Exploring Assessment Adaptation Strategies for Authentic Student Evaluation in the Age of Generative AI: Insights from Educators at an Online School

Abstract

The rapid integration of generative AI into digital learning environments has challenged traditional assessment models, raising concerns about assessment authenticity and validity. This study explores how online educators perceive and adapt assessment strategies to uphold authentic student evaluation in AI-influenced learning contexts. Grounded in Validity Theory (Messick, 1995), Authentic Assessment Theory (Wiggins, 1990), and Situated Cognition Theory (Brown, Collins & Duguid, 1989), the study examines the tensions between AI’s impact on assessment integrity and the need for adaptive, process-driven evaluation strategies. Employing a qualitative, interpretivist case study design, this study investigates the perspectives of educators in a South African fully online school. Semi-structured interviews and survey data were thematically analyzed to capture nuanced educator responses. Findings reveal that educators widely perceive AI as a threat to traditional assessment validity, particularly in text-based, output-focused assessments. Language educators tend to adopt restrictive strategies such as AI detection tools and platform limitations, whereas STEM educators favor process-driven adaptations that emphasize reasoning, iterative work, and step-by-step demonstration. Despite broad recognition that restriction alone is unsustainable, adaptation remains highly inconsistent across subjects, constrained by curriculum mandates, institutional policies, and limited AI literacy training. The study highlights four key adaptation strategies. First, shifting from product to process-based assessment reinforces construct validity by focusing on cognitive engagement rather than final outputs. Second, personalized and contextualized tasks require real-world data collection or unique student inputs to mitigate AI-generated responses. Third, oral and multimodal assessments, though underutilized, offer AI-resilient alternatives to traditional text-based evaluations. Finally, structured AI integration moves beyond restriction toward scaffolded AI use in assessment to develop digital literacy and critical engagement. Ultimately, this study concludes that authentic assessment in AI-influenced education is not about restricting AI, but about redesigning assessments. To uphold fairness, validity, and educational integrity, institutions must provide clear policy guidance, AI literacy training, and research-informed adaptation frameworks. As AI reshapes knowledge creation and learning processes, assessment practices must evolve to maintain their credibility and relevance in digital learning environments. Keywords: Generative AI, online education, assessment adaptation, authentic assessment, assessment reform.