Week 10 summaries
The Uncanny Valley: Effect of Realism on the Impression of Artificial Human Faces
-Jun’ichiro Seyama and Ruth S. Nagayama
This paper is an analysis and a critique of Mori’s prediction about the uncanny valley. The uncanny valley is a phenomenon stating that pleasantness or believability of artificial objects drops suddenly as these objects become more realistic but are not perfect, yet. The paper describes four experiments conducted over the web that used two sets of images- one for artificial faces and another for real human faces. Each experiment used a different morphing criteria between these faces to generate intermediate faces that were somewhere between artificial and real human faces. The initial results did not showcase the existence of the uncanny valley. However when only the heads or only the eyes of the faces were morphed the uncanny valley emerged in the graphs plotted from the user’s response. The researchers concluded that although the uncanny valley exists it only appears when there is a misplaced feature in the realistic faces like unnatural eye sizes. They also believe that humans have developed sophisticated algorithms to identify and distinguish artificial faces from real ones. We are highly tolerant to different degrees of unrealism in artificial faces but as they start appearing like real humans the smallest lack of detail is detected by our brains and is very off putting.
I do agree with most of the conclusions and theories presented by the author but I am not very comfortable with the means. The images shown in the paper show a very primitive form of morphing mechanism akin to alpha blending. Also many of the artificial and real image pairs used for morphing have no relation with respect to each other like dolls being morphed to a girl. I get the feeling the setup of the experiment may have negatively affected the results obtained during this study. Question for the authors: Shouldn’t the artificial images and their morphed targets be related?
Variations in Physiological Responses of Participants during Different Stages of an Immersive Virtual Environment Experiment
-Andrea Brogni, Vinoba Vinayagamoorthy , Anthony Steed, Mel Slater
The study is an attempt to use physiological stress as a means to measure mental stress in VE experiments. The prevalent method for VE assessment is to fill out questionnaires at the end of VE experiments to judge what the user’s experience from the feedback was. The problem with this system is that it is not real time and very subjective. The feedback may vary from the user’s mood or ability to recall the proceedings inside the VE. Physiological measures like the heart rate can be monitored during the study and user responses are obtained to understand the stress levels. The objective of this study was to examine if there were variations in the stress level during a VE experience. They attempted to figure out if physiological responses were good tools for evaluating the user’s reactions and whether or not they had event related stress during the experience. The study concluded that the user’s stress levels were high in the introduction and training phases, dipping during the first half of the experiment and again rising in the second half. The first increase in stress is attributed to lack of familiarity with the environment while the second increase is attributed to boredom or over familiarity. The authors also found that increasing level of realism also did impact the stress levels experienced by a user.
Questionnaire based studies do raise questions about the feedback obtained from the user. Such physiological measure monitored in real time can be used to scale the user responses such that they are independent of personal measures and only evaluate the VE at hand.
Physiological Measures of Presence in Stressful Virtual Environments
-Michael Meehan Brent Insko Mary Whitton Frederick P. Brooks, Jr.
This paper describes a study conducted to test the effectiveness of physiological measures as a reliable, valid, sensitive, and objective measure for the evaluation of VE systems. The authors use sensor data such as heart rate, skin conductance and skin temperature to measure the level of excitement of a user inside the VE system. A faster heartbeat, sweaty palms and drop in skin temperature are physical indicators of stress in humans. The measurements are compared for every user with their own base measurements in a training room to obtain consistent results independent of the user. The results show that heart rate is a good measure of stress followed closely by skin conductance due to sweat. The authors also tested the hypothesis that presence increases with the frame rate and found data supporting this claim.
The use of physical measures seems like a better measure of presence in VE systems. With the emergent technologies in the future it may become feasible to obtain direct brain scans during the experiment and predict the level of engagement from those. I would like to know the authors thought about this idea.
Comments are closed.