Speaker
Description
The rapid integration of artificial intelligence in high-stakes language testing has driven a profound paradigm shift in education, shifting proficiency certification away from traditional "brick-and-mortar" centers toward automated, digital-first assessment environments. In the Vietnamese context, where securing an English certification is critical for global mobility and domestic university graduation, the local market has long been dominated by the high financial demands of the IELTS and TOEFL "Gold Standard." Recently, the Duolingo English Test (DET) has disrupted this landscape by offering an affordable, rapid alternative. The purpose of this study is to critically evaluate the systemic affordances and limitations of the DET within the Vietnamese socio-educational landscape, exploring whether automated testing platforms truly foster educational equity or merely prioritize technological convenience at the expense of linguistic validity.
Methodologically, this paper employs a critical test evaluation framework, synthesizing Bachman and Palmer’s (1996) model of test usefulness—specifically balancing reliability, validity, impact, and practicality—with Messick’s (1989) conceptualization of consequential validity. This theoretical lens is applied to a qualitative analysis of current DET test specifications, machine-learning scoring models (including GPT-4 integration), and localized case studies of institutional adoption in Vietnam, contrasting conservative policies with progressive pathways.
The key findings indicate a stark difference between the test’s practicality and its construct limitations. On one hand, the DET significantly reduces financial burdens and removes geographic barriers for disadvantaged students in rural areas. On the other hand, the study highlights critical shortcomings: construct under-representation due to "atomistic" micro-tasks that fail to measure sustained academic writing, a "production gap" impacting long-form discourse, and a profound lack of pedagogical transparency caused by its "black box" proprietary AI scoring. Furthermore, rigid AI-driven proctoring acts as a blunt instrument that creates false positives for cheating, while infrastructure dependencies worsen the local digital divide.
This critique highlights an urgent need to cultivate AI literacy among TESOL educators so they can critically interpret automated results, decode algorithmic evaluation systems, and confidently guide students through digital formats. Furthermore, the local tutoring industry must pivot away from superficial test-prep shortcuts toward teaching authentic digital communication skills. Finally, the study calls on the Ministry of Education and Training (MoET) and institutional stakeholders to implement responsible, hybrid recognition frameworks and provide school-level infrastructure support, ensuring that emerging assessment technologies serve as secure, equitable gateways..
Keywords: Duolingo English Test, consequential validity, educational equity, AI literacy
