Keyword: assessment literacy

ABSTRACT: Artificial intelligence (AI), particularly generative AI, changes assessment by altering the evidential relationship among student work, learning, and professional judgment. Teachers must design tasks when students can readily obtain machine-generated responses, interpret automated feedback and analytics whose validity is conditional, and justify consequential decisions under uncertainty. This critical integrative review synthesizes scholarship on AI in education, assessment, teacher learning, and emerging professional-development interventions. It proposes Assessment-AI Knowledge (A-AIK) as an assessment-specific configuration of professional knowledge rather than a replacement for assessment literacy, AI literacy, or technological pedagogical content knowledge. A-AIK comprises three interdependent capacities: design competence, which establishes conditions for eliciting meaningful evidence; interpretive competence, which calibrates the meaning, uncertainty, and evidential weight of AI-mediated information; and justificatory competence, which supports transparent, fair, and reviewable decisions. Disciplinary authenticity, human agency, equity, transparency and data governance, and epistemic humility operate as cross-cutting commitments. The reviewed literature suggests that promising interventions use scenarios, authentic assessment redesign, structured critique, peer deliberation, and iterative classroom trials. Recurrent limitations include small samples, reliance on self-report, short duration, and measures of confidence or acceptance rather than demonstrated assessment performance. A four-phase professional-learning model—calibrate, redesign, enact, and justify, is proposed as a synthesis-informed design hypothesis requiring empirical evaluation. The article concludes with provisional performance indicators, institutional conditions, and a research agenda for responsible AI-mediated assessment.