Determining the Characteristics of Preferred Virtual Faces Using an Avatar Generator Valentin Schwind ab , Katrin Wolf b ,Niels Henze b , Oliver Korn b a Stuttgart Media University Stuttgart, Germany schwindv@hdm-stuttgart.de b VIS, University of Stuttgart Stuttgart, Germany {firstname.lastname}@vis.uni-stuttgart.de ABSTRACT Video game developers continuously increase the degree of details and realism in games to create more human-like char- acters. But increasing the human-likeness becomes a problem in regard to the Uncanny Valley phenomenon that predicts negative feelings of people towards artificial entities. We de- veloped an avatar creation system to examine preferences to- wards parametrized faces and explore in regard to the Un- canny Valley phenomenon how people design faces that they like or reject. Based on the 3D model of the Caucasian av- erage face, 420 participants generate 1341 faces of positively and negatively associated concepts of both gender. The re- sults show that some characteristics associated with the Un- canny Valley are used to create villains or repulsive faces. Heroic faces get attractive features but are rarely and little stylized. A voluntarily designed face is very similar to the heroine. This indicates that there is a tendency of users to design feminine and attractive but still credible faces. Author Keywords Virtual Faces; Avatar Generation; Player Preferences; Uncanny Valley; Video Games ACM Classification Keywords J.4: Computer Application: Social and Behavioral Science; K.8.0: General: Games INTRODUCTION Success of video games often relies on the appearance and credibility of their virtual characters. As Rollings and Adams write, ”a player will not play a game if its character does not interest the player or is not believable” [25]. This might be correct for most of today’s online and role-playing video games (RPGs), which depend on continuously technical im- provements in realistic computer graphics. Video games de- velopers continuously increase the degree of details to create more credible human-like characters and to improve immer- sion in virtual worlds. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Permissions@acm.org. CHI PLAY 2015, October 03 - 07, 2015, London, United Kingdom Copyright is held by the owner/author(s). Publication rights licensed to ACM. ACM 978-1-4503-3466-2/15/10$15.00 DOI: http://dx.doi.org/10.1145/2793107.2793116 Figure 1. Random selection of 15 user generated faces. But increasing the human-likeness of virtual characters be- comes a problem in regard to a counter-intuitive phe- nomenon, called the Uncanny Valley [22]. This effect is al- ready known in robotics or computer animated movies and forecasts a rejection towards too human-like entities. The Uncanny Valley hypothesis suggested by the roboticist Masahiro Mori in 1970 [22] predicts negative feelings to- wards figures or prosthetics that are not quite human-like. The term emerges from the graph in Figure 2 which illustrates the relationship between familiarity and human-like appear- ance. The more human-like characteristics a figure has, the more likely it will be accepted. However, at a certain point the similarity to humans causes a reverse effect. The affin- ity rapidly changes into repulsion. The figure appears in a negative way to its human observer and falls into the valley, and only an indistinguishable real human is fully accepted by observers again. Prediction errors increase negative feelings [21] and lead to a concept that does not match with the visual sensation. Neu- roimaging confirms this effect and show brain activity in re- gions, which has been associated with violations of predic- tion [26, 31]. That mismatch can lead to an interrupt in feel- ing empathy [20], which also means a loss of immersion in games and identification with the main character [3]. Previ- ous research confirmed, that less human-like designs trigger less negative feelings [8]. However, the effect was primarily examined using photos, videos of robots or computer animations. If people actu- ally prefer credible and human-like characters, the question arises which appearance and amount of details would they choose, when given a free design choice. Due to negative 221