What’s Happening in Practice?

Last year, artificial intelligence became a focal point in K-12 education. Now more than ever, there is significant interest in understanding not only the landscape of K-12 AI education, but also progress in related subjects like computer science or science, technology, engineering, and mathematics more broadly. Previously, the National Security Commission on AI’s (NSCAI) Final Report stated that investing in AI and STEM education was critical to future U.S. technological competitiveness and national security. However, the education space is quickly becoming saturated with guidance, curricula, materials, and opinions on what AI education is and how to teach about and with AI. To help make sense of the AI education landscape, this blog post describes new and existing K-12 AI education efforts so that U.S. policymakers and other decision-makers may better understand what is happening in practice. These insights and observations are drawn from a deep exploration of K-12 AI education over the last year, including discussions with stakeholders, conference attendances, and literature reviews.

Because the U.S. education system is largely decentralized, progress in K-12 AI education is being driven by local school districts and state departments of education, with input from non-profits and industry partners. In practice, this means that the design, approach, and implementation of K-12 AI education vary across states and school districts. Many factors influence the decisions made at the local level, such as the novelty of AI as a subject in K-12 curricula, varying definitions of what constitutes AI education, access to funding and human capital, and general variance in computer science and technology education progress across the country. For those concerned about the AI digital divide, this creates a sense of unbalance in the education system, where schools that can implement AI curricula are ahead and schools that are unable to are behind.

Our findings include:

Finding 1: High schools are adopting and implementing AI career technical education (CTE) programs. Over the last year, we identified at least 19 high schools across Maryland, Georgia, California, and Florida that have either already implemented or are preparing to implement an AI-specific CTE program. All of the 19 high schools are public schools, and at least eight are magnet schools. The school districts in Maryland, Georgia, and California each developed AI-specific CTE programs and implemented them within individual schools. The Florida Department of Education, in partnership with the University of Florida, developed an AI-specific CTE program at the state level, which 13 Florida high schools either have or are planning to adopt. 

Once known as vocational training, CTE programs are optional and supplemental methods of instruction that are designed to prepare students with the technical knowledge, skills, and abilities related to specific occupations and career clusters. CTE programs span nearly every major industry, from healthcare and cosmetology to manufacturing and computer science, and are increasingly seen as an opportunity to strengthen the connection and coherence among K-12 education, postsecondary education, and workforce development efforts. Programs can start as early as middle school but are more common at the high school and technical college levels, thus allowing students to explore future career fields, earn college credits, or obtain industry-recognized certifications. Well-designed pathways may also enable students to gain an early start in their field and shorten the time required to enter the workforce. Almost all public school districts across the United States offer a variety of CTE programs, with approximately 7.5 million secondary students presently enrolled across disciplines.

The new Florida AI CTE pathway is an example of one such program and focuses on technical skill proficiency and competency-based applied learning of AI. There are four classes in this CTE pathway: (1) AI in the World, where students explore the role of data and ethics in AI applications and how AI agents make decisions; (2) Applications of AI, designed to deepen understanding of AI applications and how to build AI models; (3) Procedural Programming, which continues the study of computer programming concepts with a focus on the creation of software applications; and (4) Foundations of Machine Learning, designed to provide students with core foundational knowledge to deepen understanding of machine learning (ML) practices and applications.

Another example is an AI CTE program out of Georgia. In 2022, the district opened an AI-themed high school and created a three-course AI CTE pathway that incorporates AI “literacy” and “core” AI skills into the curriculum. The first course in the CTE pathway, Foundations of AI, introduces students to programming, data science, math, and ethical reasoning skills. The second, AI Concepts, teaches students about the history of AI, current AI research, and the societal and ethical impacts of AI. The third course, AI Applications, teaches students to design and test AI-powered solutions. While this pathway is for students who want to dive deeper into AI-focused studies, an elementary and middle school feeder program has started piloting AI-ready learning embedded courses to develop the necessary skills for deeper student engagement at the high school level. 

Finding 2: Nonprofit organizations, academic institutions, and industry fill critical gaps in the educational ecosystem by designing and delivering content and curricula, providing teacher training programs, and emphasizing foundational skill-building. These entities also assist schools, districts, and states with developing curricula and educational frameworks in the absence of state and federal policy. In May 2018, the Association for the Advancement of Artificial Intelligence and the Computer Science Teachers Association launched the AI for K-12 working group (AI4K12) to develop some of the first AI teaching guidelines for K–12 schools. AI4K12 introduced “Five Big Ideas” in AI as its core framework: (1) Perception, (2) Representation and Reasoning, (3) Learning, (4) Natural Interactions, and (5) Societal Impact. The guidelines outline key learning objectives for students across different grade bands and can assist standards writers and curricula developers with incorporating essential knowledge and skills related to AI concepts. The framework is also recognized by the United Nations Educational, Scientific, and Cultural Organization (UNESCO) as one of the few existing AI education frameworks. 

There are nonprofits leading efforts to develop readily-adoptable AI education curricula and tools for K-12 classroom educators. One example is the MIT Media Lab, a research laboratory at the Massachusetts Institute of Technology, which developed the DAILy curriculum for middle school students to explore AI concepts and applications using hands-on and computer-based activities. The curriculum includes four units: (1) What is AI? (2) Supervised Machine Learning, (3) Generative Adversarial Networks, and (4) AI + My Future. Each unit includes slides for lessons, teacher scripts, and interactive classroom activities. The curriculum was designed to to scale programming and training for teachers which can be barriers to adoption.

The AI Education Project, another nonprofit organization, promotes general AI literacy by providing AI education to all students through interdisciplinary approaches. The organization developed two classroom-ready courses. The first course, Introduction to AI, uses project-based learning to teach students in 9-12 grade how to build, test, and refine an AI application without coding. The course is designed for computer science, STEM, or CTE teachers. The second course, AI Snapshots, is designed to facilitate a basic understanding of AI and its connections with math, science, English, and social studies beginning in middle school, and is designed for all teachers regardless of discipline.

In addition, nonprofit organizations are emerging to provide schools and policymakers with policy and implementation guidance on AI education. One example is TeachAI, an effort led by Code.org, the Educational Testing Service, the International Society for Technology in Education, and Khan Academy, that highlights both the potential benefits (e.g., assessments, personalized learning, operational efficiency) and harms (e.g., plagiarism, overreliance, perpetuating bias) of AI in the classroom. 

Certain federal STEM education initiatives support the critical work of nonprofit and educational leaders in AI K-12 education. One such program is the National Science Foundation’s Innovative Technology Experiences for Students and Teachers (ITEST) which supports research on increasing student interest in careers in STEM from preschool to high school. Recent awards totaling over $13 million support multiple projects aimed specifically at AI education initiatives such as:

• Introducing AI concepts to middle school students and teachers through summer camps, workshops, and school-based programs in rural communities where students often lack access to advanced STEM educational opportunities.
• Developing AI literacy through a research-informed educational ecosystem for after-school and summer programs where functional AI-enabled solutions to problems are presented in fictional stories that the students read in English-language arts and summer reading programs.
• Teaching fundamental machine learning concepts to middle school students by using ML models to classify shark teeth by their shape and function and address educational disparities in STEM to encourage students from underrepresented groups to consider STEM career pathways.

Finding 3: Conceptions of AI literacy can be broad. Defining AI literacy involves finding the right balance between theoretical knowledge and practical skills, as well as determining the necessary depth in each area. A key question is whether the understanding of AI concepts is adequate for achieving literacy, or if students must also possess technical abilities, like coding, to develop and work with AI models.

Our review of existing AI education literature reveals a trend toward definitions that do not mandate hard technical skills. Long and Magerko aggregated emerging themes and trends in the field of AI education by constructing 17 competencies that focus on building knowledge of AI concepts. They define  “AI literacy” as “a set of competencies that enables individuals to critically evaluate AI technologies, communicate and collaborate effectively with AI; and use AI as a tool online, at home, and in the workplace.” This framework of AI literacy complements AI4K12’s pioneering work on the “Five Big Ideas,” both emphasizing reflective and evaluative skills over hard technical skills. A proposal for an elementary school AI curriculum incorporates both frameworks, demonstrating their potential for AI education in practice. 

However, efforts to build an AI workforce often concentrate on strategies to build and maintain a workforce that is prepared to design, build, and test AI systems, software, or products. The necessary education and training needed will differ from workers with more generalist backgrounds or the average consumer. Therefore, we think the right approach might depend on the educational goal, which requires a necessary distinction between “AI in education” and “AI education.”

AI as an educational or classroom tool considers the use of AI-enabled tools to advance and improve educational practices for effective teaching and learning. This can take many forms, such as classroom management, individualized performance tracking, or other comprehensive classroom analytics that can serve as metrics of literacy in the effective use of AI tools in the classroom. Using AI-enabled tools can foster AI literacy to enable individuals to not only evaluate AI technologies but also to use AI as a tool.

AI education is concerned with actual educational policy and competencies for understanding technical AI concepts, foundations of AI technologies, how AI uses and perceives data, and how to use or build an AI tool. This also includes how concepts of AI are taught in the classroom through examining different agendas, pedagogical approaches, and definitions of what specific lessons and skill-building constitutes “foundational AI.”

Without technical requirements, the core concepts of AI literacy can be captured by, or integrated with, established instructional content in digital literacy and computer science standards across states. Common themes include computer literacy, digital citizenship, and computational thinking. Familiarity with the power and limitations of computation and communication technology is an important facet of STEM literacy, but computational literacy and computational thinking are not just limited to the STEM disciplines and are considered universally beneficial across all disciplines in K-12. States and schools already incorporate digital literacy and computational thinking into state standards and curricula. The motivation for including this type of instruction is that it will enable students to develop skills to deconstruct problems, recognize patterns, and think critically about solutions. 

As more states adopt computer science standards in their K-12 curricula, there might be a concurrent rise in either intentionally or unintentionally adopting components of AI literacy for students. However, states may vary in their approaches. For instance, California positions computer science as a standalone standard, on par with subjects like math and science, while other states fold it into a wider curriculum. Decisions on whether to introduce AI literacy as a new, independent standard or to integrate it within existing standards will influence resource allocation and the focus areas for teacher training.

Concluding Considerations

Schools are adopting diverse educational tools and methods, turning to options like CTE programs to address the growing demand for AI education. Non-profit organizations, academic institutions, and industry players are supporting these efforts by developing AI curricula and educational tools to help teachers and schools become better equipped. While the necessity of AI education is widely recognized, the key competencies needed to achieve AI literacy have yet to be established. What is evident, though, is a strong commitment from states and educators to reshape K-12 education.

Commonalities in schools offering standalone AI programming. Schools offering AI programming have existing computer science, data science, cloud computing, or cybersecurity CTE programs. This suggests that schools or states with existing computer science and technology education infrastructure may be more capable of adopting AI education standards or programming. It also suggests that these schools already have faculty that are prepared and able to teach separate AI classes. Therefore, the marginal cost of adding a separate curriculum is likely significantly lower for these schools. 

While there is no singular approach to AI education, many share common ground. Efforts to integrate or adopt computer science curricula are one way schools can introduce new technical topics or integrate them into existing ones, develop standards or assessments, and understand the required support for classrooms and educators. Nearly 60% of U.S. high schools offer at least one computer science class and 30 states require that their schools offer computer science. Disparities in access persist in rural regions, urban areas, and low-income communities. This is not to suggest that progress in equitable computer science education across the United States has been slow or stagnant. Instead, it is meant to highlight that a percentage of U.S. schools might not be prepared to begin thinking about AI education.

Schools that do not have structured AI coursework may not be falling too far behind. In its final report, the NSCAI urges the nation to prioritize investments in STEM education based on its assessment that the U.S. education system is not producing sufficient AI talent to meet U.S. demand. This is a national challenge, but education systems can fill this gap even without separate AI classes or educators or specific CTE pathways. For example, it is likely that AI concepts are taught either intentionally or unintentionally within standard computer science, mathematics, and other course work. This is why educational goals are important when considering what constitutes AI literacy—general STEM foundations, such as basic statistics or computer science, can still prepare students for future exploration of AI as a discipline or career. 

Teacher shortages are an often-overlooked challenge. Computer science, mathematics, and science teachers are most likely the educators who are more prepared to teach elements of AI education with minimal upskilling required. In some cases, one teacher may act as all three. For the 2023-2024 school year alone, 32 states report some form of a teacher shortage in the area of mathematics and science for at least one school district within the state. Seven states report teacher shortages specifically in computer science. A survey of 1,200 school and district officials by Frontline Education suggests that the top reasons for teacher shortages are: (1) a lack of fully qualified applicants; (2) salary and/or benefits are lacking compared to other careers; and (3) fewer new education school graduates. Even the NSCAI report and others acknowledge that recruiting high-quality K-12 teachers with STEM experience and proficiency is difficult.

Recognizing the need for increased resource allocation in teacher training, states have made positive strides. 2023 saw the largest increase in high schools offering computer science courses since 2018. Thirty-four states have adopted or updated policies to establish computer science as foundational, supported by an allocation of over $120 million in 25 state budgets for computer science education. These funds include increased professional learning opportunities to prepare teachers with the necessary skills to teach computer science effectively. For example, in Colorado, the state-funded Computer Science Teacher Education Grant (CSEd) Program is designed to train teachers in computer science education. Professional learning, upskilling, and certification opportunities are crucial not just for teaching computer science, but also for enabling more teachers with the skills and toolkits to integrate AI education into K-12 classrooms.

Ali Crawford is a research analyst at CSET, where she works on the CyberAI Project. She also serves in an uncompensated capacity on the advisory board of TeachAI, where she represents CSET.

Cherry Wu is a research assistant at CSET, where she works on the CyberAI Project. She is also a current master’s student at Georgetown University’s Walsh School of Foreign Service.