Design language learning with artificial intelligence (AI) chatbots based on activity theory from a systematic review

Abstract Artificial Intelligence (AI) chatbots, with their ability to engage in conversations that resemble human interactions, have been increasingly applied to language teaching. Most recent review studies overlook student learning outcomes and the methods to achieve these outcomes in chatbot-supported language learning. Activity Theory (AT) offers a framework of elements and functions inside an activity system to accomplish desired objectives. This systematic study intends to specify student learning outcomes in a chatbot-supported setting and explain how various factors such as rules, tools, and division of labor work together to enhance learning outcomes in this environment. This review included 37 papers published from January 2014 to January 2025. The findings provide two empirical contributions: the four types of outcomes and the use of AT-based approaches to achieve these outcomes. Additionally, two practical suggestions are made: creating instructional design models for teacher-AI collaboration in chatbot-assisted language learning and developing professional AI chatbots for language education. Furthermore, five research directions are proposed: teacher-AI chatbot interactions, agentic outcomes, out-of-school context, chatbot and human-chatbot collaborations, and K-12 education setting. The findings indicate how to use factors from AT to assist students leaning language effectively with AI chatbots.