 Hello everyone. My name is Yulia Vertanita. I'm a PhD student at the Department of Cognitive Science and Artificial Intelligence at Stillbrook University. I work on a project where the girl is to grade a virtual human. And today in my presentation I will explain what I mean by virtual human and also talk about some of my research while I studied how we real humans perceive the virtual ones. So to get the concept of virtual human in place let's start with something that is much more familiar, robots. Most of you would have seen a robot either in real life or in a movie. Many of you I believe are becoming more and more accustomed talking to your devices. Perhaps not yet having quite meaningful conversations as this guy in movie heard but this may welcome in the future. But today I won't talk about disembodied artificial intelligence. My talk is about virtual humans. And actually there are many names to call this technology such as embodied conversational agents, virtual agents, intelligent virtual agents. I like calling them virtual humans. These are software entities more or less intelligent and more or less autonomous. And they are meant to have conversations with humans via natural language and also using nonverbal means to communicate such as facial expressions, gestures, gaze. And quite importantly they have an embodiment. Some 15-20 years ago this is how they looked like pretty much cartoon characters. These days they are starting to resemble more and more actual humans due to advances in technology. But then the question is if those faces of virtual humans looked human like to you, why do we want to have a human like face for an entity that is definitely not a human? And actually the answer comes from some of the research in robotics. So for example with robots we saw that when we have imperfect human likeness people get negative reactions. So their reactions to a face that is imperfect almost like a human is negative. We get this airy sensation. So this is one of the reasons, the reason why we would like to have human like face for virtual humans. And in my research I set out to answer two questions. So looking at these faces of virtual humans created with the advanced technologies does that mean are we already not able to distinguish these faces from the actual human faces? That was one of the questions that I was interested in. And also what in the face makes the face appear human like? Where does human likeness reside in the face? So I did some experiments. In one of the experiments I had a pool of different images, human faces, virtual human faces and also objects that looked like the face, resembled faces. And I gave participants to rate these images on human likeness. What I found was that actually the virtual human faces they were not perceived as human like as the actual faces, even though they looked pretty decent. And then I wondered exactly what is responsible for this perception of human likeness. So we did some image processing and what we found in that analysis we found two features to be influential for the perception of human likeness. One of them was the skin texture. In humans, skin texture is not uniform. We have small wrinkles, lines, imperfection, pores. No one knows how many holes you have in your skin, but you're seriously perforated. This was in stark contrast to the skin texture of virtual humans that was smooth. Another feature that we identified was the eyes and specifically these highlights in the eyes, corneal reflections as they are called. So we found that these reflections that real humans get when there are light sources in the environment and the surface of the eye being smooth, you get these reflections. So those were less pronounced in the virtual human faces compared to actual human faces. So these two features seemed to be very important for perceiving human likeness in the face and actually I put them to the test. What I had, I had altered images of real humans. So again, I had a pool of different images of different individuals and I systematically changed two features. I altered the eyes removing the reflections of the eyes and I altered the skin making the skin texture appear smooth. And I also applied both changes to create the face without the reflections in the eyes and a smooth skin texture. And what I found was that these features were highly important for perceiving the face as human-like or not. And especially the eyes were crucial for perceiving life in the face for perceiving human likeness. So then now you may say to me, well, great, we know that the eyes and the skin texture are very important for perceiving human likeness, but so what? Why does this matter at all? So like I mentioned at the beginning, there are cases that actually the entity that is not alive, for its perception, human likeness is important. So human likeness is important as a first cue to also infer that this entity might have a mind behind that face. And this is especially crucial in several applications or in several domains. So if we wanted to apply our virtual human to healthcare, where we know that building trust is very important, we would like to have this virtual human looking human like, which would then promote seamless communication between a real human and a virtual human. And like healthcare, other domains would be tutoring or sales. So there are many different areas for which this perception of human likeness is quite at the core of communication. So today I've learned what features engender this perception of human likeness and why this may be important. So thank you for your attention.