Presentation Is Your Chatbot Helping or Harming? A Framework for Designing Context-Specific AI Language Partners more

As AI chatbots become mainstream in education, instructors in particular are wondering if these chatbots help students prevail—or fail. In English for specific purposes (ESP) environments, chatbots offer a promising solution for the speaking practice gap in which students cannot get enough feedback from an instructor. However, generic chatbots like ChatGPT are developed to be “helpful” and will autocorrect errors, “interpret” unclear responses, and ignore silences, thereby doing the cognitive work for learners. There is still a debate on whether AI helps or inhibits ESL learning. Moreover, technostress still prevails for both students navigating unfamiliar learning methods and teachers forcing chatbots to fit contexts they were not built for. Here, I present a framework for designing AI learning partners for self-directed speaking practice in a large-class, tertiary-level ESP communication course (n=117). The tasks were designed to be adaptable to different role-playing scenarios, outcome targets, and language levels. I will share lessons learned from my struggles with the frequent changes in free chatbot services and my successes in engineering a “good enough” stable AI speaking partner for my ESP context. My experiences will help educators design or refine chatbot activities to match their learners, ESP objectives, and classroom reality.