Unpacking ethics-domain of intelligent-TPACK scale in relation to in-service teachers' trust and distrust

Synthesis: Ocak and Caskurlu (2026) use an exploratory qualitative multiple case study to explore how in-service K-12 teachers' technical knowledge of AI relates to their Trust in using AI in education, and how this trust is shaped by ethical perceptions. Framed within the Ethics dimension of the intelligent-Technological Pedagogical Content Knowledge (TPACK) framework (Celik, 2023), the study examines four ethical constructs — transparency, fairness, accountability, and inclusiveness — as indicators of ethical AI, and analyzes how teachers' trust plays out relative to these constructs. Drawing on written reflections from seven purposively selected in-service teachers (categorized into lower, moderate, and higher-trust groups) following a two-week, AI-focused online asynchronous learning module, the findings suggest that without solid foundational technical knowledge, teachers struggle to recognize how human decisions shape AI systems — a prerequisite for trusting and evaluating them. The ethical constructs proved deeply interconnected and dynamic (fairness and accountability often emerging together), yet transparency and human accountability in decision-making received far less attention than fairness, inclusiveness, and accountability.

Key Findings

Design and sample. A qualitative multiple case study (following Yin) of seven in-service K-12 teachers, purposively selected from 33 consenting participants enrolled in a graduate-level dual computer-science/instructional-technology course at a southeastern US university (summer 2025). Cases were selected for divergent self-rated AI knowledge and trust: lower trust (Sara M = 2.00, Jane M = 2.17, Beth M = 2.67), moderate trust (Hannah M = 3.17, Nora M = 3.50), and higher trust (Hailey M = 3.67, Mia M = 4.17), computed from six Likert items.

Instruments and module. Teachers completed an "AI Trust Survey" adapted from Viberg et al. (demographics, AI knowledge, familiarity/experience, self-perceptions, and a six-item Trust in AI Tools scale plus two open-ended items) before and after a two-week online asynchronous module drawing on Ko et al.'s Critically Conscious Computing — Week 1 on data encoding and its limitations/bias, Week 2 on comparing human vs. AI intelligence and Machine Learning basics (training, testing, prediction), including building models with Google's Teachable Machine.

Analysis. Deductive first-cycle coding used the four Ethics categories of the Intelligent-TPACK scale (transparency, accountability, fairness, inclusiveness); a second cycle applied inductive open coding to derive indicators, followed by cross-case comparison. Trustworthiness relied on collaborative coding, purposive participant selection, a case-study protocol, and thick description.

Technical knowledge underpins ethical recognition. Teachers without solid foundational AI technical knowledge struggled to recognize how human decisions shape AI systems, undermining their ability to evaluate and trust them — consistent with prior work (e.g., Lucas et al.'s 211-teacher study) that higher AI knowledge does not automatically raise trust and may even heighten skepticism.

Interconnected, dynamic ethical constructs. Fairness and accountability frequently emerged together in teachers' reflections, indicating the four ethical categories are not independent but relational; for example, teachers linked designer bias, quantification bias, data bias, and non-neutrality of technology to both fairness and inclusiveness.

Asymmetric attention to ethical constructs. Teachers frequently referenced fairness, inclusiveness, and accountability concerns, while transparency and human accountability in decision-making were addressed far less often — emerging as areas needing greater attention in Professional Development initiatives.

Distrust as its own phenomenon. Teachers expressed concrete distrust (e.g., skepticism about automated placement based on test scores, and about AI lacking emotional, social, and moral judgment), with several cases showing trust shifts over the module (e.g., Beth declining from 2.67 to 2.33; Hannah rising from 3.17 to 3.33) — changes not tested for statistical significance.

Trust as a relational outcome. Teachers' trust in AI was shaped by their ethical perceptions as much as by technical proficiency — pointing to the interdependence of knowledge and values in AI Literacy and to human-in-the-loop responsibility.

Implication for professional development. The findings argue that AI professional development must build both technical knowledge and ethical fluency so teachers can trust — and appropriately distrust — AI systems, a central aim of the knowledge base's Professional Development concept.

What this means for practice

• Instructors. Build technical AI knowledge before expecting ethical fluency: teachers without solid foundations struggled to recognize how human decisions shape AI systems, which undermined their ability to evaluate and trust them.
• Faculty developers. Give transparency and human accountability explicit attention; teachers raised fairness, inclusiveness and accountability far more often, leaving the other two constructs underdeveloped.
• Instructors. Teach distrust as a competence: several teachers showed warranted skepticism (for example about automated placement from test scores and AI's lack of moral judgment), so professional development should help teachers calibrate trust rather than maximize it.
• Administrators. Treat the four ethical constructs as relational rather than a checklist; fairness and accountability repeatedly appeared together in teachers' reflections.

Limitations

• A qualitative multiple case study of seven in-service K-12 teachers, purposively selected from 33 consenting participants in one graduate course at a single southeastern US university in summer 2025.
• Cases were chosen for divergent self-rated AI knowledge and trust, and trust scores came from six Likert items; the reported per-teacher trust shifts (e.g., 2.67 to 2.33, 3.17 to 3.33) were not tested for statistical significance.
• Data are written reflections and survey responses from a two-week asynchronous module, so the analysis captures stated ethical perceptions rather than observed classroom practice.
• Case-study generalization was not a goal; the authors sought transferability through thick description of seven bounded cases, so findings should not be read as representative of K-12 teachers broadly.

Citation

Ocak, C., & Caskurlu, S. (2026). Unpacking ethics-domain of intelligent-TPACK scale in relation to in-service teachers' trust and distrust. Computers and Education Open, 100321. https://doi.org/10.1016/j.caeo.2025.100321