以科技接受模型三（TAM 3）探討商務華語中級學習者使用生成式AI進行模擬求職面試之接受度與挑戰

A TAM 3 Study on the Acceptance and Challenges of GenAI-Enabled Interviews for Business Chinese Intermediate Learners

本研究探討生成式人工智慧在華語作為外語教學中的應用，特別聚焦於求職面試情境下的學習輔助效果與學習者回饋。以科技接受模型三（Technology Acceptance Model 3，TAM 3）為理論架構，從感知易用性、感知有用性與行為意圖三個構面，分析中級華語學習者對生成式AI面試教學代理人的接受情形。研究對象為10位具碩士學歷的外籍商務華語學習者（美國籍4位、哥倫比亞籍1位、越南籍3位、印尼籍2位），透過ChatGPT進行中文模擬面試，並結合語音辨識與輸出技術，以貼近真實面試情境。教學活動後，透過半結構訪談蒐集學習者使用經驗與回饋。結果顯示，在感知易用性方面，學習者意見呈現分歧：部分認為系統操作簡便且有助於面試準備，但亦有學習者指出語音輸入困難，特別是在中文聲調與詞語辨識上仍有限制，影響整體使用體驗。在感知有用性方面，多數學習者肯定生成式AI能協助熟悉面試流程並提供即時回饋，但同時指出其無法有效糾正發音與語調，且回饋內容深度不足，缺乏人性化理解。在行為意圖上，多數學習者表示若語音辨識功能改善，將願意持續使用並推薦他人，惟普遍認為AI無法取代真人教師與面對面互動的重要性。整體而言，生成式AI在模擬面試教學中具備潛力，但仍需提升語音辨識與回饋品質，並與真人教學結合，以強化學習成效。TAM 3分析顯示學習者對此類工具持中度接受態度。本研究補充了生成式AI應用於華語教學的實證基礎，並為未來教育科技設計與實施提供參考。

This study examines the application of generative artificial intelligence in teaching Chinese as a foreign language, with a particular focus on its effectiveness as a learning support tool in job interview contexts and on learners' feedback. Using Technology Acceptance Model 3 (TAM 3) as its theoretical framework, the study analyzes intermediate Chinese learners' acceptance of a generative AI interview teaching agent in terms of perceived ease of use, perceived usefulness, and behavioral intention. The participants were ten international learners of Business Chinese, all holding master's degrees (4 from the United States, 1 from Colombia, 3 from Vietnam, and 2 from Indonesia). They used ChatGPT to conduct simulated Chinese job interviews, integrated with speech recognition and speech output technologies to approximate authentic interview situations. After the instructional activity, learners' experiences and feedback were collected through semi-structured interviews. The results show mixed opinions regarding perceived ease of use: some learners found the system easy to operate and helpful for interview preparation, while others reported difficulties with voice input, particularly because of limitations in recognizing Chinese tones and vocabulary, which negatively affected the overall user experience. In terms of perceived usefulness, most learners agreed that generative AI helped them become familiar with interview procedures and provided immediate feedback, but they also noted that it could not effectively correct pronunciation or intonation, and that its feedback lacked depth and human-like understanding. Regarding behavioral intention, most learners indicated that they would be willing to continue using the system and recommend it to others if the speech recognition function were improved, although they generally believed that AI could not replace the importance of human teachers and face-to-face interaction. Overall, generative AI shows promise for simulated interview instruction, but improvements in speech recognition and feedback quality, as well as integration with human teaching, are still needed to enhance learning outcomes. The TAM 3 analysis indicates that learners demonstrated a moderate level of acceptance of this type of tool. This study adds to the empirical foundation for the use of generative AI in Chinese language teaching and provides a reference for the future design and implementation of educational technology.