Competence frameworks function as structured knowledge codification systems: authoritative, domain-wide representations of professional knowledge that define what counts as competence within a field and how it should be organised. As such, they constitute a form of professional knowledge encyclopaedia, structured to comprehensively map an emerging or established knowledge domain. This paper examines how eight EU and international competence frameworks codify AI-related knowledge for STEAM educators, using a dual analytical lens comprising AI-TPACK and STEAMCompEdu. Read through this dual lens, the cross-framework analysis identifies four knowledge dimensions insufficiently addressed in the existing landscape: interdisciplinary pedagogical knowledge, creative-aesthetic AI knowledge, ethics as embedded disciplinary practice, and inquiry-oriented AI pedagogy. Based on these findings, the paper proposes three emerging theoretical extensions to AI-TPACK to address the identified lacunae and discusses implications for the design of structured professional knowledge codification systems that better represent complex, interdisciplinary, and rapidly evolving knowledge domains.
Artificial intelligence (AI) is rapidly entering classrooms, and its arrival places new demands on teachers. Integrating AI tools into teaching is not simply a matter of technical familiarity; it requires teachers to understand how AI systems work, what they can and cannot do, how they reshape learning tasks, and what ethical questions they raise. Two distinct bodies of work have developed in response to this challenge. The first is theoretical, seeking to describe the knowledge an effective teacher of AI-integrated lessons must possess. The second is operational, seeking to specify and codify that knowledge in structured, assessable terms that can guide teacher training and professional development. This paper brings these two bodies of work together and examines them in the specific context of STEAM education.
On the theoretical side, the dominant account is the Technological Pedagogical Content Knowledge (TPACK) framework
[1] and its recent artificial intelligence extension, AI-TPACK
[2]. TPACK models effective technology-integrated teaching as the interplay of three knowledge domains, technology, pedagogy, and subject matter content, while AI-TPACK adds a fourth domain, AI knowledge, to capture the distinctive understanding that AI tools demand. AI-TPACK thus provides a principled account of what a teacher needs to know to integrate AI effectively. On the operational side, a growing number of competence frameworks, structured specifications that define what counts as competence within a field and organise it into domains, areas, and descriptors, seek to translate this knowledge into concrete, trainable, and assessable terms. Produced by the European Commission, UNESCO, professional standards bodies, and academic research projects, frameworks such as DigCompEdu
[3], DigComp 3.0
[4], the UNESCO AI Competency Framework for Teachers
[5], GreenComp
[6], LifeComp
[7], EntreComp
[8], the European e-Competence Framework
[9], and the UNESCO ICT Competency Framework
[10] collectively constitute the most comprehensive and widely adopted landscape of educator and learner competence specifications currently available.
Competence frameworks of this kind function as a distinctive form of professional knowledge codification: structured, authoritative representations of what counts as competence within a field, how that knowledge is organised, and what progression within it looks like. In this sense, they operate as professional knowledge encyclopaedias, not merely lists of skills, but systematic attempts to map a knowledge domain comprehensively, to define its internal architecture, and to establish a shared vocabulary for its development and assessment. The proliferation of competence frameworks for educator AI integration is therefore not merely a policy phenomenon but an epistemological one: a collective effort to codify a rapidly evolving knowledge domain before that domain is fully understood. The quality of that codification, its comprehensiveness, its internal coherence, and its adequacy to the knowledge demands of the professionals it describes, is consequently a matter of significant practical and theoretical importance.
Yet, despite the sophistication of both the theoretical and operational dimensions of this field, a significant and consequential gap persists. Neither AI-TPACK nor any of the eight major competence frameworks was designed with the STEAM educator in mind, the professional who facilitates learning that deliberately integrates Science, Technology, Engineering, Arts, and Mathematics through interdisciplinary, inquiry-based, and creativity-oriented approaches. STEAM pedagogy has distinctive characteristics, identified and theorised across the body of literature including Yakman
[11], Honey et al.
[12], Kelley and Knowles
[13], Ring et al.
[14], and the arts integration scholarship of Hetland et al.
[15] and Perignat and Katz-Buonincontro
[16], that make AI integration qualitatively more complex and knowledge-demanding than single-discipline, direct instruction teaching.
STEAM education refers to an integrative approach that connects Science, Technology, Engineering, Arts, and Mathematics, rather than teaching them as separate subjects. Its defining commitment is interdisciplinarity: learners investigate authentic, open-ended problems that draw on several disciplines at once, typically through inquiry-based and project-based work in which creative and aesthetic judgement matters alongside technical reasoning. The STEAM educator is therefore not a single-subject specialist but a professional who designs and facilitates learning across disciplinary boundaries, and it is this professional whose knowledge demands, when AI enters the classroom, neither AI-TPACK nor the eight competence frameworks were built to describe.
To address this, the present study analyses the eight frameworks through a dual analytical lens that pairs two complementary perspectives. The first is AI-TPACK
[2], introduced above, which specifies what a teacher needs to know to integrate AI effectively. The second lens is STEAMCompEdu
[17][18][19] (the STEAM Competence Framework for Educators), the competence framework designed specifically for STEAM educators, which articulates 41 competences across 14 areas structured around five educator perspectives. Together, these two frameworks define what an AI-competent STEAM educator needs to know and do and thereby reveal what the existing landscape does not yet provide. The paper makes two connected contributions. The first is analytical: a cross-framework analysis using the dual AI-TPACK × STEAMCompEdu lens to identify four knowledge dimensions that are insufficiently addressed across the eight frameworks. The second is theoretical: three emerging extensions to AI-TPACK motivated by these dimensions, grounded in both competence framework evidence and the STEAM pedagogical literature, and informed by practitioner-level experience of framework development in European educational contexts.
These two contributions are closely related but serve different purposes. The cross-framework analysis identifies four knowledge dimensions that are insufficiently addressed across the existing framework landscape, whereas the theoretical response developed in this paper proposes three emerging extensions to AI-TPACK. This apparent asymmetry is intentional. Three of the identified dimensions, interdisciplinary content and pedagogical knowledge, creative-aesthetic AI knowledge, and ethical AI knowledge, represent conceptually distinct knowledge domains that warrant theoretical elaboration. In contrast, the fourth dimension, inquiry-specific AI pedagogy, is interpreted as a pedagogical refinement of AI-TPACK’s existing pedagogical components rather than as a separate knowledge domain. Accordingly, the proposed extensions comprise: interdisciplinary content and pedagogical knowledge (ICPK), which supports the integration of AI across STEAM disciplines through coherent interdisciplinary pedagogy; creative-aesthetic AI knowledge (CAAK), which enables educators to facilitate AI-supported creativity, aesthetic judgement, and meaningful human–AI co-creation; and ethical AI knowledge (EAK), conceptualised as a standalone knowledge domain that embeds ethical reasoning across all dimensions of AI-supported teaching and learning.
The paper proceeds as follows:
Section 2 establishes the theoretical genealogy from TPACK to AI-TPACK.
Section 3 introduces the dual analytical lens (AI-TPACK and STEAMCompEdu), the eight-framework corpus, and the relevant STEAM pedagogical models.
Section 4 conducts the cross-framework analysis.
Section 5 analyses why the identified dimensions are particularly consequential in STEAM contexts.
Section 6 proposes three AI-TPACK extensions.
Section 7 discusses implications for framework design and professional development.
Section 8 discusses limitations and future research, and
Section 9 concludes.