AI-Assisted Formative Assessment in English Language Education: Student Autonomy, Learning Outcomes, and Conditions of Implementation

Abstract

This article presents an analytical review of contemporary empirical research at the intersection of three subject areas: AI-assisted formative assessment (AI-FA), self-regulated learning, and English as a Foreign Language (EFL) pedagogy. What distinguishes this work from existing reviews is the introduction of an original analytical framework in the form of a three-level model of implementation conditions, which makes it possible to systematize conflicting findings on the effectiveness of these tools. Analyses in the field tend to be limited to describing individual instruments. The proposed model differentiates between technological, pedagogical, and institutional conditions, demonstrating that the observable effects of AI-FA are determined not by the characteristics of any particular tool, but by the configuration of all three levels simultaneously. The findings lead to the conclusion that when the technological capabilities of AI-FA meet a well-prepared instructor within a supportive institutional environment, these tools produce consistent positive effects on both academic achievement and student self-efficacy. In other cases, the risk of underutilizing the formative potential of such systems remains high, even when the technology itself is technically sophisticated.