<p>With the increasing prevalence of social robots in everyday life, it is important to design robots that are perceived as trustworthy and approachable to facilitate human engagement. Previous research has indicated that several aspects of human-robot interaction function similarly to human-human interaction. One aspect of social interaction includes social perception, which involves forming social inferences (i.e., judged qualities about a person) while evaluating the appearance of another. There are numerous social inferences we make when looking at human faces, which inform decisions about whom we choose to interact with, and how we approach such interactions. However, it is unclear how well these inferences transfer to facial features of social robots, and consequently how they might be useful toward facilitating human-robot interaction. To test this, we conducted an experiment to investigate the effects of different facial features of a social robot (i.e., projected skin texture and skin color-heterogeneity) on the perceptions of trustworthiness, realism, and preferred appearance. Results suggest that people generally prefer more homogeneously distributed skin color and moderate skin texture. However, there are different trade-offs between perceived trustworthiness, realism, and preference, depending on the manipulation of the facial features. The current work highlights the importance of visual perception and social evaluation in the design of social robot appearance.</p>