Computer-Based Social Anxiety Regulation in Virtual Reality Exposure Therapy

Dwi Hartanto

Research output: ThesisDissertation (TU Delft)

231 Downloads (Pure)


Social anxiety disorder (SAD), commonly referred to social phobia, is one of the most an immense and unreasonable fear of social interaction. Cognitive behaviour therapy (CBT) is the most thoroughly studied nonpharmacologic approach to the treatment of SAD patients. In CBT, patients are gradually, in vivo, exposed to anxiety-provoking real-life situations until habituation occurs and patients’ fear dissipates. Although effective for most patients, there are some clear limitations: some specific and required social situations are difficult to arrange due to unpredictability and possibly short duration of these naturally occurring social interactions, the therapist has limited control over anxiety provoking elements during the exposure, and individuals with social phobia have high refusal rate for in vivo exposure to the dreaded and fearful social situation.

Virtual Reality Exposure Therapy (VRET) has been suggested as an alternative to overcome many shortcomings of in vivo exposure. Contrary to exposure in vivo, in a VRET system the therapist can manipulate the exposure elements in a safer, manageable and cost-effective way. A VRET system presents fear eliciting stimuli to the patient in a Virtual Reality environment where the parameters of its anxiety evoking stimuli can be easily gradually manipulated by a therapist. The state of the art and recent enhancement of Internet and VR technology seems to be able to meet the ever-increasing demand for more accessibility and efficiency of mental healthcare services by bringing the VRET system directly at patient’s home. This thesis presents the development and evaluation of such an envisioned home-based VRET system for SAD patients.

The blue-print of the envisioned system design entails several key elements that need to be established. All those key elements were investigated and evaluated empirically in three separated studies followed up by a feasibility study. The first key element identified, is the system’s ability to measure the patient’s anxiety level automatically. Traditionally in VRET this is done using self-reported anxiety measurements, where patients are asked to report their anxiety every four or five minutes. Without the direct involvement of a therapist, it is up to the system to determine the appropriate timing, therefore, the timing is a crucial element. An empirical study involving 24 participants investigated the effects of three different types of automatic self-reported anxiety timing mechanisms (dialogue dependent, speech dependent and context independent). The results showed that the participants preferred a dialogue dependent timing mechanism above speech dependent or timing dependent mechanism, since it was considered as less interruptive. Moreover, the study also confirmed the needs for an accurate automatic self-reported anxiety timing mechanisms, as it could affect people’s experience and their behaviour in a dialogue with virtual human.

The second key element is the system’s ability to elicit and control the anxiety evoking stimuli within the social scene. This was investigated in two successive empirical studies. The first study investigated whether an exposure to various virtual social scenarios was associated with different levels of anxiety. The 24 participants were exposed to a free-speech dialogue interaction with a virtual character in a neutral world, blind date and job interview setting. The results showed that the participants’ level of anxiety increased significantly from the neutral world, the blind date to the job interview. This indicates that various virtual social scenarios are indeed able to evoke different levels of anxiety. The second study investigated anxiety control within a dialogue in VRET system. For this, the study assessed the association between the ratio of negative and positive dialogue responses made by a virtual character and individual’s level of anxiety. Twenty-four participants were exposed to two different experimental conditions: a positive, and a negative virtual job interview condition. In the positive condition, in the course of time the number of positive responses from the virtual character increased while negative responses decreased. In the negative condition, the opposite happened. The results showed that the manipulation of the dialogue style in both conditions had a significant effect on people’s level of anxiety, their attitude, their speech behaviour, their dialogue experiences, their own emotion, and how they perceived the emotion of the virtual human. These finding demonstrate that social dialogues in a virtual environment can be manipulated for therapeutic purposes effectively.

The third key element of the envisioned system is the possibility to introduce autonomous anxiety regulation. Traditionally, in the clinic a therapist tries to regulate the patients’ anxiety. However, in a home situation, a system would have to do this regulation automatically. A third empirical study was conducted to investigate and evaluate the ability and effectiveness of an automatic feedback-loop regulation mechanism for maintaining individual’s anxiety on a predefined target level. A group of 24 participants were exposed into two different system response conditions: a static and a dynamic condition. In the static condition participants were exposed to a static set of virtual reality stressors while in the dynamic condition they were exposed to a set of virtual reality stressors that changed dynamically aiming at keeping the anxiety of the participants at stable level. In the static condition, the anxiety dropped as indicated by decreased self-reported anxiety, decreased heart rate, increased heart rate variability, and longer answers. In contrary, in the dynamic condition, the participant’s anxiety level was maintained around a pre-set anxiety reference level. Therefore, the findings demonstrate that individuals’ level of anxiety can be regulated automatically using an automatic feedback-loop mechanism.

Besides those three important key elements of the system design blue-print, the envisioned system also has a number of practical and important elements such as the development of a virtual health agent, the therapist application, and a secure remote database server. Together these elements lay the foundation for a home-based VRET system. To evaluate the feasibility of such proposed system to treat people with SAD at home, an empirical study was conducted. The home-based VRET system was evaluated with a group of five social anxiety disorder patients. All patients received a complete home-based VRET system and were scheduled to perform 10 treatment sessions at home. The study findings showed that the proposed system could evoke the required anxiety, as expected, which over time dropped as patients’ self-reported anxiety and heart rate gradually decreased during the exposure sessions. To conclude, this thesis argues that the proposed home-based VRET system could evoke the required anxiety in patients with substantial level of presence. By meeting the above mentioned key challenges of our study, we showed that an effective home-based VRET system can be built and provided in due course. Therefore, this finding suggests that delivering a home-based VRET system is indeed possible, which could provide numerous benefits for both patients and therapists.
Original languageEnglish
QualificationDoctor of Philosophy
Awarding Institution
  • Delft University of Technology
  • Neerincx, M.A., Supervisor
  • Brinkman, W.P., Advisor
Thesis sponsors
Award date15 Jul 2019
Print ISBNs978-94-6186-841-1
Publication statusPublished - 2019


  • Virtual Reality Exposure Therapy (VRET)
  • Social anxiety disorder
  • Anxiety measurement
  • Emotion regulation
  • Speech recognition
  • Dialogue stressor
  • Automatic feedback-loop response
  • Intelligent systems
  • Virtual coach
  • Virtual health agent
  • Behaviour change support system
  • Self-therapy

Cite this