Using Expressive and Talkative Virtual Characters in Social Anxiety

Disorder Treatment

Ana Paula Cláudio

, Maria Beatriz Carmo

, Augusta Gaspar

2,3

and Renato Teixeira

LabMAg, Faculdade de Ciências, Universidade de Lisboa, 1749-016 Lisboa, Portugal

CIS- Centro de Investigação e Intervenção Social, ISCTE Instituto Universitário de Lisboa, 1649-026 Lisboa, Portugal

Grupo de Psicologia, Faculdade de Ciências Humanas, Universidade Católica Portuguesa, 1649-023 Lisboa, Portugal

Keywords: Virtual Reality, Virtual Reality in Exposure Therapy (VRET), Social Anxiety, Fearing of Public Speaking,

User Study, Nonverbal Behaviour Research.

Abstract: Social Anxiety affects a significant number of people, limiting their personal and social life. We describe an

interactive Virtual Reality approach to the exposure therapy for social anxiety, resorting to virtual characters

that exhibit combinations of facial and body expressions controllable in real-time by the therapist. The

application described in this paper updates and significantly improves a former version: ameliorating the

graphical quality of the virtual characters and providing them with the ability of articulating a set of sentences.

The application executes in ordinary computers and it is easily used in counselling and research contexts.

Although we have only resorted to free or very low cost 3D models of virtual humans, we adopted strategies

to obtain an adequate final quality that we were able to validate with a significant number of observers.

Moreover, a set of therapists tested the application and gave positive feedback about its potential effectiveness

in Virtual Reality Exposure Therapy.

1 INTRODUCTION

Social phobia or Social Anxiety Disorder (SAD) is a

human condition characterized by intense anxiety

when the individual faces or anticipates public

performance (APA, 2000). This condition can be very

crippling in the personal, social and professional

domains, as those bearing it withdraw from social

contact; it also has a high comorbidity with

depression (Stein, 2000). People with SAD fear

negative social judgments and are hypervigilant for

signals in other’s behavior, thereby identifying faster

and more efficiently than other people facial clues to

threatening or negative content (Douilliez, 2012).

Therapy approaches to SAD include medication

relaxation methods, and psychotherapy, mainly

Cognitive-Behavioral Therapy (CBT). CBT produces

the most efficient and persistent improvements,

especially when it is applied as Exposure Therapy

(ET) (Beidel, 2007) which consists in exposing the

patient to the feared situation.

Virtual Reality (VR) has been used in ET since the

early 90’s, being called Virtual Reality applied to

Exposure Therapy (VRET).

Several studies have concluded that VRET

produces results that are similar to traditional

exposure therapy (Klinger, 2004; Herbelin, 2005).

VRET allows a precise control over the habituation

(and extinction) to the fear of the phobic object and

offers thus, several additional advantages over classic

ET (which is based on images and later contact with

in vivo situations).

Comparing with traditional ET, VRET presents

some important advantages: i) it allows scenario

configuration and interactions in order to fulfill each

patient’s needs and progress levels along the therapy;

ii) it provides better preparation of the patient before

facing a real life scenario, avoiding the risk of a

premature exposure to a real situation; iii) it reduces

the risk of taking steps backwards because of

overreactions, allowing a more stable and progressive

environment towards predictable and solid results; iv)

it assures patient privacy.

The downside of VRET is the cost of the

immersive virtual equipment (e.g., Head-Mounted

Displays, CAVE) and sometimes the secondary

effects reported by a few users (cybersickness)

(Laviola, 2000).

This paper describes a VR approach to the

treatment of SAD, specifically to the fear of public

speaking - before a jury in an evaluation context, in a

job interview or in some other similar scenario. The

348

Paula Cláudio A., Beatriz Carmo M., Gaspar A. and Teixeira R..

Using Expressive and Talkative Virtual Characters in Social Anxiety Disorder Treatment.

DOI: 10.5220/0005312203480355

In Proceedings of the 10th International Conference on Computer Graphics Theory and Applications (GRAPP-2015), pages 348-355

ISBN: 978-989-758-087-1

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

main idea is that in a session therapy, while the patient

faces a jury of one-to-three virtual characters that

show facial and body expressions, the therapist

controls these characters according to the level of

stress he/she wants to induce in the patient. This

control is accomplished through an interface (visible

only to the therapist) which, among other options,

triggers specific facial and body movements that can

be combined to convey neutral, positive or negative

emotional content, and simulate various degrees of

attention or lack of interest.

Before a therapy session, the therapist configures

the scenario by choosing the characters and their look

(hairstyle, clothes and glasses), their position at the

table and also the appearance of the room (classical

or modern furniture, different wall colors). The

simulation scenario must be projected on a canvas or

a wall, in such way that characters are displayed in

real size, enhancing the immersive effect.

Our application is also a useful tool to the

investigation of non-verbal behavior: the

communication effect related to facial or body actions

is not entirely known, and it is currently object of

intense academic debate (Gaspar 2014; Russell,

1997). The application allows the individual

visualization of a virtual human’s face (henceforth

VH) and supports many combinations of body and

facial behavioral units. It also makes possible a fine

tuning control over a single individual facial action,

whilst others are kept constant. Evaluating the impact

of these combinations in observers may clarify the

communicative role of single and composed actions,

giving clues about the treatment of social anxiety. It

may also assist various other research lines in the field

of non-verbal behavior.

The work described in this paper follows previous

work from the same team. A first version of the

application recreates an auditorium filled with virtual

characters with controllable behaviors but with few

facial expressions (Cláudio, 2013); a second version

considers a simulation of a jury composed by a set of

virtual characters with body and facial expressions

controllable by the therapist in real time (Cláudio,

2014). The development of each version has been

closely followed by a psychologist who has played a

crucial role in the identification of facial and body

expressions that potentially convey to the observer a

positive, a neutral or a negative feedback.

This paper addresses the implements included in

the present version, thus focusing the quality of the

virtual models; we report results from a validation

study concerning the VH’s facial expressions, and the

psychologist’s assumptions underpinning the creation

of the facial actions’ menu. Usability tests were

performed with experts (therapists) in order to study

the suitability of the application and its potential

effectiveness to VRET.

The document is organized as follows: next

section presents some of the most relevant work in the

area; section 3 describes our approach and the

implemented application; section 4 reports the VH’s

facial expression validation process and the

evaluation of the application by therapists; finally,

section 5 draws conclusions and lines for future work.

2 RELATED WORK

The first VR application to treat public fear before an

audience was presented in (North, 1998). This

application included a scenario with up to 100

characters. During a therapy session, the therapist was

able to vary the number of characters and their

attitudes, using pre-recorded video sequences. Patient

used an HMD, listening to the echo of his own voice.

Slater et al. created a virtual room with 8

characters with random autonomous behaviors, such

as swinging the head and eye blinking (Slater, 1999).

The initial study gathered 10 students, with different

levels of difficulty about public speaking, and was

extended later to include phobic and non-phobic

individuals (Pertaub, 2001; Pertaub, 2002; Slater,

2006).

James et al. proposed a double scenario: one

subway wagon populated by characters that express

neutral behaviors, which is considered a non-

demanding situation from a social interaction point of

view; and a more demanding situation that took place

in a bar with characters that look uninterested (James,

2003). The characters’ behavior included eye gazing

and pre-recorded sentences.

Klinger et al. conducted a study with 36

participants to evaluate changes in fear before public

speaking during 12 sessions (Klinger, 2004). To

recreate virtual characters, they used real people’s

photos in typical situations. Participants were divided

in 2 groups: one treated with CBT and the other with

VRET. A higher reduction in social anxiety was

reported in the VRET group.

Herbelin published a 200 patient validation test,

demonstrating that his platform fulfilled the

requirements of therapeutic exposure in social phobia

(Herbelin, 2005) and that clinic evaluation can be

improved with integrated monitoring tools, such as

eye-tracking.

All the referred approaches resorted to HMD

equipment; in a study described by Pertaub et al. half

of patients tried one of the virtual environments

UsingExpressiveandTalkativeVirtualCharactersinSocialAnxietyDisorderTreatment

349

through a HMD, while the rest of the group used a

desktop (Pertaub, 2002). Herbelin and Grillon, in

addition to an HMD and a computer screen also used

a big projection surface (Herbelin, 2005; Grillon,

2009).

Haworth et al. implemented virtual scenarios to be

visualized simultaneously by patient and therapist in

computer screens, possibly in different physical

locations and over the internet (Haworth, 2012).

Scenarios are oriented to patients suffering from

acrophobia (fear of heights) or arachnophobia (fear of

spiders). A Kinect is used to control patent body

movements (url-Kinect). The few results of this study

suggest that this type of low-cost solution is effective

in these phobias.

3 VIRTUAL SPECTATORS

Virtual Spectators is our VRET approach to the fear

of public speaking. The virtual scenario is, during a

therapy session, the stage to a simulation controlled

by the therapist and observed by the patient.

The application has two types of users: the

therapist as an active user, and the patient as a passive

user. The patient, while giving a speech in front of a

set of virtual humans, receives stimulus from these

characters; the therapist, who is observing patient’s

behavior and reaction to these stimulus, interacts with

the application in order to modify the simulation

accordingly, whether it is by varying the characters’

behavior, or by triggering multiple events in the

simulation scenario. The main initial scenario of the

application is configured by the therapist as well.

The equipment required to use the application is:

a computer, two sound columns, a projector and a

canvas or wall used as a projection surface. The

application generates two separate windows: the

simulation window, that must be projected, and the

interface window which is displayed in the therapist’s

computer. The sound columns must be close to the

projection to increase the degree of realism of the

simulation.

The equipment is inexpensive and it is easy to

install; the projected image should contain the models

of the VHs in real size, making it easy for the patient

to get an immersion feeling. Moreover, it is possible

to gather several people observing simultaneously the

simulation, which can be valuable in research or, for

instance, in the training of students. Additionally, all

the unpleasant secondary effects reported by some

users when using a RV equipment are eliminated.

The version here presented follows two others:

the first version simulates an audience with virtual

humans (Cláudio, 2013) and the second one a jury

composed by one to three virtual characters (Cláudio,

2014). This paper addresses the improvements

performed in this last version.

3.1 The Present Version

The main goal of this development stage in Virtual

Spectators was to provide characters with high

realism. Such goal had to be met without

compromising application’s performance, which had

to respond in real time to the therapist’s control.

Finding a balance between the characters’ final

appearance and the most critical aspects of the

application (number of polygons of the meshes, the

textures’ resolution and the complexity of the

illumination algorithms) was crucial to have the real

time response. Early on we realized that it was not

possible to improve the characters of the previous

versions. So we adopted completely distinct models.

After a series of tests, we found that the models

generated using free tools, such as MakeHuman (url-

MakeHumans) did not satisfied the aimed quality

level. Thus we shifted our approach to combining

different models (or parts of them) from the software

Poser (url-Poser) and from online repositories. In

such way, we produced two males and one female

characters appearing different ages, and named John,

Carl and Jessi (Figure 1, 2 and 3).

Figure 1: John model exhibiting two different facial

expressions.

Next, we proceeded to the animation of these

characters considering the characteristics and

functionalities of the software tools involved in the

implementation of the application: Blender (url-

Blender) and a free version of Unity3D (url-Unity).

In Unity, animations are obtained through rigging

animation, i.e., by animating a skeleton, a hierarchical

structure of interconnected bones.

The animation also comprises the process of

skinning: association of a bone to a certain set of

neighboring vertices. When a bone moves, it drags

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

350

Figure 2: Jessi model exhibiting two different hair styles

and clothes.

Figure 3: Three virtual models exhibiting distinct facial

expressions (from left to right; John, Carl and Jessi).

the corresponding vertices with it; the vertices closest

to the bone suffer a bigger displacement than those

furthest away. In Blender, the association vertices-

bone and the weights that define the amplitude of the

translation are tuned resorting to a functionality called

weight painting.

In order to recycle as much as possible the

animations used in the previous version, we

proceeded to the integration of the skeleton of the old

characters into the new ones. However, the base

position of these skeletons does not match the

position of the mesh in the new characters, which

required a proper adjustment. Such adjustment led to

the modification of the base position of the skeleton

resulting in modifications in every animation that

relies on the bones used in the animation. Thus, each

animation had to be adjusted as well. Regarding

skinning, this process is created from scratch, both for

the face as for the body. Given that the three

characters are anatomically different, this exhaustive

process was repeated for every VH character.

3.2 Interface Functionalities

The current interface presents significant

improvements over the previous one in terms of

available functions, namely:

• Possibility of choosing between different

scenarios. The application is ready to new

scenarios that might be relevant to consider in the

future.

• A drag-and-drop mechanism that allows the user

to choose the characters: the user selects the

picture that represents the character and drags it to

the corresponding position in the table contained

in the virtual scenario; the photo of the character

turns grey as a sign that it is no longer available to

be chosen.

• Choosing glasses for a character previously

selected.

• Choosing a clothing style (formal or informal).

• Choosing a type of hairstyle between formal and

informal (except for Carl that is bald). Choosing

the hairstyle is independent from choosing clothes

allowing multiple combination of options.

• Character preview: therapist has the possibility to

visualize the selected and modified character with

the previous options. This preview allows a close-

up of its face.

3.3 Facial Expressions Interface

The interface that allows the user to control facial

expressions is illustrated in Figure 4.

Figure 4: The interface to manipulate facial expressions.

It contains buttons that correspond to Action

Units (AU) from Facial Action Coding System

(FACS) (Ekman, 2002); in two AUs (AU4, AU12)

there is an option to increase intensity. Our choice of

facial element to include and the universe of

expressions one can compose with them is based on

current validated knowledge on the content of human

facial behavior (for an updated review see (Gaspar,

2014)). Although there are applications, games and

films today with a wide range of expressive behavior,

the way expressive behavior is decoded by people and

what exactly do real people convey with their facial

and body behavior is still largely debated in emotion

Psychology (see (Russell, 1997)) so we opted for a

range of AUs and possible combinations, that has

been most consensually derived from behavioral

studies of spontaneous facial behavior (Gaspar, 2012;

Gaspar, 2014).

UsingExpressiveandTalkativeVirtualCharactersinSocialAnxietyDisorderTreatment

351

Furthermore, these facial compositions can also be

associated with several other body postures. Our

selection of postures is based on cross-cultural studies

of human non-verbal communication (Eibl-

Eibesfeldt, 1989).

Figure 5 shows the same model with different AU

combinations. The neutral face (upper left corner),

with no AU activated and a negative-emotion

expression, AU4 (upper right corner). In the lower

right, we see a combination of AU4 and AU15, both

related to negative affect. In the lower left corner we

see a combination of AU4 and AU6+12 (“smile”

button). The AU6+12 combination is the known as

“Duchenne smile”, and is consensually associated to

positive affect more than any other smile form.

Figure 5: Clockwise from top: a VH with a baseline face

(no AUs), displaying a frown (AU4), displaying a

combination of AU4 and AU15, displaying a combination

of AU4 and AU6+12 (the smile button).

3.4 Speaking Characters

Since our main goal is to attain realistic human-like

characters, it was considered a very important task to

include in the characters the ability of speaking.

Verbal communication plays an important role in

human interactions, reason why this requirement is

important. However, to provide a VH with the ability

to speak is a complex matter and difficult to

implement, requiring complex algorithms as well as

complex animation process. So, we conceived a

simple but effective solution.

The idea was to define a set of speeches that each

character could reproduce (with sound synchronized

with lip movements) and that were controllable and

triggered by the therapist during the simulation. On a

first trial, we implemented this solution for a single

character, John.

The main steps involved in this process are:

1. Record the intended set of speeches that will

figure in the application. It can easily be

accomplished using the microphone of a laptop.

2. Define and shape all mouth positions involved

in the process of animation of the characters

considering the sounds that need to be reproduced.

On a first instance we defined 5 shapes/states for

the Portuguese language: Base (baseline), A, E, O

and U. This means that every sound can be

obtained using only these 5 states.

3. Each shape corresponds to an animation.

Therefore, this includes a process of animating the

characters in order to be able to visually reproduce

these different states. The only bones affected

belong to the mouth area, namely, jaw and lips.

4. Decompose the sentence given as an input.

This is the sentence that the character should be

able to reproduce. Thus, any given sentence

should correspond to a sequence of animations

that the character has to process and verbalize. For

example, the sentence “Hello world” should be

translated to a language recognizable to the

character and it would correspond to the sequence

of animations (E,O,U,O,Base). This process

involves the following steps:

a) Decomposing the sentence into smaller

segments. The division of the sentence is

somewhat similar to division by syllabus.

However, each segment must contain a single

vowel. Given the example, it would be “He-

llo-w-o-rld” (in this approach, letter w is

considered as a vowel);

b) Evaluate each segment according to the

sounds each segment demands;

c) Add the corresponding animation to the buffer

of all animations that need to be played. For

example, a segment “He” will correspond to

state E, “llo” to O, and so on.

5. Reproduce sequentially all the animations. It

is defined a time interval between each animation

so that all animations can be synchronized with

the sound that starts playing as the therapist

triggers the event.

This approach provides a simple but realistic and

attractive solution. It is a good starting point towards

a more robust and complete approach. New states can

be considered in order to include a wider set of

sounds. As of now this solution only takes into

account Portuguese language so the proper

modifications are required for other languages.

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

352

4 EVALUATION AND

VA L I D AT I O N

This sections describes two evaluation tasks that have

been performed. A validation study was made to

confirm the assumptions of the psychologist in the

team during the creation of facial actions’ menu.

Additionally, we performed usability tests with a set

of experts (therapists) to study the suitability of the

application and its potential effectiveness in VRET.

4.1 Validation Phase

Beyond the theoretical framework that presided our

choice of expressive elements to include in the VHs,

derived from human behavior, it is necessary to

validate the content in the expressive behavior

displayed by the VH’s. This step precedes writing a

manual with guidelines for therapists who wish to

design a comprehensive intervention plan with the

different levels of positive/negative affect or

intimidation that the VH´s may convey. For such

purpose we have conducted a preliminary study of the

facial behavior content with a normative sample of 38

voluntary participants (31F; 7M), recruited in two

universities (ages 18- 25 yrs. old).

Participants were tested in groups of 10 and given

instructions on their appraisal tasks. These were

presented and controlled by an E-Prime 2.1

programmed experiment. Experiments took place in

a dim light room and consisted on participants

watching and rating each of 28 animated clips (3”

duration) with a close up view of a VH showing either

a neutral or an expressive face) projected on a canvas

2-3 meters in front of them. We tested expressions

comprised of the following AUs/AU combinations:

• Baseline eyebrows + Baseline lips

• Eyebrows brought together AU4

• Eyebrows brought together (more intense)++

(AU4)

• Full brow up and eyelid lift (AU1+2+5)

• Inner brow up (AU1)

• Outer brow up (AU2)

• Lips up (AU12)

• Smile (AU6+12)

• Intense smile ++ (AU6+12)

• Lip corners down (AU15)

• Eyebrows brought together++ (AU4) + Lips

up (AU12)

• Eyebrows brought together++ (AU4) + Lip

corners down (AU15)

• Inner brow up (AU1) + Lips up (AU12)

• Inner brow up (AU1) + Lip corners down

(AU15)

Clips were presented randomly to control for

order effects. The first task consisted of rating the clip

on each of the Self-Assessment Manikin (SAM)

(Bradley, 1994) scales, a non-verbal pictorial

assessment technique that measures the arousal,

pleasure, and dominance associated with a person’s

affective reaction to an observed stimulus. The

second task consisted in choosing from a list of

emotion labels (happy, angry, surprised, fear, other

positive, other negative or neutral) the one that best

suited the face just watched (see Figure 6 with

screenshots used in the validation experiments).

Results were encouraging as they showed great

convergence in both content attribution and

emotional impact to many compositions, thus

validating our content assumptions when choosing

which expressive elements to include.

In regard to content, an independence Χ

test

performed on the cross tabulation of composition and

content, showed a highly significant association

between images and label (Χ

=2088.07; p<=.001;

N=978), with 12 out of 28 clips presenting

convergence above 75% (6 on the label angry, 4 on

happy, 1 on surprised and 1 o neutral). Convergence

was congruent with the expected content in all

convergent pictures.

Figure 6: Types of screens in the user experience.

As to emotional impact, images generally did not

elicit high Arousal (only 3 clips scored higher than

the Median). Valence was highly convergent, fitting

a narrow bell curve, with 100% of data below SD=1.9

and Mean=4.53, approaching the Median=5.

Regarding Dominance/control, 19 clips scored

higher than the Median, in an also centered

distribution where Mean=5.16 approaching the

UsingExpressiveandTalkativeVirtualCharactersinSocialAnxietyDisorderTreatment

353

Median=5 and 100% of data fell below SD=1.9; the

highest scores on feeling dominant occurred in

response to target faces that had been convergent on

positive content in the labelling task and resulted

from exposure to the female avatar, for exact same

compositions seen in male faces.

In conclusion, we were able to tabulate values for

relevant affective impact parameters - a crucial step

towards the setting up of a therapist’s manual with

validated content. It will be challenging to create

compositions that elicit higher levels of arousal and

Dominance/control. However, these may reveal

much different when the application is tested with a

clinical population, which is a future step in our

research.

4.2 Evaluation Phase

Another study was conducted with the goal of

evaluating the application on two main features:

usability and VH realism

We recruited 6 therapists (5F; 1M) ages 34- 59

years old. Within this group, only 2 elements had not

tried the previous version of the application. Tests

were conducted individually with each therapist, in a

dim light room (to improve visualization of images

and feeling of immersiveness). The apparatus for

tests was identical to that of typical application use: a

laptop computer connected to an LCD projector

displaying the image onto a projection canvas. Both

client and server were executed in localhost. Whilst

the therapist interface is shown in the laptop screen,

the client interface is projected onto the canvas

approaching real-life size. Two sound columns were

connected close to the projection canvas.

The evaluation was divided into 4 distinct phases:

character evaluation regarding realism (compared to

the previous version), usability of the implemented

functions, realism of the application as a whole, and

an open answer questionnaire section. Each part was

evaluated as the user performed each respective task.

As a result, in every section therapists considered

the new characters to be more realistic than those of

the previous version. The modifications made to the

configuration interface were welcomed as

improvements. Suggestions made to this interface

were:

• increase the number of available scenarios and

characters for higher versatility and the options

to edit each character;

• add new body animations (such as, looking at

the watch or touching the hair).

Each therapist was further asked to trigger a

specific speech in the “John” character and none had

difficulty performing this task. The favorite feature of

therapists in the new version is the VH’s ability to

speak.

Finally, every therapist mentioned that if

available, they would likely use the application in

their therapy sessions.

5 CONCLUSION AND FUTURE

WORK

One of the main questions related to the use of VRET

by the therapists is the cost of the immersive

equipment and some discomfort associated with its

use. Having this in mind, we sought to implement a

low cost solution that effectively assists in the

treatment of SAD resorting to VRET in the specific

situation of fear of public speaking, a common

problem.

The approach we propose involves ordinary

equipment: computer, sound system (two sound

columns are enough), projector and canvas (or a wall)

to project the simulations. The software for this

project (with no budget whatsoever) was developed

using freeware and free or very low cost 3D models.

The main disadvantage of this approach is the

difficulty in obtaining photo-realistic models.

Nonetheless, results from our content validation

experiments, facial expressions in our present models

are consistently interpreted by normative observers

and are decoded according to expectations derived

from emotion and expression science, which is

indirect evidence of realism. These gives us the

confidence to use the application to generate

animation clips to be used in the research domain of

non-verbal communication.

Tests performed with therapists further confirm

that the current models are more realistic than those

of the previous version. Speech articulation was seen

as a major improvement, and in general the idea of

using this tool in a clinical environment was

welcomed with enthusiasm.

The main follow up steps will be: i) to develop and

integrate in the application an artificial intelligence

module toward the simulation of emotions; ii)

afterwards, to validate the usefulness of the

application with a clinical population in a therapeutic

context.

ACKNOWLEDGMENTS

We acknowledge all the therapists and volunteers that

GRAPP2015-InternationalConferenceonComputerGraphicsTheoryandApplications

354

participated in our user study with no reward. We

thank our national funding agency for science,

technology and innovation (Fundação para a Ciência

e a Tecnologia- FCT) and the Research Unit LabMAg

for the financial support (through PEst-

OE/EEI/UI0434/2014) given to this work.

REFERENCES

APA- American Psychiatric Association, 2000. DSM-IV-

TR. American Psychiatric Publishing, Inc..

Beidel, D. C., Turner, S. M., 2007. Shy Children, Phobic

Adults: Nature and Treatment of Social Anxiety

Disorder. Washington, DC: APA, 2nd edition.

Bradley, M. M. and LANG, P. J., 1994. Measuring

Emotion: the Self-Assessment Manikin and the

Semantic Differential. J. Behav. Therap.. and Exp.

Psvchyatry, 25, 1, 49-59.

Cláudio, A.P., Carmo, M.B., Pinheiro, T., Esteves, F., 2013.

A Virtual Reality Solution to Handle Social Anxiety.

Int. J. of Creative Interfaces and Computer Graphics,

4(2):57-72.

Cláudio, A.P., Gaspar, A., Lopes, E., Carmo, M. B., 2014.

Virtual Characters with Affective Facial Behavior. In

Proc. GRAPP 2014, pp 348-355, SciTePress.

Douilliez, C. Yzerbyt, V., Gilboa-Schechtman, E.,

Philippot, P., 2012. Social anxiety biases the evaluation

of facial displays: Evidence from single face and multi-

facial stimuli. Cog and Emo, 26(6): 1107-1115.

Eibl-Eibesfeldt, I., 1989. Human Ethology. NY: Aldine de

Gruyter.

Ekman, P, Friesen, WV., Hager, J C., 2002. Facial action

coding system. Salt Lake City, UT: Research Nexus.

Gaspar, A., Esteves, F., 2012. Preschoolers faces in

spontaneous emotional contexts – how well do they

match adult facial expression prototypes? Int J Behav

Dev. 36(5), 348–357.

Gaspar, A., Esteves, F., Arriaga, P., 2014. On prototypical

facial expressions vs variation in facial behavior:

lessons learned on the “visibility” of emotions from

measuring facial actions in humans and apes. In M. Pina

and N. Gontier (Eds), The Evol. of Social Comm. in

poserPrimates: A Multidisciplinary Approach,

Interdisciplinary Evolution Research. New York:

Springer (pp.101-145).

Grillon, H., 2009. Simulating interactions with virtual

characters for the treatment of social phobia. Doctoral

dissertation, EPFL.

Haworth, M. B., Baljko, M., Faloutsos, P., 2012. PhoVR: a

virtual reality system to treat phobias. In Proceedings of

the 11th ACM SIGGRAPH International Conference

on Virtual-Reality Continuum and its Applications in

Industry, 171-174.

Herbelin, B., 2005. Virtual reality exposure therapy for

social phobia. Doctoral dissertation, EPFL.

James, L. K., et al., 2003. Social anxiety in virtual

environments: Results of a pilot study.

Cyberpsychology and behavior, 6(3), 237-243.

Klinger, E., et al., 2004. Virtual reality exposure in the

treatment of social phobia. Studies in health tech. and

informatics, 99, 91.

LaViola Jr, J. J., 2000. A discussion of cybersickness in

virtual environments. ACM SIGCHI Bulletin, 32(1),

47-56.

North, M. M., et al., 1998. Virtual reality therapy: an

effective treatment for the fear of public speaking. The

International Journal of Virtual Reality 3, 1-6.

Pertaub, D. P., Slater, M., Barker, C., 2001. An experiment

on fear of public speaking in virtual reality. Studies in

health tech. and informatics, 372-378.

Pertaub, D. P., Slater, M., Barker, C., 2002. An experiment

on public speaking anxiety in response to three different

types of virtual audience. Presence: Teleoperators and

Virtual Environments, 11(1), 68-78.

Russell, J. A., Fernandez-Dols, J. M., 1997. What does a

facial expression mean? In The psychology of facial

expression, pp. 3–30. New York, NY: Cambridge

University Press.

Slater, M., Pertaub, D. P., Steed, A., 1999. Public speaking

in virtual reality: Facing an audience of avatars. IEEE

CGandA, 19(2), 6-9.

Slater, M., et al., 2006. A virtual reprise of the Stanley

Milgram obedience experiments. PloS one, 1(1), e39.

Stein, M.B. and Kean, Y.M., 2000. Disability and Quality

of Life in Social Phobia: Epidemiologic Findings. Am

J Psychiatry 157: 1606–1613.

(last access to these sites on Dec 2014).

url-Blender http://www.blender.org/

url-Kinect http://www.microsoft.com/en-

us/kinectforwindows/

url-MakeHuman http://www.makehuman.org.

url-Poser http://my.smithmicro.com/poser-3d-animation-

software.html.

url-Unity http://Unity.com/

UsingExpressiveandTalkativeVirtualCharactersinSocialAnxietyDisorderTreatment

355