This paper focuses on conversational coordination skills, essential for learners of a second language. The “trial and error conversation process” often occurs in real life; clarification and interaction are carried out when misunderstandings or insufficient communication of intentions happen. This paper poses the research question, treating conversation not as guessing the right answer but as interaction through trial-and-error. We then utilize the Pragmatics Understanding Benchmark(PUB) to design an evaluation framework that measures contextual interpretation ability, with a focus on implicature understanding. Based onthese ideas, we develop a prototype CALL system using a large language model, providing learners with a learning environment where they can “experience communication failures and try conversations again through trial-and-error”. Although the experimental results showed no significant short-term improvement in learners’ interpretation ability, many participants engaged in trial-and-error strategies: rephrasing across multiple turns and asking clarification questions. These results suggested that the system provided a “safe conversation practice environment.” It was suggested that learners could experience failure without psychological burden and express their learning through trial and error.
