Fighting Substance Dependency Combining AAT Therapy and Virtual

Reality with Game Design Elements

Tanja Joan Eiler, Armin Gr

unewald and Rainer Br

uck

Medical Informatics and Microsystems Engineering, University of Siegen, H

olderlinstraße 3, 57076 Siegen, Germany

Keywords:

VR Applications, Virtual Reality, Digital Medicine, Game Design, Therapy, Substance Dependency, Ap-

proach Avoidance Task, Cognitive Bias Modiﬁcation, Approach Bias.

Abstract:

Smoking poses a signiﬁcant health risk and is still the main cause of premature mortality today. The Approach

Avoidance Task (AAT) developed by Rinck and Becker aims to develop a substance dependence therapy that

can reach the digital society. In this paper, a demonstrator that transfers the AAT procedure into virtual reality

(VR) is presented. This demonstrator was used to carry out an evaluation with twenty participants who were

asked to use the program and evaluate it by means of questionnaires and interviews. In addition, the reaction

times (RTs) of the test persons were recorded and evaluated. The results show that the transfer of the AAT

procedure to VR is possible and promising. Above all, the use of three-dimensional scenarios and objects,

with which one interacts during the training, were well received and increased the immersion as well as the

felt embodiment. The use of game design elements has also proved helpful and has had a positive inﬂuence

on user motivation.

1 INTRODUCTION

1.1 Motivation

Tobacco products are among the most consumed ad-

dictive substances along with alcohol (Batra et al.,

2015). In Germany, approx. 29 % of adults and 9.6

% of adolescents between twelve and seventeen years

are smokers, which is one of the many reasons why

smoking is still one of the leading causes of prema-

ture mortality today (Donath, 2017).

Nevertheless, little attention is paid to this addic-

tion disorder. Hardly any therapy or prevention pos-

sibilities are offered or (further) developed. Anyone

who wants to start a therapy must expect long waiting

times and stigmatization.

The aim of this and our further studies is to de-

velop new and innovative methods of intervention us-

ing digital media, which can be used for the treat-

ment of substance dependency and thereby improve

the success of therapy. The use of digital medicine

(Elenko et al., 2015), in the form of virtual reality

(VR) and mobile applications, is not only intended

to reduce the inhibition threshold for starting a ther-

apy, but also to improve its availability to the general

public and its effectiveness.

The Approach Avoidance Task (AAT) was cho-

sen as the basis therapy procedure for this cause, as

past and recent research has already provided promis-

ing results, which show that AAT is an effective addi-

tional method for the treatment of addiction diseases

like smoking or eating disorders (for an overview see

(Kakoschke et al., 2017)).

Whereas conventional methods of smoking ces-

sation address reﬂective processes by, for example,

informing patients about the negative consequences

of their behavior, AAT uses the cognitive bias mod-

iﬁcation (CBM) method to cover implicit processes

in addition to conventional therapy. Cognitive biases

are responsible for the selective processing of stimuli

in the environment, and thus inﬂuences the emotions

and motivation of the viewer (MacLeod and Mathews,

2012). In this context, smokers show a speciﬁc ap-

proximation behavior for images containing smoking-

related stimuli, and, at the same time a comparatively

reduced approximation to alternative images. This be-

havior is measurable and can be observed in the ap-

proach and avoidance biases. Due to this fact, the

AAT approach is well suited for additional smoking

cessation therapies using smoking-related stimuli, as

described in (Machulska et al., 2016).

The rest of this paper is structured as follows:

First, the theoretical background is explained. This

includes the AAT procedure, VR and game design

elements. Subsequently, the state of the art and the

Eiler, T., Grünewald, A. and Brück, R.

Fighting Substance Dependency Combining AAT Therapy and Virtual Reality with Game Design Elements.

DOI: 10.5220/0007362100280037

In Proceedings of the 14th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2019), pages 28-37

ISBN: 978-989-758-354-4

background knowledge will be summarized. Then,

the concept as well as the game elements used in this

application will be presented. The following section

describes the implementation itself and the evalua-

tion. Finally, the results are summarized, conclusions

will be drawn and an outlook is given.

1.2 Theoretical Background

1.2.1 The Approach Avoidance Task (AAT)

The AAT is a psychological procedure based on the

hierarchical model of Approach Avoidance Motiva-

tion. This assumes that each individual follows posi-

tive stimuli and moves towards them (”approach”) to

capture and hold them. In contrast, negative stimuli,

such as pain, are avoided, resulting in a movement

away from them (”avoidance”) to correct or escape

them. These movements are not only physical, but

also psychological, making this model a fundamen-

tal component for understanding and generating mo-

tivation or self-regulation (Elliot, 2006). Past stud-

ies have shown that people evaluate almost all stimuli

as positive or negative (Bargh, 1997), and according

to their evaluation react with approach or avoidance

to them (Lewin, 2013). Moreover, it was repeatedly

noted that this is accompanied by a pushing or pulling

movement of the arms (Chen and Bargh, 2016; Marsh

et al., 2005).

Based on this knowledge, the AAT procedure was

developed by Rinck and Becker (2007) . For this

method, participants were shown individual pictures

on a computer screen, to which they should react uti-

lizing a joystick. Depending on its movement, the size

of the currently displayed image changed: by pulling

the joystick, it became larger, while pushing made it

smaller (see Fig. 1). Instructed not to pay attention

to the content of the images, but only to a certain dis-

tinguishing feature, like the image format (portrait or

landscape) or its clearly visible tilting, the test per-

sons should thus subconsciously learn to push away

stimuli related to their addiction, as these are always

shown in the format that should be avoided. In ad-

dition, the reaction times (RTs) were measured and

evaluated to determine the approach/avoidance bias

of the participants. The calculation for each stimuli

type is as follows:

Bias = MedianRT [PUSH]− MedianRT [PULL] (1)

A positive value implies an approach bias and

suggests an afﬁrmative attitude towards the stimulus,

which is reﬂected in an approaching behavior. A

negative value represents an avoidance bias, result-

ing in an unfavorable attitude and averting behavior.

Figure 1: Idea of the Approach Avoidance Task.

Smokers should therefore get a higher approach bias

in smoke-related stimuli than non-smokers, as this

would imply an approaching behavior towards them.

The results of various studies (Kakoschke et al.,

2017) have shown that the AAT method could be used

for therapeutic purposes. Studies to this topic are still

undergoing, and this motivated us to transfer the AAT

procedure to VR in order to treat addiction disorders

more efﬁciently.

1.2.2 Virtual Reality and Game Design Elements

VR is a computer-generated three-dimensional alter-

native reality in which a person can interact and move

(Sherman and Craig, 2003; Simpson et al., 2000).

This virtual environment (VE), into which the user

can immerse, can be used in a wide range of applica-

tions, such as the simulation of an alternative reality

(duPont, 1995) or for the representation of complex

data (Pickover and Tewksbury, 1994; Simpson et al.,

2000). But it is also possible to use VR in quite dif-

ferent environments, e.g. in medical or military con-

text. Applications are also seen in education, archi-

tecture (virtual house inspections), psychology (treat-

ment of phobias and dependence illnesses), entertain-

ment (video games) and many other domains (Giraldi

et al., 2003).

Since users can immerse into the virtual world on

a visual, auditive, and motoric level, the feeling of

immersion is signiﬁcantly higher compared to con-

ventional systems. This type of human-computer in-

teraction is called ”presence” and can increase the ef-

fectiveness of interactions in VE (Schultze, 2010). If

implemented correctly, the VE can generate a high

cognitive absorption in the users, which means that

they lose the sense of time, can focus better, and feel

that the content is entertaining and arouses the curios-

ity for more. Temporarily, the user tends to get the

feeling that he has full control over the system and is

not depending on it (Agarwal and Karahanna, 2000).

This effect can be enhanced by the use of game de-

sign elements. The aim is to motivate users to act in a

Fighting Substance Dependency Combining AAT Therapy and Virtual Reality with Game Design Elements

more targeted manner and thus increase their general

motivation to interact with the system (Sailer et al.,

2017). In addition, AAT can contribute a lot to early

intervention when used in a timely manner, especially

to ensure that the users stay motivated for a sufﬁcient

period of time to carry out the therapy and do not drop

out prematurely (Boendermaker et al., 2015).

2 RELATED WORK

Lee et al. (2004) linked a VE to cue exposure ther-

apy (CET), which aim is to reprogram the behavior of

patients by learning to deal differently with situations

that trigger addiction. They are regularly exposed to

stimuli that cause craving until their respective toler-

ance level towards them has increased to such an ex-

tent that they can think about their actions in risk sit-

uations instead of involuntarily giving in to their urge

(Murphy, 2014). In the experiment of Lee and his col-

leagues, a VE was created which was composed of ad-

dictive environments and components. In their case,

it was a bar containing various objects such as ash-

trays, cigarette packs, or lighters, a smoking avatar,

and the audio track of a noisy restaurant. For all ob-

jects, three-dimensional models were used, since Lee

et al. had already found out in a previous study that

these trigger a much higher addiction pressure than

two-dimensional images (Lee et al., 2003). The study

participants, sixteen late adolescent men who con-

sumed at least ten cigarettes a day, were repeatedly

exposed to addictive stimuli in six sessions. The re-

sults showed that in the course of the test series the

number of cigarettes consumed as well as the addic-

tion pressure has gradually decreased, which is why it

can be assumed that the use of a VE within treatment

programs can be extremely helpful.

In the pilot study of Girard et al. (2009) , the

participants should ﬁnd, grasp and destroy up to 60

cigarettes, which were hidden in a medieval VE. Over

a period of four weeks, one session was held every

week, each lasting 30 minutes. The control group

had similar conditions, however, no cigarettes were to

be destroyed here, but balls were to be collected and

taken along. The results showed a statistically sig-

niﬁcant reduction in nicotine use and the abort rate of

the accompanying treatment program as well as an in-

crease in the abstinence rate compared to the control

group. In addition, 23 % of the participants stated that

they repeatedly remembered destroying the cigarettes

within the VE, which only 3 % of the participants in

the control environment did. This could indicate that

an increased belief in self-efﬁcacy can be achieved

if an individual watches himself destroying cigarettes

and also invests time and effort in ﬁnding others to de-

stroy them as well. The perceived embodiment within

the VE additionally contributed to the effectiveness.

A ﬁrst attempt to transfer CBM to VR for the

treatment of eating disorders was made in 2016 by

Schroeder et al. (2016) . Here, 23 participants were

asked to interact in a VR scenario with 3D objects

representing either a food item or a ball by making

a rejecting or gripping hand movement. The RT is

measured at three different points in time: At the be-

ginning of the hand movement, at object contact, and

as soon as the object has been collected. During the

study, the participants sat on a chair and wore an Ocu-

lus Rift DK2 as a Head-Mounted-Display (HMD). In

addition, the Leap Motion infrared sensor (Leap Mo-

tion, 2018) was used to track the hand movements

and transfer them to the VE. The visible virtual hand

should help the users think that it was their own hand

(body-ownership) (Slater et al., 2009). As soon as the

starting conditions (placing the dominant hand at a

predeﬁned starting position, HMD oriented centrally

and no head movements for 1.000 ms) were met, a 3D

object appeared in front of the players, to which they

should either react with a defensive hand movement,

or grab and collect it. During the game, a progress

bar ﬁlled up, and the last six collected objects were

displayed at the top of the screen. The study results

showed that food objects, especially with increasing

body mass index (BMI) of the test persons, were col-

lected signiﬁcantly faster than ball objects. In sum-

mary, VR in conjunction with CBM, possibly also to-

gether with other technologies, such as eye-tracking,

can be a helpful tool for the detection and treatment

of addiction disorders.

3 CONCEPT AND GAME DESIGN

3.1 Requirement Analysis

Figure 2: Three-dimensional mock-up of the virtual house.

As a pilot study (Eiler, 2018), a basic VR demon-

strator had to be implemented that should compare

the original desktop AAT (DAAT) procedure with a

slightly extended version that contains more game de-

HUCAPP 2019 - 3rd International Conference on Human Computer Interaction Theory and Applications

sign elements. The Unreal Engine and a HTC VIVE

HMD with a resolution of 2160 x 1200 pixels and a

refresh rate of 90 Hz will be used. The training con-

cept, e.g. the tilting of the images as a distinguishing

feature, is based on the AAT studies of Machulska

et al. (2016) . To make the two scenarios compa-

rable, the VE should consist of two rooms, one for

each procedure. Free movement within an area of 2

m x 2 m will be permitted. A mock-up for the three-

dimensional layout and design of the rooms, created

with eTeks Sweet Home 3D, Version 5.6, is shown in

Fig. 2. Within these rooms, in the sense of the AAT

procedure, various stimuli ought to be shown one after

another and have to be interacted with. They display

either addiction-related or neutral stimuli. As imple-

mented in the studies of Machulska et al. (2016) , the

addiction-related 2D images in the ﬁrst room have to

be tilted to the right and should be pushed, while the

neutral ones have to be tilted to the left and require to

be pulled. In the second room, where 3D objects in-

stead of 2D images should be used, the distinguishing

feature will be the border color of the stimuli: Red-

bordered objects ought to be thrown away into a trash

can and blue-bordered ones pulled and collected in-

side a cardboard box. Correctly treated stimuli will

disappear immediately. Following the original proce-

dure the next stimulus appears after the thumb stick

on a controller is pressed. Within the VE short in-

structions on the wall will show how the task should

be fulﬁlled in case the participants need to reassure

themselves within the training scenario.

While the participants complete the AAT train-

ing, their reaction times have to be recorded twice:

The time that has passed between the appearance of

the stimulus and the ﬁrst contact by the user, and the

time that has passed until the correct action has been

taken. Time measurements have to be taken as ac-

curate as possible, ideally in the millisecond range,

to calculate the approach/avoidance bias. This is due

to the fact that we want to measure and alter auto-

matic processes, which happen very fast. Reactions

lasting longer than 300 ms are accessible to rational

consciousness and therefore are no longer involuntary

(see P300 wave). In addition, the differences resulting

from AAT training are often only a few milliseconds,

and such accuracy is needed to validate these differ-

ences.

A conﬁguration ﬁle should allow the test leader to

specify the test subject ID, the degree of tilting, the

number of stimuli shown, how many of these stimuli

should show an addiction-independent distinguishing

feature and if a 3D model of the HTC Vive controller

or a robotic hand should be used to represent the in-

teractions of the users.

Our aim was to ﬁnd out if a transfer of the AAT

procedure to VR is possible and if so, which room

works better and what has to be changed in the de-

sign for the further course of the study. The measured

RTs and the biases derived from the measurement re-

sults were compared to the results of the studies by

Machulska et al. (2015) . Due to the fact that the

psychologists involved wanted as few distracting el-

ements as possible in order to make the results com-

parable, game design elements were kept to a mini-

mum in this pilot study. The evaluation carried out

should provide initial results in terms of design, user-

friendliness and immersion in order to develop an im-

proved demonstrator which will be used for the elab-

orated studies in the future.

The two rooms, the VE, and the training execution

are described in more detail below.

3.2 The Desktop Room

Figure 3: Interaction within the desktop room.

The ﬁrst room, called ”desktop room” (DR), should

differ as little as possible from the DAAT and con-

tains hardly any game design elements. For this rea-

son, 2D images ﬂoating in the middle of the room are

used here (see Fig. 3). These images, kindly provided

by Rinck and his colleagues (2007) , are either tilted

to the left or right because the distinguishing feature

ought to be clearly recognizable, but should not dis-

tract too much from the image content, as the avoid-

ance of addiction-speciﬁc stimuli is to be automated.

Since the zooming factor is extremely decisive in the

AAT procedure, the images pushed away are artiﬁ-

cially reduced in size, whereas the enlarging perspec-

tive effect is sufﬁcient when pulled. Correct actions

are rewarded with a positive sound effect, and press-

ing the thumb stick will display the next image. In the

event of an error, a negative-sounding tone rings out

and the ceiling light turns red until the correct move-

ment is done. In addition, the interactions are visu-

ally represented with two different 3D models (see

Fig. 4), which can be set via the conﬁguration ﬁle.

The ﬁrst model represents the HTC Vive Controllers,

which the test subjects will hold in their hands during

Fighting Substance Dependency Combining AAT Therapy and Virtual Reality with Game Design Elements

the experiment and which can also be seen on the in-

structions. The second model is a robotic hand with

gripping animations.

Figure 4: The two models that can be used for the interac-

tion.

3.3 The VR Enrichment Room

In the second room, the ”VR enrichment room”

(VRER), more game design elements are used, like

3D modeled surroundings as well as a particle effect

accompanying the appearance of the stimuli (Fig. 5).

The most important elements for the training are a ta-

ble and two containers: a cardboard box and a trash

can. The box stands between the users and the table

on which the objects will appear. Since the users stand

directly in front of the box and the trash can is lo-

cated behind the table, arm movements necessary for

the AAT training can thus be transferred into three-

dimensional space. Instead of 2D images, 3D objects

are used to further increase immersion since the stud-

ies by Lee et al. (2003) and Gorini et al. (2010) could

prove that they produce a higher craving than 2D im-

ages. The objects are to be sorted according to their

border color: In case of having a blue border, they

should be placed in the box, if they have a red bor-

der, they have to be thrown into the trash can. Visual

and acoustic feedback is the same as in the DR. The

appearance of the objects is accompanied by a purple

cloud of smoke, so that they, unlike the images in the

ﬁrst room, do not appear ”out of nowhere”. The added

elements should have the beneﬁt that the training is

more entertaining, which would increase motivation

and therefore the success rate of the therapy.

3.4 The Virtual World

The two rooms are connected by an elevator, which

has an operating panel and a level indicator. After

the successful completion of the ﬁrst room, which is

indicated by a green light, the elevator opens its door

so that the user can enter and proceed to the second

ﬂoor. The ride is accompanied by a sound ﬁle that

reproduces the sounds of a moving elevator. After

the user has reached the upper level, a bell sounds,

and at the same time, the door opens again so that

the user can step outside. The main purpose of the

elevator is to ensure that the start and end points are

always identical when changing rooms. If the rooms

were connected by a door, the user would have to turn

180 degrees to enter the next room without leaving

the boundaries of the tracking area, which would feel

unnatural and therefore disrupts the immersion.

Outside the house a virtual environment was cre-

ated, which is visible through the windows. It consists

of an extensive grassy landscape with trees, ﬂowers

and bushes. This should help ensure that the VE has

no visible end and appears more natural. In addition,

the oppressive feeling within the small rooms should

be reduced so that the users feel more comfortable

during the training.

3.5 Conducting the Training

Figure 5: The VR enrichment room.

In summary, the training is conducted as follows: The

user starts in the DR, where the thumb stick of the

controller needs to be pressed to start the training and

the ﬁrst run. Each run includes one stimulus. The

recording of the RT also starts at this point. After

pressing the button, the ﬁrst image appears, to which

the user must react according to its tilting. When the

image is grasped, by pressing the trigger, the ﬁrst RT

is saved. If the user’s subsequent action was cor-

rect, the image disappears, a positive-sounding tone

is played, the second RT is saved and the run is con-

sidered ﬁnished. However, if the action is faulty, the

ceiling light in the room turns red and a negative-

sounding tone rings out to signal the error. The im-

age is also moved back to its original location and the

current run continues until the correct action has been

executed. Once the image has been moved correctly,

the ceiling light returns to its natural color. In this case

the run is ﬁnished and the next one can be started by

pressing the thumb stick. Errors are also recorded,

since incorrect runs must not be taken into account

when calculating the bias. Once the amount of images

speciﬁed in the conﬁguration ﬁle have been handled

correctly by the user, the light turns green for one sec-

ond and the elevator opens. If this happens, the DR is

HUCAPP 2019 - 3rd International Conference on Human Computer Interaction Theory and Applications

considered ﬁnished. Via the lift the user now enters

the VRER, which is similar in functionality, with the

exception that 3D objects are used here whose bor-

der color speciﬁes into which container they must be

sorted. The ID of the user and the recorded data are

exported to an external .csv ﬁle.

4 IMPLEMENTATION

The Unreal Engine (UE) (Epic Games, 2017), in the

version 4.18, was used for the implementation of this

Windows x64 application. Mainly Blueprints were

used, a kind of visualized programming language of

the UE. However, some functions, e.g. for reading

and writing external ﬁles, have been programmed in

C++.

A ”canvas” is placed in the middle of the DR (see

Fig. 3), whose blueprint controls the entire function-

ality of this room. It contains an array in which all

pictures that can appear are stored. File names and

substring queries are used to determine whether the

stimulus is negative or positive. To determine whether

the executed movement was an pulling or pushing ac-

tion, the difference between the image distance be-

fore and after the participant grasps and releases it is

calculated. A buffer area of 6 cm in both directions

was implemented in order to ensure that the program

does not evaluate the slightest movement as pushing

or pulling away.

The VRER works very similarly, although 3D ob-

jects are used here. These are taken from the Google

Poly database (Google LLC, 2018). The border color

that decides into which container the item is to be

sorted, was implemented using Tom Looman’s ”Multi

Color Outline Post Process” extension (Tom Looman,

2015), which requires a post-process volume and the

use of a custom stencil. In order to be able to deter-

mine whether the objects have been sorted correctly,

they were provided with tags, which are queried in

the event of a collision with one of the trigger boxes

inside the containers. An ongoing sorting process can

be seen in Fig. 6. The smoke effect was taken from

the ”Inﬁnity Blade Effects” asset collection made by

Epic Games and adapted for the use inside the demon-

strator.

A challenge was the accuracy of the time record-

ing. This must be very precise, since even a few mil-

liseconds make a difference in the evaluation of the

recorded RTs in order to determine the cognitive bias.

Initially, it was attempted to achieve this over the sys-

tem time, but the result was not satisfactory because

the function call, implemented in C++, was made via

a custom blueprint node. Due to our inexperience in

Figure 6: Sorting within the VR enrichment room.

VR programming at this time, after some research we

found out that blueprints have a frame dependency.

This means that function calls are always executed

with the next frame, never between two frames, which

in turn means that the accuracy depends on the per-

formance of the used computer. Due to these circum-

stances, a time recording accurate to the millisecond

is not feasible using blueprints. Next, the time mea-

surement was implemented using a time line, which,

while it is running, continuously updates a variable in

which the time that has passed so far is stored. For

time measurement, this variable can simply be read

and the return value saved in further variables. This

resulted in an accuracy of 12 ms to 16 ms on the work-

ing machine, which is critical in terms of precision,

especially since less powerful computers are therefore

unsuitable.

5 METHOD

5.1 Participants and Design

Twenty participants (ten females and ten males; mean

age: 29.74 years, range: 18-60; ﬁve smokers) took

part in the evaluation. For 75 % of the test persons,

the experiment was their ﬁrst point of contact with

VR.

The experimental design required the participants

to be shown twenty stimuli in each room, the ﬁrst ten

of which show a tilting or border color independent of

the image content, which means that addiction-related

stimuli may be pulled and neutral ones pushed. In the

last ten stimuli, each smoking-related stimulus had

to be pushed away and vice versa. All participants

started in the DR and completed both rooms. They

conducted one test run to learn the controls and to de-

termine how understandable the demonstrator is only

with the contained instructions.

After signing a declaration of consent, the test per-

sons were introduced to the framework plot and func-

tionality of the program, while it was already known

that the demonstrator was aimed at the therapy of

Fighting Substance Dependency Combining AAT Therapy and Virtual Reality with Game Design Elements

nicotine addiction. After the participants put on the

HTC Vive HMD the application was started. During

the test run, the controller model was used in the DR

to visually represent the interaction, while the hand

model was used in the legitimate run whose RTs were

used for bias calculations.

Afterwards, all subjects were asked to complete

a non-standardized questionnaire containing ﬁfteen

questions to determine how they perceived the pro-

gram with regard to various components like compre-

hensibility of the task, control, user-friendliness, en-

joyment or the perception of certain game design el-

ements. By means of a short interview, some of the

answers had to be explained in more detail.

6 RESULTS

6.1 Observations

As anticipated, observations during the experiment

have shown that while using the DR especially those

participants who have not had any previous experi-

ence with VR have found it difﬁcult to perform the

correct arm movements. Only after a verbal instruc-

tion on how the movement should be carried out,

these test persons were able to continue the AAT

training correctly. Further problems that could be ob-

served in the DR include participants having difﬁculty

gripping the pictures. Often, despite the feedback

given by the vibration of the controller, they were un-

certain whether they were close enough to the image

to be able to grasp it. For the majority of the test sub-

jects, it was difﬁcult to determine which arm move-

ment is required by the tilting of the images, which is

why they often positioned themselves in such a way

that they always had the instructions in view. In addi-

tion, it was tedious for most participants to press the

thumb stick after each stimulus to make the next one

appear. Each subject forgot this at least once during

the training.

The observations in the VRER differ greatly from

those in the DR. Surprisingly, even the inexperienced

participants had signiﬁcantly fewer problems in carry-

ing out the required task, and instead began to experi-

ment with the objects during the test run. By using the

three-dimensional scenario and virtual hands, the sub-

jects had virtually no problems gripping and sorting

the objects. In addition, signiﬁcantly fewer mistakes

were made. Furthermore, the test persons felt more

comfortable and less oppressed in this room, which

led to a stronger immersion. Moreover, they had sig-

niﬁcantly more enjoyment performing the task, which

contributes positively to their motivation. However,

pressing the thumb stick to make the next object ap-

pear was also perceived as disrupting in this room. It

should be noted that the participants forgot to trigger

it much less frequently, which may be caused by the

fact that they could already internalize this behavior

through training in the DR.

6.2 Questionnaires

The evaluation of the questionnaires showed that both

immersion and embodiment were rated very well (on

average 7.5 and 7.4 points on a scale between 1 and

10). User-friendliness was generally perceived as

very good and beginner-friendly. The execution of the

task did not cause any further problems, except for the

difﬁculties in the DR mentioned above. The move-

ments felt natural, especially the VRER received very

good ratings. However, six people (30 %) stated that

they found the movements in the DR unnatural and

unusual. Two of them explained that this was due to

the fact that grasping ”ﬂoating images” felt peculiar.

Regarding the 3D models, 65 % of the participants

thought that the robotic hand model is more realis-

tic and makes the gripping movement, not only vi-

sually, more intuitive. 15 % of them found the hand

model to be better in principle, but noted that the con-

troller model is better suited for beginners, since the

instructions are easier to understand with it. 10 %

considered the controller model better because it re-

ﬂects reality due to the fact that this controller is held

in the real world. However, it was emphasized that the

hand model would be preferred instead if a glove or

another technique would be used that would transfer

own hand movements realistically into the VE. The

remaining 25 % of the respondents stated that they

found both models to be of equal value.

Only 5 % of the participants expressed a negative

opinion about the particle effect used because it felt

too intrusive. Another 5 % stated that although the

smoking effect was not disturbing, ideally it should

only be used for smoking-related stimuli and the pos-

itive stimuli should have a different effect. 55 % con-

sidered the effect to be appropriate, as the objects do

not appear ”out of nowhere”, instead their appearance

is ”spectacular and made interesting”. Surprisingly,

35 %, including 60 % of the participating smokers,

stated that they did not notice the effect during the

experiment.

6.3 Evaluation of RTs and Comparison

with the Desktop AAT

Fig. 7 and Fig. 8 summarize the RT distribution

of smokers and non-smokers per image category and

HUCAPP 2019 - 3rd International Conference on Human Computer Interaction Theory and Applications

Table 1: Evaluation of RTs.

Desktop Room VR Enrichment Room Desktop AAT

median time until ﬁrst contact 886 ms 851 ms /

median time until correct reaction 1416 ms 1319 ms 620 ms

mean duration of arm movement 530 ms 469 ms not necessary

Error rate smokers (%) 10 0 8

Error rate non-smokers (%) 8 6 11

Ø smoking-related bias (smokers) 256 44 25

Ø smoking-related bias (non-smokers) -312 -11 10

Ø neutral bias (smokers) 123,5 21.5 0

Ø neutral bias (non-smokers) 95 27.5 11

Figure 7: RT distribution of smokers.

Figure 8: RT distribution of non-smokers.

room. The results show that both groups were faster

in the VRER, whereby smokers were signiﬁcantly

quicker to pull smoking-related stimuli in the DR (M

= 964.7,SD = 118.9) than in the VRER (M = 1252,SD

= 146.7). The lower deviation from the mean value

when pulling smoking-related stimuli also supports

the statement that smokers have a higher approach

bias for these stimuli. In addition, it is noticeable

that smokers have generally reacted faster than non-

smokers in all categories.

The evaluation of the measured RTs compared to

the DAAT can be seen in Table 1. The DAAT val-

ues refer to the study by Machulska et al. (2015) , in

which 92 smokers and 51 non-smokers participated

and responded to smoking-related vs. non-smoking-

related pictures.

Our measured RTs show that the median time

elapsed in the DR before the ﬁrst contact with the

stimulus is 886 ms and 1416 ms until the correct in-

teraction with it, 530 ms were required to execute the

arm movement. The overall result of the VRER shows

that the subjects needed a median time of 851 ms un-

til the ﬁrst contact and 1319 ms until the correct arm

movement, the execution took 469 ms. As expected,

due to Fitt’s Law our measured RTs are longer than

with the DAAT, where the median RT was 610 ms

until the correct reaction.

The calculated bias values are comparable with

those of the DAAT, with the exception that smok-

ers show an approach bias for neutral stimuli instead

of being neutral towards them. The bias values of

the VRER come closer to those of the desktop ATT,

as the values of the DR are much larger. Neverthe-

less, in both rooms smokers had an approach bias to-

wards smoking-related stimuli, which suggests that

those were pulled faster than they were pushed away,

whereas non-smokers had an avoidance bias towards

them. Interestingly, non-smokers had no avoidance

bias when using the DAAT, but an approach bias,

which at ﬁrst would not be suspected.

The recorded data also show similarities between

the VRER and the DAAT regarding error rate, as

smokers made signiﬁcantly fewer mistakes (0 % vs.

8 %) than non-smokers (6 % vs. 11 %) in these pro-

cesses. In the DR, on the other hand, more mistakes

were made in general, as smokers had an error rate of

10 % and non-smokers of 8 %.

Fighting Substance Dependency Combining AAT Therapy and Virtual Reality with Game Design Elements

7 CONCLUSION

On the basis of our ﬁndings, the transfer of the AAT

procedure into virtual reality is promising, and should

be pursued further, as almost all required functions

are possible in VR, except for the time measurement,

which is not sufﬁciently accurate due to the frame de-

pendency of the blueprints.

An added value can be recognized by using

VR and game design elements, since the evaluation

showed that the VRER was appraised more favorably

in all areas, especially considering enjoyment and

therefore motivation. The use of border colors as the

distinguishing feature also resulted in a lower error

rate. As anticipated, participants found interactions

with 3D objects more realistic, intuitive and visually

appealing, which contributed to an increased pres-

ence. The recorded data also reﬂect that the VRER

works better, as not only the RTs are faster, but the

VRER also shows more similarities with the DAAT

compared to the DR, especially considering the cal-

culated bias values, which are an essential part of the

AAT therapy.

The built-in game design elements were positively

received by the participants and therefore receive a le-

gitimization for implementation. Still, it had to be

examined whether the smoke effect is a disturbing

element, since the test subjects could possibly form

an association between the effect and smoking itself,

which would lead to biased results. However, the as-

sessments of the subjects lead to the conclusion that

the smoke effect could be maintained in the further

development of the project and can represent an added

value, whereby alternative particle effects should be

considered.

With regard to control, most respondents favored

the robotic hand model, as it feels more natural and in-

creases the feeling of embodiment. However, the con-

troller model was considered useful to make it easier

for beginners, as the controller is mapped in the in-

structions. It was often criticized that the thumb stick

has to be pressed repeatedly to make the next image

or object appear. Here it would make more sense for

the next stimulus to appear immediately after the pre-

vious one has been treated correctly.

For the future development of the project, a new

demonstrator, which incorporates the results of this

study, ﬁxed the problems regarding the time measure-

ment and corrects design errors, is already in devel-

opment. Solutions suitable for therapeutic purposes,

which allow a visualization of the entire body within

the VE without interfering with the training, will be

examined. This would make the VR experience even

more realistic and pleasant, as it would improve em-

bodiment and therefore presence. In addition, this

would allow users to move more securely within the

VE. Regarding controls, the usage of the Leap Mo-

tion sensor (Leap Motion, 2018) or data gloves will be

evaluated, since hand movements can be transferred

even more faithfully to the VE. In this way, a positive

effect on embodiment can be expected.

In addition, consideration can be given to intro-

ducing gamiﬁcation elements like progress bars, a

scoring system, achievements, or other elements, as

there is a signiﬁcant increase in user motivation when

the context in which points or other rewards are re-

ceived is more closely related to the actions per-

formed - even if there are no stories or characters

(Dovis et al., 2012). The increased enjoyment while

playing should thereby raise the motivation to con-

tinue the therapy.

REFERENCES

Agarwal, R. and Karahanna, E. (2000). Time ﬂies when

you’re having fun: Cognitive absorption and beliefs

about information technology usage. MIS Quarterly,

24(4):665.

Bargh, J. A. (1997). Advances in social cognition. In Wyer,

R. S., editor, The automaticity of everyday life, vol-

ume 10 of Advances in social cognition, pages 20–27.

Erlbaum, Mahwah, NJ.

Batra, A., Hoch, E., Mann, K., and Petersen, K. U. (2015).

S3-Leitlinie Screening, Diagnose und Behandlung des

sch

adlichen und abh

angigen Tabakkonsums. Springer

Berlin Heidelberg, Berlin, Heidelberg and s.l., 1. auﬂ.

2015 edition.

Boendermaker, W. J., Prins, P. J. M., and Wiers, R. W.

(2015). Cognitive bias modiﬁcation for adolescents

with substance use problems–can serious games help?

Journal of behavior therapy and experimental psychi-

atry, 49(Pt A):13–20.

Chen, M. and Bargh, J. A. (2016). Consequences of au-

tomatic evaluation: Immediate behavioral predisposi-

tions to approach or avoid the stimulus. Personality

and Social Psychology Bulletin, 25(2):215–224.

Donath, C. (2017). Drogen- und suchtbericht 2017.

https://www.drogenbeauftragte.de/ﬁleadmin/dateien-

dba/Drogenbeauftragte/Drogen

und Suchtbericht/

ﬂipbook/DuS 2017/ﬁles/DuS 2017 download.pdf.

Last checked on Oct 31, 2017.

Dovis, S., van der Oord, S., Wiers, R. W., and Prins, P. J. M.

(2012). Can motivation normalize working mem-

ory and task persistence in children with attention-

deﬁcit/hyperactivity disorder? the effects of money

and computer-gaming. Journal of abnormal child psy-

chology, 40(5):669–681.

duPont, P. (1995). Building complex virtual worlds without

programming. EUROGRAPHICS’95 State Of The Art

Reports, pages 61–70.

Eiler, T. J. (2018). Conception and implementation of an

application for the treatment of dependency disorders

HUCAPP 2019 - 3rd International Conference on Human Computer Interaction Theory and Applications

in virtual reality using gamiﬁcation of the approach-

avoidance-task (aat). Master’s thesis, University of

Siegen.

Elenko, E., Underwood, L., and Zohar, D. (2015). Deﬁning

digital medicine. Nature biotechnology, 33(5):456–

461.

Elliot, A. J. (2006). The hierarchical model of approach-

avoidance motivation. Motivation and Emotion,

30(2):111–116.

Epic Games (2017). Unreal engine features.

https://www.unrealengine.com/en-US/features.

Last checked on Oct 27, 2017.

Giraldi, G., Silva, R., and Oliveira, J. (2003). Introduction

to virtual reality. LNCC Research Report, 6.

Girard, B., Turcotte, V., Bouchard, S., and Girard, B.

(2009). Crushing virtual cigarettes reduces tobacco

addiction and treatment discontinuation. Cyberpsy-

chology & behavior : the impact of the Internet, mul-

timedia and virtual reality on behavior and society,

12(5):477–483.

Google LLC (2018). Poly: Explore the world of 3d.

https://poly.google.com/. Last checked on Mar 27,

2018.

Gorini, A., Griez, E., Petrova, A., and Riva, G. (2010). As-

sessment of the emotional responses produced by ex-

posure to real food, virtual food and photographs of

food in patients affected by eating disorders. Annals

of general psychiatry, 9:30.

Kakoschke, N., Kemps, E., and Tiggemann, M. (2017). Ap-

proach bias modiﬁcation training and consumption: A

review of the literature. Addictive behaviors, 64:21–

28.

Leap Motion (2018). Leap motion: Reach

into virtual reality with your bare hands.

https://www.leapmotion.com/. Last checked on

Apr 23, 2018.

Lee, J., Lim, Y., Graham, S. J., Kim, G., Wiederhold, B. K.,

Wiederhold, M. D., Kim, I. Y., and Kim, S. I. (2004).

Nicotine craving and cue exposure therapy by using

virtual environments. Cyberpsychology & behavior

: the impact of the Internet, multimedia and virtual

reality on behavior and society, 7(6):705–713.

Lee, J. H., Ku, J., Kim, K., Kim, B., Kim, I. Y., Yang, B.-

H., Kim, S. H., Wiederhold, B. K., Wiederhold, M. D.,

Park, D.-W., Lim, Y., and Kim, S. I. (2003). Experi-

mental application of virtual reality for nicotine crav-

ing through cue exposure. Cyberpsychology & behav-

ior : the impact of the Internet, multimedia and virtual

reality on behavior and society, 6(3):275–280.

Lewin, K. (2013). Dynamic Theory of Personality - Selected

Papers. McGraw-Hill paperbacks. Read Books Ltd.

Machulska, A., Zlomuzica, A., Adolph, D., Rinck, M., and

Margraf, J. (2015). A cigarette a day keeps the goodies

away: smokers show automatic approach tendencies

for smoking–but not for food-related stimuli. PloS

one, 10(2):e0116464.

Machulska, A., Zlomuzica, A., Rinck, M., Assion, H.-J.,

and Margraf, J. (2016). Approach bias modiﬁcation in

inpatient psychiatric smokers. Journal of psychiatric

research, 76:44–51.

MacLeod, C. and Mathews, A. (2012). Cognitive bias mod-

iﬁcation approaches to anxiety. Annual review of clin-

ical psychology, 8:189–217.

Marsh, A. A., Ambady, N., and Kleck, R. E. (2005). The ef-

fects of fear and anger facial expressions on approach-

and avoidance-related behaviors. Emotion (Washing-

ton, D.C.), 5(1):119–124.

Murphy, K. (2014). Cue exposure therapy: What the

future holds. https://www.rehabs.com/pro-talk-

articles/cue-exposure-therapy-what-the-future-holds/.

Last checked on Nov 11, 2017.

Pickover, C. A. and Tewksbury, S. K., editors (1994). Fron-

tiers of scientiﬁc visualization. A Wiley-Interscience

publication. Wiley, New York.

Rinck, M. and Becker, E. S. (2007). Approach and avoid-

ance in fear of spiders. Journal of behavior therapy

and experimental psychiatry, 38(2):105–120.

Sailer, M., Hense, J. U., Mayr, S. K., and Mandl, H. (2017).

How gamiﬁcation motivates: An experimental study

of the effects of speciﬁc game design elements on psy-

chological need satisfaction. Computers in Human

Behavior, 69:371–380.

Schroeder, P. A., Lohmann, J., Butz, M. V., and Plewnia,

C. (2016). Behavioral bias for food reﬂected in hand

movements: A preliminary study with healthy sub-

jects. Cyberpsychology, behavior and social network-

ing, 19(2):120–126.

Schultze, U. (2010). Embodiment and presence in virtual

worlds: A review. Journal of Information Technology,

25(4):434–449.

Sherman, W. R. and Craig, A. B. (2003). Understand-

ing virtual reality: Interface, application, and design.

Morgan Kaufmann series in computer graphics and

geometric modeling. Morgan Kaufmann, San Fran-

cisco, CA.

Simpson, R. M., LaViola, J. J., Laidlaw, D. H., Forsberg,

A. S., and van Dam, A. (2000). Immersive vr for sci-

entiﬁc visualization: A progress report. IEEE Com-

puter Graphics and Applications, 20(6):26–52.

Slater, M., Perez-Marcos, D., Ehrsson, H. H., and Sanchez-

Vives, M. V. (2009). Inducing illusory ownership of a

virtual body. Frontiers in neuroscience, 3(2):214–220.

Tom Looman (2015). Multi-color outline post process

in unreal engine 4. http://www.tomlooman.com/

multi-color-outline-post-process-in-unreal-engine-4/.

Last checked on Mar 27, 2018.

Fighting Substance Dependency Combining AAT Therapy and Virtual Reality with Game Design Elements