피지컬 AI의 외형 디자인과 생산자․소비자 담론의 인지의미론적 탐색 = A Cognitive Semantic Exploration of Physical AI Design and Its Discourse

다국어 초록 (Multilingual Abstract)

This paper classifies Physical AI into humanoid and non-humanoid robots and examines, from a cognitive perspective, both their appearance design and the robot-related discourses constructed by producers and users. The main findings of this study are as follows.
First, a robot’s appearance design begins with identifying a source domain that provides conceptual material to be referenced through conceptual metaphor. Once a source domain is established, only a subset of its elements is selectively incorporated into the robot’s appearance through conceptual metonymy.
Second, the appearance design of humanoid robots adopts person as its source domain and models human morphology and behavioral patterns. However, such design does not remain at the level of simple imitation; depending on contextual demands grounded in social needs, different elements within the source domain may be selectively recruited and undergo transformation. Focusing on industrial-site robots whose prototype is the “adult male,” this paper analyzes diverse patterns of transformation and their motivations, and, through comparison with educational robots, identifies differences in intra-domain element selection according to the robot’s intended purpose.
Third, among non-humanoid robots, the appearance design of robot dogs takes the “biological dog” as its source domain and models canine form and behavior. The analysis further confirms that, depending on use contexts, robot dogs may incorporate additional elements absent from the source domain, such as object grasping or high-temperature tolerance. Moreover, a comparison between Unitree robot dogs and Sony robot dogs shows that, even when the referenced source domain is identical and the intended use is similar, different elaboration processes may emerge depending on producers’ design intentions. Such differences are reflected not only in the robots’ appearance design but also in producers’ accompanying texts.
Fourth, robot discourse is analyzed by distinguishing between producers’ advertisements and consumers’ product reviews. The results indicate that the selection of a source domain and the ensuing elaboration process are evident not only in robots’ appearances but also in producers’ advertising texts. While producer and consumer discourses display convergent tendencies for educational humanoid robots, robot dogs exhibit discrepancies between the two. This is attributable to the fact that producers may preconfigure prior discourse through advertising and consumers may further specify it through use, yet context may also give rise to unanticipated relationships between users and robots.