人機互動方式對建立夥伴關係的影響:服務型機器人與陪伴型機器人的差異

Abstract

科技的進步使機器人的功能由傳統提供機械式勞力的性質發展為具有陪伴和服務性質的功能，因此對於機器人的需求就不僅是硬體功能上的正確性或精準性，而是注重使用者與機器人的互動品質，其中機器人讓使用者信賴並願意與之合作是重要的關鍵。本研究採用合作式知覺匹配作業做為量測參與者對機器人接納意願的指標，以人機間不同的互動方式來探討哪些是增進人機關係的要素。在知覺匹配作業中，參與者需要記憶三種形狀（圓形、正方形、和三角形）與三個名字（最好朋友、機器人、與陌生人）的配對連結關係，作業要求參與者對隨機出現的名字和圖形組合的正確性做出快速且正確的判斷。參與者被告知將與機器人一起進行此作業，量測是否有夥伴優勢的出現（即參與者對機器人名字和圖形的組合反應速度快於陌生人名字和圖形的配對，或更進一步與好友名字圖形的配對一樣快）。系列一的研究比較人-人互動和人-機互動的差異，探討參與者對機器人是否也和人類一樣會產生夥伴優勢。結果顯示在僅見機器人一面而沒有互動的情況下，無論是參與者獨自完成作業（實驗一）、名字與形狀隨機搭配（實驗二）、與機器人合作完成作業（實驗三）、甚至實驗前未見到機器人實體（實驗四）皆有對機器人表現優於陌生人的夥伴優勢。此結果與人-人互動的夥伴優勢類似，顯示無論對人類或機器人不需實際見面，只需口頭被引介有此夥伴共同工作即可建立夥伴優勢。系列二則進一步研究機器人與受試者不同的互動方式（服務型功能或陪伴型功能）對人機關係的影響。結果發現當機器人展現服務功能（如自然語音系統、觸控、移動）的情境時，對機器人的夥伴優勢程度可增加到和好友相同的程度（實驗五），但如果在後續測驗時機器人持續待在參與者身旁則夥伴優勢退到僅優於陌生人（實驗六）。與機器人作為陪伴功能的結果則相反，向受試者求助時須在作業過程中陪伴在參與者身旁才得到作業表現與好友無異的較強夥伴優勢（實驗七、八）。綜合本研究的發現顯示，使用者對機器人的接受程度會因機器人所扮演的角色不同而有影響：使參與者感到被需求且持續陪伴在身邊的情況下則能有和好友相同的夥伴優勢。不同於一般的推測，機器人除了需有為人類服務的功能外，讓人類反向協助機器人產生被需要的感受也是影響人機接受度的重要因素。

With the advancement of technology, especially artificial intelligence, functions of robots have been developed from merely providing mechanical labor to companionship, and thus the demands for robots have extended from functional utility to quality of human-robot interactions. We examined the factors that affect the establishment of a partnership when the partner is a robot, using the perceptual matching task. In the task, participants were asked to remember the matching relationship between shapes (circle, square, and triangle) and names (best friend, robot, and stranger). Participants’ acceptance of the robot as a partner was defined by the presence of partner advantage (i.e., participants responded to the combination of the robot's name and shape faster than stranger's name and shape, or, one step further, as quickly as best friend's name and shape). To compare human-human interaction with human-robot interaction, in the first series of studies, we used a similar experimental design as in a previous study examining human-human interaction (Cheng Tseng, 2019) to measure whether participants had similar partner advantage toward robots as that toward human. Results showed the same level of partner advantage when participants saw the robot once and then finished the task alone (Experiment 1), with correspondent shape randomized (Experiment 2), remotely cooperated with the robot (Experiment 3), or even without seeing the robot al all (Experiment 4). These results are consistent with human-human interaction in that no face-to-face interaction is required for partner advantage as long as the participant is introduced to the robot as a partner even by instruction alone. In the second series of studies, we further investigated the effect of different interactions (robot providing service or companionship) on building a human-robot partnership. Results showed that when the robot displayed service functions (such as natural language system, touch control, and movement), the strength of partner advantage increased to be like a best friend (Experiment 5). However, the advantage returned to be like a stranger if the robot continued to stay near the participants during the task (Experiment 6). In contrast, in the situation in which the robot as a companion (asking for help), the robot had to accompany the participant during the task to obtain the partner advantage as good as with a best friend (Experiment 7 and Experiment 8). We conclude that the role plays by the robot would affect the subsequent human-robot interaction, and if the situation is consistent in context, participants’ feeling of partnership toward the robot would increase. In addition to having the robot to provide service functions, let humans assist the robot to produce the feeling of being needed is also important for a better human-robot partnership.