AI in Formative Assessment Scoring Practices: A Policy Research Study

Abstract

This study investigates the role of artificial intelligence (AI) in formative assessment scoring practices within a private high school (Grades 9–12), with a specific focus on consistency, efficiency, and ethical governance. Grounded in a qualitative, policy-oriented research design, the study analyzes international frameworks and school-level policy documents to examine how AI-supported assessment practices are conceptualized, regulated, and aligned with pedagogical principles. Drawing on Assessment for Learning (AfL) and Messick’s Unified Theory of Validity, the research positions consistency and efficiency as enabling conditions that support valid interpretation and use of assessment evidence rather than as indicators of validity in themselves. Findings indicate that AI-based scoring systems are widely framed as tools that can enhance consistency by standardizing evaluation criteria and improve efficiency through timely feedback delivery. However, policy documents consistently emphasize the necessity of teacher professional judgment and human oversight to ensure appropriate interpretation and instructional use of AI-generated outputs. The study also reveals that while international frameworks provide broad ethical principles—such as transparency, fairness, and accountability—school-level policies tend to remain general and lack detailed operational guidance for implementing AI-supported formative assessment practices. The research highlights a gap between technological innovation and policy development, where AI tools are often adopted faster than governance structures are established. This misalignment may lead to variability in implementation and uncertainty in professional responsibilities. The study concludes that effective integration of AI in formative assessment requires coherent policy frameworks that balance efficiency with ethical responsibility, reinforce teacher interpretive authority, and ensure alignment with learning-centered assessment practices. These findings contribute to the development of governance-informed approaches that support responsible and pedagogically sound use of AI in education.