Power is arguably the most significant factor in human social interactions, yet its role has remained elusive in prior research on interactions with robots and AI agents. This dissertation uncovers the pervasive yet largely unexplored role of power. I begin by systematically examining the definitions of power across various social science disciplines to understand its integral relationship with artificial agents. I then explore the existing literature on Human-Robot Interaction (HRI) through a lens of power, revealing its implicit yet omnipresent nature. Two empirical studies are presented to illustrate how power can enrich research in HRI and Human-AI Interaction (HAI), reconciling disparate streams of study and unveiling novel technological opportunities. Finally, I discuss how these insights can guide future researchers, engineers, and designers to develop technologies that are both responsible and influential. The key contributions of this dissertation are threefold: First, it underscores the critical role of power in Human-Agent interactions, urging us to view power not merely as a background factor but as a central theme in the development and analysis of intelligent technologies. Second, by applying French and Raven's theory of power bases, this dissertation showcases the varied forms and effects of power, such as how expert, reward, and legitimate power influence interactions with robots and AI systems. Lastly, it identifies new directions and guidelines for future research and practice, calling for a deeper exploration of power dynamics and in-depth ethical reflections on power. 權力是人類社會互動行為中最關鍵的要素，但在先前機器人及人工智慧代理人的互動研究文獻中，與權力有關的探討卻相當缺乏。本篇論文因此試圖解釋在這些互動中，權力關係其實無所不在，只是之前大多隱而不顯。此篇論文首先系統性地檢視社會科學不同領域對權力的定義，以理解其與人造代理人的密切關係。接著，我透過權力的視角探討現有的機器人互動（Human-Robot Interaction, HRI）文獻，說明權力關係在之前的研究中雖然很少被直接提及，但其實無所不在。接下來，我透過兩項實證研究，說明權力相關的觀點可以如何應用於機器人互動及人工智慧互動（Human-AI Interaction, HAI）的研究：可以整合先前看似不一致的研究結果，並展示了新的科技應用可能性。最後，我討論了這些洞見如何引導未來的研究人員、工程師和設計師——開發有影響力的新科技時，必須同時注意到科技應肩負的社會責任。 本論文的三大貢獻如下：首先，本論文強調了權力在人機互動中的關鍵角色：我們不應將權力關係僅視為背景因素，而應將其作為開發和分析人工智慧相關技術時的中心考量。其次，本論文透過應用 French 與 Raven 的權力理論展示了權力的多樣形式和影響，例如專家權力、獎勵權力及合法權力如何影響人與機器人或人工智慧系統的互動。最後，本論文為未來的研究和實作提出了新的方向和原則，呼籲人機互動（Human-Computer Interaction, HCI）學界與業界更深入地探索權力的動態及對其進行更深入的倫理思考。