A cultural-context framework of AI-mediated formative assessment: perceptions and proficiency in Iran’s TEFL context

Abstract This convergent mixed-methods study introduces a novel cultural-context framework to explore Iranian TEFL learners’ perceptions of AI-mediated formative assessments compared to teacher-led assessments, emphasizing validity, reliability, fairness, and their links to self-regulated learning (SRL) and proficiency gains. Data from 346 learners (B1–C1) across Tehran, Isfahan, and Shiraz were collected over 12 weeks using surveys, interviews, focus groups, and Cambridge tests. Results reveal significant improvements in AI perceptions (e.g., validity: M = 28.4 to 34.6, p < 0.001), with C1 learners rating AI higher (validity: M = 36.7 vs. B1: M = 32.8). Both B1 and C1 learners demonstrated significant proficiency gains (writing: B1 +1.5/20, C1 +1.6/20; speaking: B1 +1.0/20, C1 +1.5/20), though between-group differences in gains were small and not formally tested. However, cultural deference sustained teacher-led preference (fairness: M = 42.9 vs. AI: M = 30.8). Institutional trust and technological familiarity drove adoption, while urban-rural disparities (e.g., Shiraz’s connectivity issues) tempered it. The proposed cultural-context framework highlights AI’s perceived usefulness and associations with proficiency gains in Iran’s collectivist TEFL setting. Qualitative findings suggest that these perceptions are likely shaped by cultural norms and infrastructure, supporting calls for hybrid AI-teacher models to enhance equitable technology-enhanced language learning in high-context educational settings.