There is a growing recognition in K-12 education of the importance of artificial intelligence (AI) literacy to help students navigate and shape an AI-driven world thoughtfully and responsibly (Wang et al., 2021). Yet, many schools and districts lack the necessary resources (e.g., trained teachers, professional development [PD], instructional materials) to provide students with meaningful opportunities to learn AI concepts, limitations, and ethical considerations (Code.org, 2024). Currently, AI instructional responsibilities largely fall on computer science (CS) teachers, who may lack specific training in AI but possess a solid understanding of CS concepts needed to support AI instruction (Ng et al., 2023). Thus, most existing research on teacher training and the implementation of AI education has focused on CS educators (Grover, 2024). However, to address resource gaps and expand AI learning opportunities, other educators (e.g., library media specialists, Career and Technical Education [CTE], and STEM instructors) are increasingly called upon to teach AI (Grover, 2024). More research is needed to determine the best resources to prepare non-CS educators to teach AI effectively, identify potential obstacles they may encounter, and conceptualize best teaching practices (Yao et al., 2023).
The AI PLAY project is a four-year, federally-funded research effort aimed at designing, developing, and investigating curricular materials that support middle-grade students' AI learning-aligned with the Five Big Ideas in AI (Touretzky et al., 2019)-in rural communities. This design-based research project (Sandoval, 2014) positions CS and AI researchers from a regional university in collaboration with teachers to create engaging activities that can be adapted and integrated for classroom use. An integral component of AI PLAY is the design and development of a PD program that supports teacher learning of AI and offers resources to facilitate student learning in the classroom. Teachers and researchers work collaboratively to build and refine resources and best pedagogical practices to support students' AI learning. Teachers attend a workshop to learn AI concepts and tools to teach students about AI and then implement those resources in their classrooms. In the first year of the project, we recruited a small sample of teacher participants to establish a greater understanding of the resources, support, skills, knowledge, and potential barriers to implementation to refine and develop a more robust PD program for subsequent years. The research team designed an initial one-day PD workshop that provided teachers with an introduction to AI and AI concepts as well as classroom resources to support their AI teaching. The majority of teachers in the program are not CS teachers. Thus, the team must build a sustainable PD program for this demographic. We studied one teacher's professional learning journey to get a better understanding of how the AI PLAY PD program could support these educators.
Cultural-Historical Activity Theory (CHAT) provides a comprehensive framework for studying how teachers integrate new skills and resources into their teaching practices (Ke et al., 2023;Rasmussen & Ludvigsen, 2009). Originating from Vygotsky's ideas, CHAT views the individual actions of a subject, as part of broader, goaloriented activities (object) that are mediated by tools, language, and other cultural artifacts within their social, cultural, and material contexts (Vygotsky, 1978). CHAT emphasizes that these activities do not occur in isolation; they are shaped by a community of people working together, a division of labor that allocates roles and tasks, and rules that guide acceptable actions and implicit norms within the group (Engestrom, 1987(Engestrom, /2015)). Ultimately, human activity is analyzed as part of a system where contradictions or conflicts between elements-such as competing goals or insufficient resources-often become triggers for change and growth. Investigating these contradictions can help researchers understand where educators need additional support or where new opportunities exist.
A key concept within CHAT is expansive learning theory where people create new approaches to resolve contradictions and tackle problems that lack concrete solutions (Engestrom & Sannino, 2010). Expansive learning is a collaborative process occurring within and across activity systems (AS), whereby undefined concepts are transformed into innovative practices as existing structures and methods are reshaped. Thus, it is useful for analyzing situations where an ideal outcome does not exist yet-such as non-CS educators' best teaching practices with AI-as it provides a holistic way to understand how learning, development, and change unfold in practice within complex environments like classrooms (Pareto & Willermark, 2022).
We applied CHAT and expansive learning theory as an analytical lens to understand how a non-CS teacher, a library media specialist, translates new skills and resources from participating in the AI PLAY PD program into classroom practices for teaching AI. This context constituted the AS under study, emphasizing the interconnectedness of her actions within the broader socio-cultural context of her school and classroom environment. The theoretical lens also highlights collective learning, as she collaborated with researchers and colleagues and engaged students, potential contradictions that arise in the implementation process, and how transformations can lead to profound changes such as the development of new knowledge and practices. The following research questions guided our analysis: RQ1. How does participation in the AI PLAY PD program influence a non-CS educator's implementation of AI teaching strategies in the classroom? 1a) What sociocultural and contextual factors influenced implementation? 1b) What contradictions and transformations emerged?
We chose an exploratory, single-case study design (Yin, 2014), to conduct an in-depth analysis of one educator's experiences in the AI PLAY PD program. Ann (a pseudonym) was a library media specialist at a middle school in the southeastern region of the United States. Ann participated in a one-day PD program led by our research team. With their support, she implemented AI learning activities with students immediately following the PD and independently with a new group of students the next semester. Thus, we sought a longitudinal perspective of Ann's adoption and implementation of new skills and resources, providing valuable insights into the changing conditions and underlying processes that impacted her outcomes. Because Ann did not teach classes on a fixed schedule, she collaborated with a colleague to implement the AI activities with her students. In the first semester, the classes consisted of academically and intellectually gifted (AIG) students, while during the second semester classes were comprised of regular education students.
The data consisted of Ann's responses to open-ended survey items before and after the PD, transcribed audio recordings of 20-30-minute semi-structured interviews with a member of the research team after each classroom implementation, and field notes from the first classroom implementation. One researcher performed an initial coding cycle of a line-by-line analysis of the data, with labels assigned according to the CHAT framework (e.g., Subject, Tools, Object, Rules, Community, Division of Labor) to characterize the AS. Throughout this initial phase, the researcher also recorded analytical memos to capture emergent themes as they surfaced. Subsequently, labeled data were organized into a CHAT categorization table, enabling the identification of patterns and interrelationships between elements. This structured data informed the generation of further analytical memos and refinement of initial observations.
In the second cycle, focused coding was applied to the categorized CHAT data using expansive learning theory, with labels such as "tension/contradiction," "mediating/enabling," and "expansion/transformation" to discern patterns and processes. Emerging themes were further noted and elaborated on. A second researcher independently reviewed the coded data and themes to ensure reliability through both phases. Discrepancies were discussed and resolved through consensus-building.
To answer Question 1a, we present the AS for Ann in relation to her participation in the AI PLAY project developed from our initial analysis, which is illustrated in Figure 1.
Regarding Question 1b, we report only on the most notable contradiction within Ann's AS due to page constraints. The most significant tension prior to the PD lay between Ann as the subject and the object: unfamiliar teaching practices with a new discipline, AI. Ann stated she lacked the confidence and expertise to teach AI content effectively, "I would consider this unknown territory since it is my first ever training on teaching AI to my students." This tension was amplified by the established norm (rule) within the United States educational system that teachers are subject matter experts responsible for didactically transmitting knowledge to students (Popper-Giveon & Shayshon, 2017). This expectation made it difficult for Ann to initially view herself as anything other than a novice, which caused hesitation around teaching the subject. "I didn't want to look like an idiot. I mean, those 8th graders will call you out." Accordingly, Ann selected resources within the AS that she felt the most comfortable with and relied heavily on the support of the researchers including using learning materials in their original forms. At the same time, we observed that these choices gradually built her confidence to teach AI, as she refined tools and practices, and ultimately adopted a collaborative mindset that positioned her as a learner alongside her students.
Initially, Ann chose to work with a class of AIG students during the first semester. She perceived these students as high achievers and early learners capable of quickly grasping new concepts, which gave her the initial confidence to attempt implementing AI content. Additionally, her prior relationship with these students fostered a supportive environment where Ann felt more at ease trying an unfamiliar activity. "I mean I've taught these kids a lot. This isn't my first time teaching them, so I have a relationship with them." Once she observed the students' favorable responses-characterized by high engagement, motivation, and successful attainment of learning objectives-Ann felt encouraged to continue teaching AI to other student groups in the school the next semester. She shared this experience, "It was really valuable to see how engaged they were and how they were making these connections. Kids know a lot more about AI than I thought and it showed me we need to meet them where they are."
The PD and ongoing support from the researchers further enabled Ann to develop her competence. Having a supportive team available during the PD, while she was preparing to teach the lesson, and during the initial classroom implementation provided Ann with reassurance and additional resources that reinforced her sense of capability in delivering the AI lesson. She commented, "Right before I taught a lesson if I had questions, [the researchers] were very quick to respond and would even go into the slideshow and provide notes and then I thought of things I could say to the students." The scaffolding from the research team during the PD (e.g., modeling AI activities, teachers as hands-on, active learners) and the positive classroom experiences with students encountered in the first implementation helped Ann internalize new teaching practices.
Additionally, Ann's strategic choice of instructional tools and content was essential during her early AI teaching experience. She selected an unplugged activity focused on Representation & Reasoning (Lim et al., 2024)-something she felt comfortable with and believed her students would find accessible. "I went with face recognition first because I was more interested in this and one [the students] can connect. I think the self-driving cars and the sensors, that is one that I might not be as confident." This choice was significant in reducing her apprehension, allowing her to start with a manageable topic, and building her confidence through encouraging teaching experiences with materials she was comfortable with. She modified the same activity the next semester to better meet student needs and added an advanced computer-based pathfinding activity.
Ultimately, through these interactions within her AS, Ann was able to shift her understanding of teacher expertise, moving from a traditional expectation of holding all knowledge to a more flexible perspective where learning alongside students became acceptable. She reflected on this shift in the final interview, noting, "I mean, I'm learning too. I tell them that. I went to a training, I have this new information I've learned and I want to share it with you. Just being upfront, like I don't know everything." In the end, expanding her outlook and approach empowered Ann to create new learning opportunities for her students.
Despite the limited duration of this study, we observed meaningful growth in Ann's approach to teaching AI, facilitated by her participation in the AI PLAY PD program. These developments offer promising insights for our ongoing research and for others who wish to foster effective AI teaching practices, particularly for non-CS teachers. Ann's experiences highlighted common barriers to introducing a new subject like AI, as well as resources and strategies that supported her in overcoming these challenges. Her journey emphasizes the importance of viewing teacher learning of AI as a gradual progression, enriched by a network of supportive elements within the AS that nurture confidence and skill development (Warford, 2011).
The main barrier was her initial lack of confidence and knowledge in teaching AI which was compounded by the institutional norm regarding teacher expertise (Beijaard, 2000;Bobis et al., 2020). However, several supportive elements within Ann's AS helped her overcome this obstacle. First, the collaborative and supportive nature of the PD played a critical role in building her confidence and competence. Ann's limited experience with AI was initially a source of apprehension, but continuous guidance and resource-sharing from the research team alleviated her concerns. This finding suggests that a collaborative PD approach, especially one with ongoing support, is essential for engaging teachers with complex topics (Desimone, 2009). Thus, researchers should determine what supports educators want and need prior to designing a PD program for AI education so that they are readily available.
Ann was also able to leverage several elements within her AS that comfortably placed Ann within her zone of proximal development (Vygotsky, 1978), enabling her to gradually take on new challenges with growing confidence and expertise. For example, Ann's decision to begin implementing AI with a familiar group of AIG students provided a safe environment for experimenting with new materials. The trust and rapport she had with these students helped reduce her anxiety, underscoring the value of initial, low-stakes implementations for teachers new to AI (Reich, 2022). Another important strategy was the selection of accessible, introductory AI topics, such as Representation & Reasoning with an unplugged activity, which felt manageable and engaging for both Ann and her students. By starting with familiar concepts rather than more advanced topics, Ann was able to experience early successes that motivated future actions. Therefore, PD designers should consider introducing AI topics in an approachable way to allow teachers to build foundational confidence and gradually expand to more complex concepts (Long & Magerko, 2020).
Finally, the most significant finding was Ann's shift from a traditional "expert" role to a "co-learner" role. This transformation, an instance of expansive learning as described by CHAT, allowed her to adopt a more adaptive teaching identity, where learning alongside her students became acceptable. This shift enabled Ann to overcome the tension between traditional norms of teacher expertise and the new demands of AI teaching. Such flexibility is essential for engaging with rapidly evolving subjects (Johnston, 2019), and PD programs should support these identity shifts by including discussions around teacher roles and expertise when teaching unfamiliar, complex content. This finding also aligns with prior CS education work that advocates teachers as co-learners and lead learners (Tucker-Raymond et al., 2020). Embracing a collaborative learner mindset will be critical as educators adapt to continuous advancements in AI, engaging in co-learning and exploring new tools.
This paper represents an exploratory investigation that traced Ann's AI PLAY experience during one school year. Our team will continue to investigate Ann's experiences with AI teaching, as well as conduct indepth analyses with other teacher participants in AI PLAY over the course of the project. With additional data, we hope to conduct cross-case analyses that will enable us to generalize findings to design a larger more scalable program. We will integrate outcomes from this investigation to refine the program with new participants.