尽管人工智能在常识推理和语义理解领域均取得了显著进展，但在实现人机互动有效互解方面仍面临重大挑战。本文以会话分析为理论依据，将人机互动视为社会互动的一个特殊分支，从时间性与交际者预期在互解中的作用、序列性与互解机制的关联、以及多模态会话分析在人机互动中的应用这三方面，综合阐释了会话分析在人机互动研究中的重要作用，最终提出会话分析对对话系统设计、评估与优化的针对性启示。

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.