Human-Centric Artificial Intelligence Pedagogy (HCAP) framework developed from TPACK through integration of artificial intelligence literacy and competency

Synthesis: Chiu (2026) proposes the Human-Centric AI Pedagogies and Teaching Strategies (HCAP) framework, an evolution of the Technological Pedagogical Content Knowledge (TPACK) model designed for the generative AI era. Arguing that AI's agentic autonomy, epistemic complexities, and ethical dimensions render the established TPACK framework insufficient, HCAP integrates five knowledge domains: AI-Technological, AI-Content, AI-Pedagogical, Human-AI Collaborative, and Ethical Knowledge. A three-round Delphi study with 30 teachers across diverse subjects produced a consensus on 25 critical knowledge items, providing an empirically grounded model that translates theoretical AI pedagogy into actionable teacher competencies and equips educators to move from merely using AI to strategically orchestrating human-AI collaborative learning.

Key Findings

• HCAP extends TPACK with five interdependent knowledge domains: three that transform the original TPACK cores through a sociotechnical lens — AI-Technological Knowledge (I-TK), AI-Content Knowledge (I-CK), and AI-Pedagogical Knowledge (I-PK) — plus two entirely novel domains: Human-AI Collaboration Knowledge (HAIC-K) and Ethical Knowledge (Ethics-K).
• The framework addresses four critical gaps in TPACK: tool agency (understanding AI's operational logic, probabilistic nature, and prompt engineering), epistemic challenges (teachers shifting from content deliverers to critical validators), ethical complexity (making ethical reasoning a core, explicit domain), and dynamic collaboration (conceptualizing human-AI collaboration where cognitive tasks are shared).
• A three-round Delphi study with 30 teachers (adopting a 75% agreement threshold) refined and reached consensus on 25 critical knowledge items across the five domains, including: prompt engineering, limitation awareness, system auditing, data fundamentals, tool integration (I-TK); critical validation/source triangulation, bias detection, contextualization, disciplinary methodological critique (I-CK); reflective practice, assessment redesign, scaffolded co-agency design, personalization, ethical integration (I-PK); dynamic role allocation, interaction flow design, critical dialogue, critical interdependence, group dynamics collaboration (HAIC-K); and equity auditing, data privacy, usage protocols, societal impact, Well-Being advocacy, inclusive design (Ethics-K).
• The knowledge domains are anchored in foundational literacies: I-TK in data and computational literacy, I-CK/I-PK in critical/algorithmic/media and scientific literacies, HAIC-K in collaborative and epistemic literacies, and Ethics-K in ethical and well-being literacies.

Study Design & Method

This study used a three-round Delphi method with a panel of 30 teachers from diverse subjects to establish expert consensus on the knowledge and skills required within each HCAP domain. An initial list of 20 knowledge items was refined iteratively: in Round One, seven items reached consensus and two new items were proposed; in Round Two, four more reached consensus, one new item (well-being advocacy) was proposed, and some items were split or renamed; in Round Three, nine more reached consensus and one new item (ethical integration) was proposed. The final list comprised 25 knowledge items. The study adopted a 75% agreement threshold (in line with prior Delphi research), and items endorsed for removal by more than half of the teachers were excluded. Teacher modifications to item labels and descriptions were incorporated across rounds.

What this means for practice

• Faculty developers. Assess HCAP proficiency with a competency-based portfolio of human-AI teaching artifacts — a video of facilitated critical dialogue, annotated student work, a reflective journal on system auditing, and a curated set of prompts with their validated outputs — rather than written assignments, and position teachers as strategic orchestrators of human-AI collaboration rather than tool users.
• Faculty developers. Build cross-disciplinary professional learning communities in which pre- and in-service teachers, technology specialists, and subject-matter experts collaborate, so a math teacher working on disciplinary methodological critique can partner with a computer science colleague who has strong algorithmic literacy.
• Instructors. Redesign assessment toward process portfolios that include students' prompts, their critique of AI outputs, and a reflection, instead of summative products that assume individual authorship.
• Instructors. Sequence AI integration deliberately: audit tools and engineer prompts first (I-TK), then triangulate and critique outputs against reputable sources (I-CK), then design scaffolded co-agency and dynamic role allocation (I-PK and HAIC-K), with ethical protocols such as disclosure requirements running throughout.
• Administrators. Treat the five domains as interdependent and fund development accordingly, because deficiency in any one domain — technical skill without ethical competence, for instance — undermines the integrity of the whole framework.

Limitations

• Consensus came from a panel of 30 teachers, and the 75% agreement threshold and consensus criteria are inherently subjective: as the author notes, there are no universal standards for Delphi consensus.
• The knowledge items reflect a teacher-perspective view and were validated through expert consensus rather than classroom outcome data, so the framework’s implementation effectiveness still requires empirical validation in practice.
• The Delphi refinement process involved renaming and restructuring items, so the final 25-item list reflects the specific panel’s judgments and may not generalize to all educational contexts.

Citation

Chiu, T. K. F. (2026). Human-Centric Artificial Intelligence Pedagogy (HCAP) framework developed from TPACK through integration of artificial intelligence literacy and competency. Interactive Learning Environments.