김혜영 Kim Heyoung , 이성숙 Lee Sung-suk
DOI: Vol.26(No.1) 31-58, 2023
Abstract
AI-based programs, especially voice chatbots, have recently been involved in second language research, but chatbots for effective L2 learning do not seem to exist yet. The essence of chatbots is human-like communication, exchanging meaning to perform living tasks, but unfortunately, current Korean applications are designed far from that essence. This study aims to verify the design effect of task-based chatbots compared to script role-playing chatbots commonly found in AI speaking programs. Six third-grade students participated in four different chatbot-based speaking sessions. Chatbots were designed in two forms: 'role-play chatbots' and 'assignment chatbots.' Chat logs, pre-and post-test, and questionnaire data were analyzed quantitatively and qualitatively. Participants actively engaged in conversation and acquired about 50% of the target words during both trials. However, a much greater variety of semantic negotiation was observed in the 'task chatbot,' and more participants successfully completed the task chatbot conversation than in the role play. They also memorized more words in the delay test and showed more interest and preference for 'task chatbots'. Chatbot design is essential, and task-based design should be applied to future speaking programs.
Key Words
artificial intelligence, AI, voice chatbot, task-based design, English speaking