Rethinking AI: bias in speech-recognition chatbots for ELT

Abstract Artificial intelligence (AI) technologies, particularly chatbots with speech-recognition, are gaining attention as tools for ELT. However, this frontline development in contemporary ELT seems to stand in stark contrast to the multilingual effort, another innovative trend, as chatbots’ speech recognition capabilities are primarily attuned to native English speaker pronunciation. Addressing this concern, this study adopted a Global Englishes (GE) perspective and examined students’ attitudes toward English after their use of a speech-recognition chatbot. Text-mining analysis of reflection papers from 38 students and thematic analysis of interviews with seven students revealed that the chatbot hindered the incorporation of diverse English use into ELT and students expressed a conception of English not as a language in flux but rather as a subject with specific content to be mastered. The findings are discussed concerning GE, AI ethics, and computer-assisted language learning, and implications for future chatbot use in ELT are provided.