近年來，人工智慧（AI）研究在語言教育領域日益受到重視。相關應用研究指出，其主要優勢之一是可以讓學習者在不感到害羞的情況下練習會話。然而， 現有研究甚少從第二語言焦慮（Second Language Anxiety，以下簡稱SLA）的理論框架提供實證支持。本研究以SLA 理論為基礎，針對與雜談AI（以下簡稱AI） 進行雜談的初級學習者LY（以下簡稱LY）進行研究，並採用追蹤訪談（以下簡稱FI）及談話分析來探討學習者的語言使用與體驗。FI 結果顯示，LY 主要在理 解層面上經歷SLA。學習者反映，對AI 的許多發話內容，包括詞彙與語法結構，理解困難，並希望教師能以台灣華語提供說明。談話分析亦顯示，LY 不僅對片假 名詞彙存在困難，對於與台灣華語意思不同的漢字詞彙亦理解有限。此外，也觀察到與雜談語境不相符的提問與話題轉換現象。本研究結果顯示，初級學習者特 別需要在理解方面獲得教師的針對性支援，並且需要在雜談情境中進行適切提問與話題轉換的練習。

Recent research in language education has highlighted the benefits of using AI, such as allowing learners to practice conversation without embarrassment. However, few studies have examined these interactions from the perspective of Second Language Anxiety (SLA). This study investigated beginner learners LY (LY) engaging in casual chats with a chatbot AI (AI), using follow-up interviews (FI) and discourse analysis. Results showed that LY experienced SLA in comprehension. LY struggled with vocabulary in AI’s utterances. Discourse analysis also revealed difficulties with katakana and kanji words whose meanings differ from Taiwanese Mandarin, as well as inappropriate questions or topic shifts. These findings suggest that beginner learners need support in comprehension and practice in contextually appropriate questioning and topic management during casual conversations.