The Impact of Explainable AI on Teachers' Trust and Acceptance of AI EdTech Recommendations: The Power of Domain-specific Explanations

The Impact of Explainable AI on Teachers' Trust and Acceptance of AI EdTech Recommendations: The Power of Domain-specific Explanations — Feldman-Maggor, Cukurova, Kent, and Alexandron (2025) adapt Hoff and Bashir's "trust in automation" model to AI in education, proposing that explainable AI (XAI) builds Trust in AI EdTech recommendations indirectly by increasing their understandability. In a mixed-methods, within-subject experiment with 41 in-service chemistry teachers using the AI recommendation tool GrouPer, they find that understandability, trust, and acceptance of AI recommendations are positively correlated and that domain-driven explanations — framed in curricular/pedagogical language — foster greater understandability and trust than purely data-driven (feature-importance) explanations. The authors also surface two situational factors shaping acceptance beyond trust: pedagogical alignment and workload-reduction potential.

Key Findings

• Understandability is the bridge between XAI and trust. Adapting Hoff and Bashir's (2013, 2015) model, the paper positions explainable AI as a design feature whose effect on dynamic, learned Trust is mediated by users' understandability of the system's performance — explainability raises trust by letting teachers validate AI outputs against expectations. All correlations among understandability, trust, and acceptance were positive and mostly strong (r > 0.5, one moderate r = 0.439), strengthening after domain-driven explanations were added (Trust Calibration).

• Domain-driven explanations outperform data-driven ones. Moving teachers from feature-importance ("data-driven") explanations to semantic, curricular-language ("domain-driven") explanations significantly increased understandability (W = 80.5, p = 0.005), learned trust (W = 52, p = 0.002), and acceptance (W = 22.5, p = 0.003). All seven think-aloud teachers reported domain-driven explanations as more influential for building their trust than data-driven ones — evidence that explanations should "speak" the teacher's pedagogical language, not just expose model internals.

• Trust is dynamic and requires validation, not just explanation. Qualitative analysis showed that for some teachers, understandability alone was insufficient to establish trust — several stressed that real classroom experience with the tool was needed before they would fully rely on it, highlighting the situated, dynamic nature of learned Trust rather than a one-shot transparency effect.

• Acceptance also depends on situational factors beyond trust. Bottom-up analysis of think-aloud protocols revealed two additional factors influencing teachers' willingness to accept AI recommendations: pedagogical considerations (whether the tool aligns with their approach to differentiating instruction, reported by 8 of 11) and the workload-reduction potential of the tool (reported by 6 of 11), which the authors treat as acceptance drivers not directly tied to trust and grounded in exploratory qualitative evidence.

Connected Concepts

• Trust
• Trust Calibration
• Teacher Role
• Teacher AI Competency
• Technology Acceptance Model
• Edtech Platform
• AI Education

Connected Articles

• Intelligent TPACK Ethics Teachers Trust Distrust 2026
• Mind The Trust Gap Teacher Student Views Control Agency K12 Classroom AI
• Activity Theory Teachers Adoption AI Sem 2026
• XAI Education Framework
• Mejia Domenzain ML Findings Teachers Blended 2026

Citation

Feldman-Maggor, Y., Cukurova, M., Kent, C., & Alexandron, G. (2025). The Impact of Explainable AI on Teachers' Trust and Acceptance of AI EdTech Recommendations: The Power of Domain-specific Explanations. International Journal of Artificial Intelligence in Education, 35, 2889–2922.