Uncanny Valley, Physiological Responses to VEs, and Measuring Presence
Meehan. Physiological Measures of Presence in Stressful Virtual Environments
Given that presence is a measure of the quality or effectiveness of a VE, the paper gives a physiological measure of presence to satisfy the general requirements of any useful measure: to be reliable, valid, sensitive, and objective.
The physiological response in someone in a VE should be like or similar to the response evoked in a similar real environment, to the extent that the VE provides the sensations of presence.
The paper describes three experiments in which heart rate, skin conductance, and skin temperature are measured for different VEs: stressful and non stressful situations. Variation in frame rate is then demonstrated to have a positive effect on physiological response: increased frame rate yields increased response.
Since presence is a subjective response, it is difficult to quantify the sensation. The degree to which someone feels that the VE is present is hard to measure because the feeling is transient, deeply tied to the current sense, and breaks in presence (BIPs) can interrupt the sensation. Asking the person in the VE how they feel causes a disruptive BIP, spoiling the validity of the measurement. To the extent that these physiological measures access a side-channel of human emotional response, they satisfy the need for a constant and non disruptive quantitative measure of human emotional state.
Since we have very clear ideas about frame rate and the smoothness of a view being a strong contributor to the sense of the believability of a VE, showing a positive main effect of frame rate on heart rate doesn’t disprove that presence could be measured by physiological means, but it does strengthen the hypothesis that presence can still be a measurable quality teased out by watching the subtle responses of the body. Including a haptic element in the experiment shows a main effect as well, suggesting that whole-body approaches to immersion that include both visual and haptic interactions increase presence by this measure.
This all reminds me of Phillip Dick’s fictional Voigt-Kampff test in “Do Androids Dream of Electric Sheep”; in which the android robots’ identities are betrayed to human investigators by the distinct lack of a proper physiological response when placed in an imaginary scenario meant to evoke emotional reaction.
This paper is significant by suggesting that validation of post hoc analysis is possible by fusing physiological measures with survey questionnaires and interview methods. The overall problem of sensing is still difficult: even this method forces a potentially problematic issue forward- how can sensing physiological response be done in a way that doesn’t contribute more risks of BIPs. Wearing of sensors may help validate the VE, but there is an interaction effect that may be caused by increased stress of wearing the sensor arrays. Users reported being able to ignore the sensors, but how can we trust that they can report some autosomal response reliably?
I was convinced by the paper that these physiological measures meet the requirements of being valid, reliable, multi-level sensitive, and objective and expect that more refined measurement or discovery of other autonomic responses such as blushing, eye-movements, rate of movement/confident movement, or breathing rate could be good measures as well if sensitive enough readings can be made.
Brogni. Variations in Physiological Responses of ParticipantsDuring Different Stages of an Immersive Virtual Environment Experiment
This paper builds on the findings of the Meehan paper by increasing the sensitivity of time-based fluctuations in heart rate. By devising a more thorough approach to measuring and interpreting heart rate, both the high and low frequency characteristics, the researchers confirmed the value of heart rate response as a physiological measure of presence, even in low-stress VEs. While the results confirm the efficacy of HR as a measure, and recover the advantages of subtler understanding of heart beat pattern traits, they introduce some problems with the measurement technique. Namely, it exposes more of the finer grain difficulties of creating convincing VEs. In addition, the experiment design: limitations in time, sample size, and overall experience all contributed to mixed results for HR as a presence measure. It seems that a VE can induce stress in a person, but that this stress may not be related to the absolute quality of presence. Physiological response is also affected significantly by confounding factors in the complete experience of research subjects: the enjoyment of the task, the stress of wondering what the point of the VE actually is, the psychological distance between the VE being a believable environment and the interaction with that environment being perceived as useful or worthwhile.
Seyama. The Uncanny Valley: Effect of Realism on the Impression of Artificial Human Faces
In an attempt to validate the hypothesized Uncanny Valley (UV), researchers devised a web-based experiment to elicit emotionally reasoned responses to the morphed visages of faces across a controlled scale of variation between fully real human and human-like doll faces. The found that the UV feature did emerge, but only when the human-like face had non subtle abnormal features. The paper avoids issues like behavior, movements, and many other exemplary traits of human-ness, but it does thoroughly and scientifically tease out a functional relationship between visage human-ness and comfort level.
While the original Mori hypothesis figured that the uncanny valley was a prominent feature in the scale of human-like appearance from industrial robot (very non-human) to humanoid robot imperceptible from real human being, this set of experiments focused on recognizable visages only in order to more carefully target the proposed UV feature in terms of facial features alone. The experiments in the study were carefully crafted to control for perceptions in sex, mood, expression, and feature abnormality (eye size).
In the paper, the first interesting result is “Participants gave the lowest pleasantness score when the eyes were 100% real human and the head was 0% real human… [and] when the head was 100% real human and the eyes were 0% real human” This finding seems to get at the nature of our level of comfort and why close but not close enough is a significant issue. The finding suggests that completeness of the experience is more important to acceptance rather than overall human-ness or artificial-ness. The “mismatched realism may be a necessary condition for the uncanny valley’s emergence.”
Then, as the experimental process develops further, the researchers reveal that “abnormally scaled eyes decreased pleasantness scores the most when the faces were the most realistic.” Therefore, the mismatched realism may be a necessary condition, but the exact part to whole relationship of real to unreal elements is critical for eye size.
Comments are closed.