Educational Assessment, Evaluation and Accountability Education

Securing validity in GenAI-mediated essay-writing assessment: an ecological study of Vietnamese EFL teachers’ agency

Vo, Thanh Huynh, Bui, Hung Phu
Springer Nature · 2026-10-10 · DOI 10.1007/s11092-026-09515-x

Abstract

Generative artificial intelligence (GenAI) is acknowledged to serve as a key feature of automated essay scoring and AI-powered feedback on academic writing. It has sparked concerns about its validity and equity in assessing second or foreign language (L2) writing. The present study examines Vietnamese English as a foreign language (EFL) teachers’ professional agency in using GenAI to assess undergraduate students’ academic essays. The researchers used an ecological approach to teacher agency, together with a unified perspective of validity, to conduct a multiple-case study with eight teachers from four universities. Semi-structured interviews, grading observations, focus group discussions, and assessment artefacts were used to collect data. Data were then analysed thematically. The findings indicate that the teachers were not merely passive adopters but were selective about GenAI tools for several reasons. They chose GenAI tools based on perceived availability and reliability, used institutional rubrics and task requirements as prompts, and considered AI-generated scores a reference to benchmark with human grading. In addition, they adapted scores, reworded and/or expanded feedback from GenAI, and identified the role of the teachers and GenAI tools in the writing assessment process, prioritising human professional judgment in grading decision-making. Such practices of teacher agency were affected by grading loads, facility constraints, inequitable access to professional learning, and a shortage of institutional guidance. The current study highlights the need for institutional guidelines and regular teacher training on the use of GenAI-mediated writing assessment.

Read at publisher →