organizations. The decision by local authorities about whether and how to use AI must be open and transparent, she added.

When asked by Conrad Wolfram about assessments in the countries, Farrell said Scotland has no formal assessments for skills up to age 16. Students demonstrate mastery and move up when they are ready. Formal assessment at age 16 is more challenging. For data science the decision was made not to go on an exam-based route because an examination would have taken years to develop. Through internal assessment, young people demonstrate they can go through a data science cycle, come to solutions, and communicate them. In Australia, Sakurai said, achievement standards are linked to modeling or problem solving. From kindergarten to grade 10, they could be based on real-life contexts. The final 2 years of school for those in an academic pathway are very prescriptive with final examinations, whereas the vocational curriculum is competency based.

Nick Fisher pointed out the learning value of getting things wrong, but also noted the peril of getting things wrong with AI. Farrell agreed and commented that a core aspect of AI literacy needs to be critical literacy: “Is this telling me the truth or not?” Lessons are being built for Scottish students from a young age, but it is getting harder to do. Cheong stressed that the teacher in Singapore is always in the loop to counter AI errors. Lee said informatics and mathematics courses in Korea include AI literacy. Sakurai said teaching critical data literacy is already on the list to consider during Australia’s next curriculum review in 2027.

When asked by Deborah Stine about the biggest challenges related to policy making and views of other stakeholders, Farrell responded that experiences with data science and AI policies have been different in Scotland. One advantage of working in a small country is the connections within the government, including with Education Scotland and Qualifications

Scotland. “We can all get in one room and talk about issues and come up with solutions and strategies, which has been fantastic,” she said. Education Scotland offers webinars for parents to help them understand AI, AI safety, and how AI can be used with learning. In Singapore, Cheong said, the largest concern is that generative AI can stunt the thinking process if not done well. Teacher mediation and lesson design really matter, he emphasized, which is reflected in the Ministry of Education’s AI framework. Lee said stakeholders in Korea agree with the intention to include AI in the curriculum. The two main challenges are that teachers do not want to add more elective courses and assessment. Entrance to Korea universities is very competitive, so assessment is important. In Australia, Sakurai said, the biggest barrier is budget to develop resources and conduct professional learning. States and territories have agreed to let the federal government take the lead on AI, but they can enact policies differently. Although there has been no survey, parents seem to support AI in the schools. They see it in their workplace and they want students to be prepared.

In terms of input from industry about what to teach in K-12, Farrell described several efforts in which industry groups have provided examples of how they use data science to share with teachers. She stressed, however, that schools should provide students with skills for life, not train them with specific job skills, which is what industry training should do.

Picking up on the point about more content continually being added to the curriculum, Sakurai said nothing has been dropped in Australia but the goal is to show the relevance of the new content. Lee said the other courses were not removed when the electives he described earlier were added in Korea. Cheong said the curriculum committee in Singapore has the philosophy to “learn better and deeper.” Not all newly added activities appear in high-stakes exams. Some topics are learned in context. For example, vectors are taught as a concept, but the word “vector” may not appear.

This page intentionally left blank.