Artificial intelligence as a catalyst for transformative assessment: designing teacher literacy at the crossroads of ethics, pedagogy, and human relationships

The rapid proliferation of Generative Artificial Intelligence (GenAI) in educational settings necessitates a redefinition of assessment practices, raising crucial questions regarding the relationship between automation and pedagogical ethics. While AI offers opportunities for efficiency and personalization, there persists a risk of a technocratic drift that threatens the relational dimension of teaching. This study explores the perceptions of trainee teachers regarding the integration of AI in assessment, investigating the transformation of practices (RQ1) and the competencies required for an “AI-literate” teacher profile (RQ2). Adopting a convergent mixed-methods design, a survey was conducted on a sample of 662 teachers enrolled in qualifying courses (60 ECTS) at the University of Foggia and the University of Bergamo. Data collected via a CAWI survey were analyzed using Spearman’s correlation and thematic analysis supported by text mining. The data reveal a “conscious resistance”: despite possessing a good level of technical comfort, pedagogical trust in AI remains low (M = 4.83/10). Inferential analysis highlights a training paradox: trust does not correlate with the technical training received (ρ = 0.36), but rather with the perception of concrete support (ρ = 0.57). Qualitatively, concerns regarding dehumanization and a demand for a principle of subsidiarity emerge. The study suggests that technical competence is insufficient for AI adoption. A Critical AI Literacy that integrates ethics and pedagogy is required, restoring decision-making control to the teacher to ensure fair and humanizing assessment.