Student Perspectives on Hybrid AI–Human Writing Assessment in Bangladeshi Tertiary EFL Education: A Sequential Explanatory Mixed-Methods Study

Abstract

Artificial intelligence (AI) tools are increasingly used by university students to revise English writing, yet their role in formal assessment remains contested. This sequential explanatory mixed-methods study examined Bangladeshi tertiary EFL students’ perceptions of hybrid AI–human assessment for academic writing. Eighty survey responses were received; after two invalid test entries were removed, 78 cases were analyzed. Eleven survey respondents then completed asynchronous written interviews. The instrument was informed by the Technology Acceptance Model and TPACK, while quantitative data were analyzed descriptively and with paired-samples tests and correlations. Students rated AI significantly more effective for mechanical skills (M = 3.756, SD = 0.805) than for higher-order skills (M = 3.128, SD = 0.697), t(77) = 5.765, p < .001, dz = .653. Two-thirds of the sample (84.62%) preferred a hybrid workflow, whereas one respondent preferred AI-only assessment. Thematic analysis identified efficiency in mechanics, trust in human contextual judgment, a speed–depth trade-off, and concern about opaque errors and grade security. The findings support low-stakes, transparent AI assistance with teachers retaining responsibility for interpretation, feedback, appeals, and final grades.