EFL Students’ Error Detection, Correction, and Acceptance Calibration: A Performance-Based Study at the Faculty of Education–Surman

Abstract

Generative artificial intelligence (GenAI) is increasingly used in English language learning for writing support, grammar correction, feedback, translation, and academic language production. Its educational value depends not only on students’ ability to use AI tools but also on their ability to evaluate AI output and make appropriately calibrated reliance decisions. This study investigated EFL students’ performance in detecting and correcting linguistic errors in a 30-item English sentence evaluation task and examined the correspondence between their acceptance or rejection decisions and the observed correctness of the items. A quantitative descriptive-correlational design was used with 30 EFL students who evaluated 17 correct sentences and 13 sentences containing targeted errors. For each item, participants judged correctness, provided a correction when necessary, indicated whether they would accept the sentence without further checking, and rated their confidence on a five-point scale. Detection, correction, acceptance calibration, overacceptance, and underacceptance were summarized using means and standard deviations. Spearman’s rank-order correlation was used to examine associations between performance and acceptance calibration. Mean detection performance was 92.22% (SD = 4.23), mean correction performance was 91.28% (SD = 8.01), and mean acceptance calibration was 75.33% (SD = 8.51). Mean overacceptance was 17.18% (SD = 10.24), whereas mean underacceptance was 30.39% (SD = 13.56). No statistically significant association was observed between detection and acceptance calibration (ρ = .056, p = .770) or between correction and acceptance calibration (ρ = .003, p = .987). These findings indicate that accurate linguistic evaluation and calibrated acceptance decisions should not be treated as interchangeable outcomes. Within the limits of the small, non-probability sample and the single task used in this study, the results support the inclusion of explicit verification and critical evaluation activities in EFL instruction.