[Summaries Week 10] Uncanny Valley, Variations in Physiological Responses, Physiological Measures of Presence
The Uncanny Valley: Effect of Realism on the Impression of Artificial Human Faces
In this paper, the authors attempt to empirically establish Mori’s conjecture of the existence of an “uncanny valley” – an area in the realism-appeal space where there is a paradoxical reduction in the appeal of an artificial entity with increasing realism. Mori’s 1970 conjecture was that the uncanny valley exists when there is a high degree of realism in the artificial entity, making it appear almost human-like.
The authors try to validate this conjecture by carrying out various experiments that vary the realism of an image of a face by morphing between a starting image and an image of the human face, and gauge participant’s response by asking them to rate the appearance of generated images.
The 1st experiment involves the authors morphing an artificial face, in its entirety to a human face. This however, doesn’t lead to a dip in pleasantness with realism. The negative results of experiment 1 cause the authors to concentrate on morphing local features. In their second experiment, the authors first morph the eyes from artificial to real, following by the rest of the head undergoing a similar transformation, and carry out another set of morphs with the order reversed (first the head is morphed and then the eyes). The authors discover a sharp low peak in the realism-appeal graph, when one of the feature was human while the other was not, thus causing them to surmise that uncanny valley occurs when there exist abnormalities that break the realism.
The authors also perform another experiment where they try to determine if abnormalities are a sufficient condition for the uncanny valley, irrespective of the realism. They found that an abnormality in an artificial face didn’t elicit the strong decrease in appeal as it did for real faces. Finally, the authors perform another experiment that seeks to reduce the effect of co-founding factors such as age, etc. to conclusively establish the link between realism and the uncanny valley. They obtain encouraging results her as well.
Thus, the authors establish that the perceptual sensitivity to facial features is higher for real faces, and that abnormalities in such features produce the uncanny valley for faces with high degree of realism.
****
Variations in Physiological Responses of Participants During Different Stages of an Immersive Virtual Environment Experiment
In this paper, the authors study the variations in physiological responses of participants as they are exposed to a VE. The novelty is that the authors seek to study these responses at a much finer granularity than in contemporary research, which places focus on identifying high variations in the physiological responses and correlating them with “shocking” events in the VE.
In keeping with this goal, the VE was a simple street occupied by virtual pedestrians, with behavioral characteristics akin to those of humans, and the participants were place in a CAVE like environment. The authors varied two main parameters in the VE – the texture quality of the environment and the visual aspect of the characters, and kept track of the heart rate and heart rate variability(HRV) of the participants. As part of the experiment, the authors asked the participants to signal breaks in presence (BIP) by the use of a joystick type device.
The results of the four minute long experiment revealed that there was an increase in stress at the start of the VE exposure, during the training phase. The authors attribute this initial stress increase to an increase in the physical activity. This was followed by a dip in stress, and then an increase as the experiment drew to a close. The authors conjecture that the dip is due to the habituation of the participant to the system.
****
Physiological Measures of Presence in Stressful Virtual Environments
Through this paper, the authors vouch for the use of physiological reaction measures as reliable, valid, sensitive and objective presence measures for a virtual environment. They substantiate their claims by performing 3 different experiments, using the Heart Rate, Skin Conductance and Skin Temperature as physiological metrics.
The authors perform their experiments in a stressful virtual environment, to elicit significant physiological response. The environment has a simulated pit, 20 feet deep. The authors first describe how they use questionnaires on Reported Presence and Reported Behavioral Presence to obtain ratings in the scale of 1-7, which is used as the ground truth. They then describe their 3 experiments, aimed at exploring the effects of: multiple exposures, passive haptics and frame-rate on presence.
The “multi-exposure experiment” required users to experience the VE multiple times, and was aimed at determining if the presence evoked in a VE decreases with multiple exposures. The “haptics experiment” had users experience the VE with/without a wooden-ledge that provides haptic feedback akin to being on the edge of a pit. The third experiment involved the variation of frame-rate for each VE session for the participants, who were also asked to perform basic interactions in the VE.
The authors subsequently present their result, stating that they found significant physiological reactions to exposure to the pit room, and that these reactions decreased in intensity over multiple exposures, consistent with prior studies that demonstrate the effect of habituation to stress reactions, thus establishing these metrics to be reliable. The authors measure the validity of the physiological measurements by seeing how closely they correlate to the questionnaire results, and report that heart rate had a high correlation with the ground truth.
The authors also found heart-rate to be a sensitive measure of the presence in the pit VE, with metrics like the heart-rate jumping when the participants tumbled near the pit. They also find the heart rate metric to be muti-level sensitive, with its sensitivity to stimuli depending of the sense of presence evoked by the VE as a whole. Finally, the authors make a case for the objectivity of physiological measurement, stating that these measures are inherently better shielded from bias. Thus, the authors successfully demonstrate the viability of using physiological metrics as measures of presence.
Comments are closed.