聊天機器人互動體驗元素對聊天機器人個性和滿意度之影響

隨著人工智慧的興起與社群網路平台發展蓬勃，許多企業紛紛建立起品牌專屬的聊天機器人(Chatbot)，透過串連通訊、社群軟體的對話式商務，達到強化客戶情感聯繫的目的。為此，形塑Chatbot的個性以符合企業形象，就成為了Chatbot 互動設計的重要議題之一。本研究欲透過實驗設計來探索非語言互動的設計元素如何影響使用者對Chatbot的個性感受。本研究實驗分為兩階段，前導研究的自變項有三：Chatbot頭像設計、表情符號的使用與否、貼圖的使用與否，依此設計 2X2X2的受測者內實驗。以田口直交表縮減成6個實驗條件，並以動態模擬方式製作出6個條件下的Chatbot互動過程片段。依變項為Aaker提出的五種品牌個性尺度，透過線上問卷平台施測回收185份有效問卷，了解非語言互動設計對Chatbot品牌個性之影響，並依據複迴歸分析找出Chatbot 品牌個性的預測值，前導研究結果為：(1)Chatbot的頭像、表情符號與貼圖確實會對受測者的「真誠」與「振奮」個性感知產生影響。(2)頭像設計影響Chatbot給人「雅緻」的感受。(3)貼圖影響Chatbot給人「真誠」、「振奮」、「稱職」的感受。(4)表情符號Chatbot 給人的「振奮」感受影響較大。第二階段為正式實驗，透過前導研究分析的結果，組合出在「真誠」構面下最高分與最低分的樣本組合，作為自變項。在Facebook Messenger平台上，以募資平台的智能助理角色，與受試者進行線上互動。此階段採組間設計，以線上問卷與點擊率為測量使用者滿意度的方法進行分析。研究結果發現：參與者對非語言互動體驗元素所組合的真誠品牌個性提示之聊天機器人的滿意度感受，高於不真誠聊天機器人。非語言互動體驗元素在訊息品質與持續意圖的構面是具有影響的。

With the rise of artificial intelligence and the vigorous development of social network platforms, many companies have established brand-specific chatbots to strengthen customer emotional connections through conversational commerce that connects communications and social software. For this reason, shaping the chatbot's personality to conform to the corporate image has become one of the crucial issues in the interactive design of the chatbot. This research explores how non-verbal interaction design elements affect users' personal feelings on chatbot through the experimental design.Ａtwo-phase study was designed to explore the non-verbal interaction design on chatbot personality and user satisfaction. The present study discussed the role of three variables on chatbot avatar design, use emoticons, and use stickers of chatbot. According to this, a 2X2X2 between-subject factorial design was employed. We used Orthogonal Array was reduced to 6 experimental conditions, and chatbot interaction process fragments under six conditions were produced by dynamic simulation. The dependent variables are the five brand personality scales proposed by Aaker. Subjects were recruited through an online participant; a total of 185 participants were returned to understand the impact of non-verbal interaction design on the personality of the chatbot brand. Analyses of variance were used to detect the significant difference among varieties. The pilot study results show: (1) Chatbot's avatars, emoticons, and stickers striking are striking effects of "sincerity" and "stimulation." (2) The avatar design affects the "elegant" feeling of chatbot. (3) Stickers affect the feelings of "sincerity," "stimulation," and "competence" of chatbot. (4) The emoticons elements a more significant impact of chatbot "stimulus" personality on people. Finally, multiple regression analysis the predictive value of the personality of the chatbot. The second stage is a formal experiment. The sample combined with the highest and lowest scores under the "sincere" dimension is combined as an independent variable through the pilot study and analysis results. We used an intelligent assistant in the fundraising platform on the Facebook Messenger platform to interact with subjects online. An inter-group design is adopted at this stage, and the online questionnaire and click-through rate are used to measure user satisfaction for analysis. The results of the study found that participants' satisfaction with the chatbots prompted by the sincere brand personality combined with non-verbal interactive experience elements was higher than that of the insincere chatbots. Non-verbal interactive experience elements are influential in the aspects of information quality and continuous intention.