游曉貞蕭方雯2025-08-112025-08-112025-06-30U0061-1309202211162200https://www.airitilibrary.com/Publication/alDetailedMesh1?DocID=U0061-1309202211162200https://nutcir-lib.nutc.edu.tw/handle/123456789/740隨著科技的進步，人工智能（Artificial Intelligence）被運用到許多地方，近年來AI已經成為個人與企業的重要話題，特別是考慮到語音助理用戶的增長，公司與消費者的互動需要改變，各科技企業將語音助理的功能添加到行動裝置及家電。不少學者發現塑造語音助理的個性可以為使用者帶來良好的互動體驗，因此語音助理個性成為近年語音互動設計重要的議題。但是目前鮮少透過聲音屬性來操控或影響語音助理個性的設計準則，也缺少針對不同應用情境下該如何為語音助理選擇適當的個性知識。 因此本研究分成兩個階段，第一階段是透過實驗設計探討如何透過聲音去塑造語音助理的個性，自變項有四個：性別元素（男性聲音、女性聲音）、語速（快、中、慢）、音調（高、中、低）、音量（大、中、小），依此設計2X3X3X3組內設計實驗，透過田口式直交表縮減為九項不同的聲音樣本組合，並在YouTuber平台播放，依變項是語音助理給人的個性感受，依據Aaker (1997)品牌個性尺度量表所設計的線上問卷讓受測者填寫。透過線上問卷平台回收之280份問卷進行統計分析，第一階段研究結果發現：（1）以上述聲音元素組合，受測者可以感知「真誠」、「振奮」、「稱職」、「高雅」、「強健」的個性差異；（2）音調（中」為真誠個性感知中最具影響的聲音元素，最不具影響的聲音元素為性別（男）。音調（高）為振奮個性中最具影響的聲音元素，最不具影響的聲音元素為語速（快）。語速（中）為稱職個性感知中最具影響的聲音元素，最不具影響的聲音元素為音（高）。音調（中）為高雅個性感知中最具影響的聲音元素，最不具影響的聲音元素為性別（男）。語速（快）為強健個性感知中最具影響的聲音元素，最不具影響的聲音元素為性別（女）。 第二階段是檢視不同語音助理個性對網購使用者信任度與滿意感受之影響，透過前導實驗的結果，組合出「真誠」與「不真誠」的聲音樣本組合當作自變項藉由奧茲巫師法，將實驗情境設定為語音網路購物，招募兩組受測者分別與「真誠」及「不真誠」的語音助理互動，之後再透過信任度量表與滿意度量表收集受測者的感受與評價當作依變項，第二階段研究結果發現：（1）受測者對具備真誠個性的語音有較高的信任與滿意感受；（2）受測者比較傾向與具備真誠個性的語音助理互動次數較多。With the advancement of technology, AI (Artificial Intelligence) has been applied in many fields. In recent years, AI has become a popular topic for individuals and enterprises. Especially considering the growth of voice assistant users, the interaction between companies and consumers gradually became necessary to change. Various technology companies are also emerging to add the function of voice assistants to mobile devices and home appliances. Many scholars found that shaping the personality of voice assistants can bring users a positive interactive experience. The personality of voice assistants has become a critical topic in voice interaction design in recent years. However, there is currently a lack of design criteria for manipulating or affecting the personality of voice assistants through sound attributes and a lack of knowledge on choosing an appropriate personality for voice assistants in different application scenarios. Therefore, this research is divided into two stages. The first stage is to explore how to shape the voice assistant's personality through experimental design. There are four independent variables: gender elements (male and female voice), speech speed (fast, medium, and slow), pitch (high, medium, and low), and volume (large, medium, and small). Through the Taguchi-style orthogonal table, the 2X3X3X3 in-group design experiments and reduced to nine different sound sample combinations. Using the YouTube platform, the dependent variable is the personal feeling of the voice assistant, and the subjects are asked to fill in the online questionnaire designed according to the Aaker (1997) brand personality scale. Statistical analysis was conducted on 280 questionnaires collected through the online questionnaire platform. The results of the pilot experiment revealed that: (1) The combination of the above sound elements, participants could perceive differences between sincerity, excitement, competence, sophistication, and ruggedness personality (2) Pitch (middle) was the most influential vocal element in the perception of sincerity, and gender (male) was the least influential vocal element. Pitch (high) is the most influential sonic element in excitement personality, and the least influential sonic element is speech rate (fast). Speech rate (middle) is the most influential sound element in the perception of competence personality, and the least influential sound element is pitch (high). Pitch (middle) was the most influential vocal element in the perception of a sophisticated personality, and gender (male) was the least influential vocal element. Speech rate (fast) was the most influential vocal element in ruggedness personality perception, and gender (female) was the least influential vocal element. The second stage experiment is to examine the influence of voice assistant personalities effecting trust and satisfaction in online shoppers’ perceptions. Through the results of the pilot experiment, a combination of "Sincerity" and "Insincerity" voice samples were used as an independent variable by the Austrian. Using the Wizard Method, the experimental situation was set as online voice shopping. And two groups of subjects were recruited to interact with the "Sincerity" and "Insincerity" voice assistants and then collected through the trust and satisfaction scales. Subjects' feelings and evaluations were regarded as dependent variables. The results of the second stage experiment found that: (1) the participants had higher trust and satisfaction with the voice with a sincere personality, (2) the participants were more inclined to interact with the voice assistant with a sincere personality.zh語音助理語音電子商務購物品牌個性信任滿意voice assistantvoice e-commerce shoppingbrand personalitytrustsatisfactionmaster thesis