Usability Study of GenAI for English Learning in VR
Bee Sian Tan · Pakistan Journal of Life and Social Sciences (PJLSS) · 2024
Traditional English learning environments in university often hindered by outdated focus on reading and writing, limited textbook content, insufficient speaking practice opportunities, and pre-programmed artificial intelligence (AI) in English speaking practice.This study explores the potential of leveraging VR technologies and generative AI (GenAI) to overcome these barriers.EasyEnglish is a real-time conversation game with GenAI non-player character (NPC)s.This game utilizes voice input recognition, large language model (LLM) for language assessment and textto-speech (TTS) for NPC lip-sync animation.The content validity was assessed by 3 experts to evaluate the conversation quality.A usability test was conducted using a purposive sampling method with 7 undergraduate non-native English speakers who have less than 1 year experience in using VR technology.This study employed System Usability Scale (SUS) and Content Validity Index (CVI) metrics for assessment.The CVI result showed satisfactory agreement in conversation quality but highlighted areas for improvement in learning objectives.The SUS result revealed satisfaction in consistency of user interface (UI) and learnability of EasyEnglish, while also highlighting the need for improvement in UI, setup, and visual cues.The significance of the study lies in the GenAI's ability to provide diverse response, avoid repetitive dialogue and speak using gestures to undergraduate students.GenAI effectively identify and assess irrelevant words in conversation, provide immediate grammar and vocabulary correction and suggestion of conversation improvement accurately.Future research should focus on improving accessibility, include more multimodal interactions, and mapping learning objectives with the soft skills emphasized by the World Economic Forum (WEF).