This project is the thesis work for a master degree in Digital and Interaction Design at Politecnico di Milano. The research and the results are published for "International Design Engineering Technical Conference & Computers and Information in Engineering" Conference.
Social phobias require the presence of other people or avatars in virtual environment. This raises the question about the uncanny valley effect. This is a hypothesis, confirmed via a number of experiments, that perception of a robot is positive until the point it starts to look too humanlike. After a certain point, the emotions become negative, causing the valley. It is calculated that the effect is observed at 70-85% human likeness. Even though the hypothesis is originally about robots, the effect is believed to be even stronger in Virtual Reality and avatars.
With the understanding of the uncanny valley effect and the causes of social phobia, a design problem comes up. All solutions on the market for VR-based therapy of social phobia are developed with high realism. The research on the uncanny valley effect so far has not considered the individual characteristics of people. This leads to our research question – How the uncanny valley effect is influenced by the presence of social phobia?
A prototype is developed to test the hypothesis. Four key points on the realism axis are selected. The first point is low level of realism at 10%, second is medium with around 50%, and third is high realism around 70%. The fourth one is considered the full realism. The selected social phobia is fear of public speaking or also referred to as glossophobia.
For each level of realism three versions are made for positive, negative and neutral audience reactions. In total 9 versions.
The audience is made up by 8 characters - 4 male and four female.
All avatars have body movements. This includes blinking and breathing as well as shaking head in agreement or disagreement. Some of them are looking at the speaker.
They have facial expressions to convey the audience reactions.
At the end of the speech, the audience claps. The enthusiasm in the sound and clap represent the audience reaction as well – for positive, the clap is cheering while for negative is very short.
HIGH LEVEL OF REALISM
MEDIUM LEVEL OF REALISM
LOW LEVEL OF REALISM
SURVEY WEBSITE
The survey was conducted online with no requirements for participation. Both people with and without phobia are needed in order to create control group for comparison. The survey was conducted in a website, developed for this purpose due to the dynamic prototype assignment.
Step 1.
The first step includes generic questions like gender, nationality and age.
Step 2.
The next step is the questionnaire for evaluating the phobia severity. A standard one for self-evaluation is used, called Personal Report for Public Speaking Anxiety. It includes direct questions towards experiencing some of the symptoms like sweating, increase heart rate and shaky hands. The answers are in 5-step scale from Strongly Disagree to Strongly Agree.
Once this step is concluded, calculation whether the participant has phobia or not is performed. Based on it, the prototype version is assigned, targeting balance between all versions.
Step 3.
In Step 3, the participant is asked to give a 5-minute speech in a virtual environment. The prototype is provided as a VR video in Youtube to avoid additional installations. It can be watch on a screen or with a basic headset like Google cardboard.
Step 4.
Next is evaluation of the virtual speech by using questions from PRPSA for the particular experience.
PARTICIPANTS
CONCLUSION
We can conclude that the perception of avatars in virtual environment is individual and depends on the user’s individual characteristics, in particular whether they experience a social phobia.
This research suggests that the level of realism in Virtual Reality is a variable, which can be manipulated by a designer in attempt to meet a certain design brief.
The conducted experiment is at a proof of concept stage, suggesting that the hypothesis has ground. The planned next steps are to conduct an experiment in lab conditions, following the requirements for clinical psychology experiment. Physiological sensors will be used instead of self-evaluation to retrieve more precisely the levels of anxiety.
We already know that VRET is effective at high level of realism. From this research we suggest that at the same realism, the emotions are positive in people with phobia. Therefore, future experimentation could be to explore the relation between effectiveness of the therapy and the positive emotions by comparing the patient’s improvement in the other levels of realism.
