具不同性格特質之語音助理對於使用者的信任度與行為之影響

The Influence of Voice Assistant's Personality Traits on User's Trust and Behavior

謝子萱(Tzu-Hsuan Hsieh)

指導教授 ： 游曉貞

國立臺中科技大學/設計學院/多媒體設計系碩士班/碩士(2019年)

摘要

隨著人工智能技術的進步與成熟，語音助理等智能設備已逐漸成為我們日常中不可或缺的夥伴，如Amazon Echo、Google Home…等，而該如何使語音助理在協助我們解決困難，甚至是減少生活孤獨感的同時，更具穩定的增加互動之間的信任感受，學者們提出了賦予語音助理性格或是人類象徵的建議，並且認為若要建立和維持人機互動情感，增加對其之信任度，也是其中關鍵因素之一，因此本研究旨在探討語音助理的性格與使用者信任之間的關聯性。研究包含兩個部分：第一部分前導研究是調查不同語音屬性對於語音助理的性格感受之影響，首先利用文字轉語音服務，控制音量(大小)、音頻(高低)、語速(快慢)，組成8個不同聲音樣本，並由線上問卷收集人們對每個樣本在五大性格的主觀感受評價，分析122份有效問卷以了解語音變項與性格感受之關係。第二部分正式實驗，探討語音助理具有不同性格特質，是否影響參與者對其的信任度與行為。利用語音助理擔任大學研究室秘書，並根據前導研究分析結果，選取具內向和外向性格的兩個語音樣本，作為語音助理的性格提示。實驗採取組間設計，共招募60位參與者，每位參與者單獨進入研究室內與其中一位語音助理互動，並完成實驗任務。過程中，研究者以奧茲巫師法於單面鏡後控制語音助理，並以參與者的自陳式量表和任務執行程度來蒐集他們對語音助理的信任度與行為數據。研究結果發現，語音助理的性格不同時，會影響參與者對其之信任感受與行為；參與者對具外向性格特質的語音助理較信任，且較願意聽從其指示。最後，根據本研究之結果，提出有關語音助理性格提示之建議，作為未來設計參考依據。

關鍵字

語音助理 ； 性格特質 ； 信任度 ； 奧茲巫師法

並列摘要

With the maturity of artificial intelligence technology, intelligent voice assistants (VAs), such as Amazon Echo and Google Home, have already become indispensable partners in our daily lives. However, how to design VA not only to assist people with basic tasks, reduce loneliness but also maintain social connections, is still unclear? AI experts have suggested that giving intelligent virtual assistants a personality could increase users' trust in them. Therefore, this study aimed to investigate the relevance between the personality of VA and users' trust. The study consisted of two parts, the pilot study and the experiment. The pilot study is to investigate the relationship between the voice variables and the perceived personality of VAs. We used the text-to-speech service to configure the speaking rate, pitch, volume; and created eight VA voice samples. Participants were recruited via an online questionnaire to rate each voice sample on the Big-Five personality scales. A total of 122 complete responses were received. The experiment aims to explore whether the personality of a VA affects its trustworthiness and participants' behavior. In the experiment, the VA was used as a virtual secretary for a laboratory in a university. Two voice samples, the perceived introvert voice and the perceived extrovert voice, were created and used as the personality cue of VA based on the finding of the pilot study. A total of 60 participants were recruited and distributed to one of the two experimental conditions, an introvert VA and an extrovert VA. Each participant individually entered the laboratory to interact with one of the VA and to complete the experimental task. Throughout the whole process, the Wizard of Oz method was adopted. One researcher remotely controlled the VA and recorded the participant's behavior behind the one-way mirror. Afterward, participants were requested to fill in a Likert scale questionnaire regarding their trust in the VA basing on the experience. The participants' self-report inventories and their degree of completion of the experimental task were analyzed. The study found that the VA's personality correlated with participants' trust and behaviors; the participants were more willing to trust the extroverted VA and follow their instructions. Finally, implications for VA personality design and suggestions for future work in this area are also provided based on this study.

並列關鍵字

Voice Assistant (VA) ； Personality Traits ； Trust ； The Wizard of Oz Method

參考文獻

英文文獻

1.Allport, G. W. (1961). Pattern and growth in personality. Oxford, England: Holt, Reinhart & Winston.

2.Aly, A., & Tapus, A. (2013). A model for synthesizing a combined verbal and nonverbal behavior based on personality traits in human-robot interaction. Paper presented at the Proceedings of the 8th ACM/IEEE international conference on Human-robot interaction.

3.Andrews, P. Y. (2012). System Personality and Persuasion in Human-Computer Dialogue. ACM Transactions on Interactive Intelligent Systems, 2(2), 1–27.

4.Ballard, D. I., & Seibold, D. R. (2006). The experience of time at work: Relationship to communication load, job satisfaction, and interdepartmental communication. Communication Studies, 57(3), 317-340.