Automatic Vs Human Evaluation of Students' Spoken English Proficiency: A Study of Students' Perceptions

Abstract

Assessing students' spoken English proficiency is a vital component of language education. With recent advances in artificial intelligence (AI) and natural language processing (NLP), automated speaking assessments have emerged as alternatives to traditional human evaluations. This study investigates undergraduate students’ perceptions of human versus automated speaking assessments within an English Language Education program at the State University of Makassar. Grounded in Swain’s (2006) Languaging Theory, the research explores how different assessment modes influence learners’ motivation, anxiety, engagement, and perceived feedback quality. Employing a descriptive qualitative design, data were collected through semi-structured interviews with nine students who had experienced both human and automated assessment types. Thematic analysis revealed a strong preference for human evaluations due to their ability to capture contextual, emotional, and expressive aspects of speech. Human feedback was perceived as more personalized, meaningful, and motivational. In contrast, automated systems were valued for their efficiency, objectivity, and usefulness in practicing technical elements such as pronunciation and fluency. However, students noted limitations in the depth and clarity of automated feedback. Most participants advocated for a hybrid model that combines the strengths of both approaches. These findings yield important pedagogical implications for the incorporation of artificial intelligence in language assessment, while emphasizing the continued relevance of human involvement in fostering communicative competence