表達技巧智慧回饋系統介面設計對使用者體驗的影響研究

Abstract

隨著人工智慧技術迅速發展，智慧回饋系統已成為提升表達、領導、抗壓等軟實力的重要工具。智慧回饋系統結合圖形使用者介面與對話式使用者介面，提供即時且個性化的學習回饋，並因生成式人工智慧的應用，使人機互動更加自然。然而，介面設計對於使用者體驗的影響及其作用機制，相關實證研究仍有限。本實驗以本實驗室產學合作開發之情緒表達人工智慧助手為實驗場域，操弄五種介面設計（包括純圖形介面、三種結合對話介面的AI角色互動，以及角色自主選擇介面），並隨機分派101名受試者至五組。參與者於使用系統後，填寫涵蓋確認性、有用性、滿意度與續用意圖四個構面的使用者體驗問卷，並以共變數分析法檢驗不同介面設計對使用者體驗的影響。研究結果顯示，不同介面設計在提升使用者體驗的四個構面上具有顯著差異。其中，結合對話介面並提供角色選擇權的設計，顯著提升使用者的確認性、滿意度與續用意圖；而純圖形介面成效相對較低。根據本實驗實證，建議教育科技開發應強化多元互動角色與自主選擇功能，以提升智慧回饋系統的友善度與學習成效。The recent advancement of artificial intelligence has enabled auto-mated feedback systems to become indispensable tools for improving presentational skills. The systems utilize graphical and conversational in-terfaces to provide real-time personalized feedback, and the incorporation of generative AI significantly improves the naturalness of human-computer interaction. Empirical studies investigating the contribution of interface design to user experience and underlying mechanisms are still scarce.This study utilized an industry-academia cooperation AI-based emotional expression feedback system as the experiment platform. Five types of interface designs were implemented: pure graphical interface, three conversational AI role-based interfaces, and one interface that al-lowed users to choose their AI role freely. A total of 101 participants were randomly assigned to these five groups. After system usage, the users completed a user experience questionnaire on four constructs: confirma-tion, perceived usefulness, satisfaction, and continuance intention. Re-gression analysis was employed to investigate the impact of various inter-face designs on user experience.The findings revealed substantial differences among the five design models across all four user experience dimensions. Most importantly, in-terfaces incorporating conversational interaction and role selection result-ed in considerable improvement in users' confirmation, satisfaction, and intention to use the system compared to the pure graphical interface alone. Based on these findings, incorporating various interaction roles and user autonomy is suggested for educational technology developers to enhance the usability and learning effectiveness of automated feedback systems.