In the past few years, many projects have begun to use rehabilitation robots to solve a series of social problems in family management, rehabilitation monitoring, communication and social interaction [1,2,3]. In the future, more and more robots will be used to support people's more intelligent life at home. One of the main differences between robots and other household machines or objects is the social interaction between users and the robot. It is necessary to know how to build a close relationship between robots and humans, and how to cooperate actively with the robots. Generally speaking, the current social robot research is mainly focused on expanding the use cases of human robot interaction technology in related fields, such as eye tracking, predictive learning, and human-machine relationship [4,5,6]. In these studies of robots, robot appearance design is the basis of human-computer interaction, so it has received great attention.
In the product design, the emotional design of the robot shape can make a good user experience, especially in the era of man-machine harmonious symbiosis, the robot with emotional characteristics has increasingly become the urgent need of human life. As an agent, appearance is the physical carrier of all other aspects and has great significance [7,8,9]. It affects people’s first impression and expectation of the robot. There are many questions that can be studied around robot appearance, and one of the issues that are often debated is whether and to what extent smart products should look like people. Several products on the market are designed with consideration for the similarity between products and human body forms, such as bottle, chair, and desk lamp. Anthropomorphism is a method of assigning human characteristics to objects in order to help people understand and interpret their behaviors and abilities [10, 11]. Therefore, it has attracted extensive attention in fields such as information science, cognitive science, art design science, and psychology that robots express emotions through shapes.
In terms of the influence of robot appearance design on human-computer interaction, some people think that for humans and robots to have credible and meaningful interactions, robots should be similar to human beings in structure and function [12,13,14]. The tendency to use humanoid features as attributes is considered to be a useful tool to enable people to interact with robots. So that researchers begin to pay attention to the classification of robot appearance. In recent years, many types of humanoid robots developed around the world can be divided into two types: those with extremely human-like appearances are generally called “figurative type”, while others only have the vague features of the human form are called “abstract type” [15]. In the current research, scholars have studied the influence of robot appearance and human similarity on human emotion. There is evidence that users want robots to be “intelligent machines” that can show more humanoid communication skills, but not like human beings in appearance [16,17,18]. Through these studies, it is found that the robot head shape is more likely to stimulate good impression if it is close to the baby’s head shape. However, robots that look like but not human can seem strange and weird. This phenomenon is often explained in the context of the uncanny valley hypothesis. This is a theoretical hypothesis that explains the human emotional response to robot design. It is believed that if a robot looks like a human, but not enough to evoke a sense of familiarity, it may look like a terrible creature. These findings have a considerable impact on the field of robot design because they mean that robots close to humans may cause unwelcome, negative, and emotional reactions. Therefore, robot design is not only a design science, but also a cognitive science.
Affective processing in the human brain is receiving more and more attention. People focus on robots that look more attractive, and the first impression of these robots is stable for a long time, which also means that humanoid robots with excellent appearance design will inspire positive emotions in users [19, 20]. And these positive emotions are controlled by the left frontal cortex, while negative emotions depend on the full function of the right frontal cortex [21, 22].
More and more, research is concerned with the design implications for the desired end-user groups, whether children, adults, or the elderly [23, 24], looking at the appearance design of robots from their point of view. In addition, there was a difference between young and older respondents: the older the elderly, the less they liked the appearance of a human being [25].
However, at present, the research on robot appearance preference is mainly based on questionnaires measurements, interviews, behavioral experiments, and few studies use advanced neuroscience measurement instruments to enter the level of cognitive neuroscience [26, 27]. In comparison, the cost of subjective measurement is relatively low, but it does not go deep into the human brain to accurately grasp the mechanism of human response to the appearance of the robot. Studies have shown that in many cases, questionnaire responses are affected by external factors and may be changed intentionally by the interviewees, rather than the initial direct brain response [28]. In the past few years, researchers have also been interested in using brain-imaging tools to provide objective data to determine the possible relationship between robot appearance and preferences, rather than simply asking people what they like. In this way, cognitive neuroscience goes deep into the interior of the human brain, which can explore the neural response process and cognitive mechanism of people to things, and then make the research go deep into a more scientific and accurate basic level. In this respect, electroencephalogram is a fairly objective approach to know people’s perception process and processing mechanism of robot appearance from the brain cognitive level [29]. By using electroencephalogram, it is possible to understand their emotional state when they are looking at different robot designs.
Event-related potentials are a good method for facial perception analysis. In the field of design, there has been some work that uses these methods to study aesthetic preferences. The brain response related to preference and rank of faces is that the emotional judgment process precedes the cognitive process [30]. In other words, before we make a conscious choice about a robot, we have unwittingly given the answer in advance. Event-related potentials can assess the neural response to emotional events at millisecond time resolution, and we know that decisions to like/dislike are made milliseconds before we really realize it [31].
In summary, the existing research still lacks enough understanding of people’s preference for the appearance of robots. Most of these studies use questionnaires with closed questions to investigate this question, which may bring some bias to the experimenter. Considering these questions, we decided not to find the answer directly through specific questions (for example, what kind of robot do you like). Instead, we try to explore some of the aesthetic criteria of robots by collecting people’s electroencephalogram responses to different types of robots.