綜上所述，本研究提出一個研究模型來探索使用者自我效能 （隱私自我效能、科技自我效能）是否會透過對語音助理的確認 （隱私確認、績效確認）與感知 （感知隱私風險、感知有用性）對手機語音助理服務的滿意度與持續使用行為產生影響，並探索各人口統計變數 （性別、年齡、教育程度、語音助理使用經驗、語音助理使用頻率）對整個研究模型的調節作用。本研究採用線上問卷調查的方式來收集樣本，總共回收835份有效問卷，並利用偏最小平方法 （Partial Least Square, PLS）來對樣本資料進行分析並驗證模型之假說。研究結果顯示，本研究所提出之研究假說全數成立，隱私自我效能對隱私確認有正面的影響；科技自我效能對績效確認有正面的影響；隱私確認對感知隱私風險有負面的影響；績效確認對感知有用性有正面的影響；感知隱私風險對滿意度有負面的影響；感知有用性對滿意度有正面的影響；滿意度對黏著度有正面的影響。此外，本研究也針對性別、語音助理使用經驗與語音助理使用頻率對特定假說的調節做進一步的檢驗與說明，最後本研究也針對研究中的限制進行討論，並提出未來的研究方向。 ;In recent years, the emergence of voice assistant has drawn increasing attention as it brings technological freshness and convenience to people and the daily lives. Despite its increasing popularity, the success of voice assistant lies in whether uses are willing to continue to use the service. To understand the factors influencing the continued use of voice assistants, this research adopts the expectation-confirmation model （ECM） as the theoretical basis to develop a research model for investigation. Specifically, the proposed model modifies ECM by replacing continuous use intention with stickiness to better represent the continuous use behavior in this study. As voice assistants involve disclosure of users’ privacy information, this study utilizes the privacy-calculus theory by including privacy confirmation and perceived privacy risk as potential factors to investigate the willingness to use the service given the privacy threat. The model also innovatively considers users′ psychological factors by including two types of user self-efficacy, i.e., privacy and technology self-efficacy in the investigation. A total of 835 questionnaires are collected and are analyzed by using PLS. The findings of the results are highlighted as follows. Firstly, privacy self-efficacy has a positive impact on privacy confirmation, which has a negative impact on perceived privacy risk. Technology self-efficacy has a positive impact on performance confirmation, which is found to also have a positive impact on perceived usefulness. Perceived usefulness and privacy risk have opposite effects on satisfaction, which has a positive effect on stickiness. This study also further illustrates the moderation of specific hypotheses by gender, experience with voice assistants, and frequency of voice assistant usage.
