Despite notable advancements in artificial intelligence in the domains of common-sense reasoning and semantic understanding, achieving effective mutual understanding in human-machine interactions remains a significant challenge. Drawing on Conversation Analysis as the theoretical foundation, this paper conceptualizes human-machine interaction as a specialized branch of social interaction, and examines the pivotal role of Conversation Analysis in human-machine interaction research, focusing on three core aspects: the role of temporality and participants, anticipations in mutual understanding, the relationship between sequentiality and intersubjectivity, and the application of multimodal conversation analysis in this field. Finally, this paper proposes targeted implications of Conversation Analysis for the design, evaluation, and refinement of conversational systems.