| 摘要: | 在數位行銷與電子商務的發展下，個人化推薦技術已成為提升消費者體驗的核心策略。而過度個人化可能導致個人化疲勞、個人化悖論及過濾氣泡等問題，降低對平台的互動體驗。為應對問題，過往研究利用調整演算法來優化推薦系統，而本研究旨在探討激將與鼓勵風格的個人化銷售機器人如何透過促進因素(感知有趣、親密感)與抑制因素(使用者疲勞、操縱感知)的心理感知構面影響其行為意圖(分享意圖、持續使用意圖)，以改善個人化推薦技術所面臨的挑戰，並探討使用者個性（樂觀、悲觀）與機器人風格交互作用對其影響。
為了深入探討這一現象，本研究基於SOR模型並納入雙因子理論而發展出研究模型，並實際設計了三種不同風格的個人化銷售機器人。首先，標準型機器人提供個人化推薦，專注於根據使用者歷史行為推送相關商品；正向型機器人透過積極鼓勵回饋來提升使用者的參與度，增強情感連結與接受度；挑戰型機器人利用反向心理學設計，以激將法激發使用者的興趣與動機。透過實驗設計、開發實際機器人系統、問卷調查，進一步將問卷調查中有效樣本資料進行結構方程模型的分析。結果顯示，標準型風格機器人對抑制因素有顯著影響，並進一步減少使用者的行為意圖；正向型與挑戰型風格則能顯著增強促進因素，並有效緩解使用者疲勞感，但對操縱感知構面影響不大；此外，使用者個性調節的結果顯示，樂觀情緒的使用者對正向型和挑戰型風格的機器人有更積極的反應；而悲觀情緒的使用者則對標準型風格的機器人有較好的適應性。
本研究補足對風格設計與個性調節因素探討的不足，為實務界提供有價值的見解，驗證不同風格的個人化銷售機器人的有效性，並提供電商平台在個人化推薦領域的新策略，幫助品牌優化使用者體驗，提升與平台或系統的互動性。
;With the rise of digital marketing and e-commerce, personalized recommendation technology has become a core strategy for enhancing consumer experience. However, over-personalization can lead to personalization fatigue, the personalization paradox, and filter bubbles, causing consumers to resist repetitive, non-novel content and reducing their engagement with the platform. To address these issues, prior studies have adjusted recommendation algorithms to mitigate the negative effects of over-personalization. This study examines how different styles of personalized sales robots influence users’ behavioral intentions (sharing intention, continuance intention) via promotive factors (perceived fun, intimacy) and inhibitive factors (user fatigue, perceived manipulation), and further investigates the interaction between user personality (optimism, pessimism) and chatbot style.
Based on the SOR model and dual-factor theory, we developed a research framework and implemented three styles (standard, positive, and challenge) of personalized sales robots. We conducted an experiment using are fully functional chatbot system combined with a questionnaire survey, then analyzed valid responses with structural equation modeling. Results show that the standard style significantly amplifies inhibitive factors, thereby reducing users’ behavioral intentions; in contrast, the positive and challenge styles both significantly enhance enabling factors and effectively alleviate user fatigue, while exerting minimal influence on perceived manipulation. Additionally, moderation analysis indicates that optimistic users respond more favorably to the positive and challenge styles, whereas pessimistic users adapt better to the standard style.
This study addresses the research gap in style design and personality moderation, provides valuable insights for practical applications, validates the effectiveness of personalized sales chatbots with different interaction styles, and offers new strategic directions for e-commerce platforms in personalized recommendation. These findings can help brands optimize user experience and enhance engagement with the platform or system. |
