Using consumer story scripting software to evoke emotions and empathy

- A pilot within-group experiment

Anna Hermannsdottir Supervisor: Philip Lindner BACHELOR’S THESIS, 15 HP PSYCHOLOGY III, SPRING 2018

STOCKHOLM UNIVERSITY DEPARTMENT OF PSYCHOLOGY

USING CONSUMER STORY SCRIPTING SOFTWARE TO EVOKE EMOTIONS AND EMPATHY: A PILOT WITHIN-GROUP EXPERIMENT

Anna Hermannsdottir

New technology provides the possibility of delivering exposure therapy for socially anxious individuals through virtual environments. This study investigated whether emotional responses and empathy can be experienced for virtual characters (avatars) and whether the evaluations differ depending on level of social anxiety. Six scenes depicting avatars interacting were created through the consumer story scripting software Plotagon and then replicated with real humans. 102 participants viewed the scenes and appraised their emotional response and level of empathy. Results revealed the avatars varied in ability to elicit positive emotions, yet were equally successful in the negative conditions. An association was found between high social anxiety and a more negative emotional response of the scenes with humans but not with avatars. In conclusion it was found possible to feel emotions and empathy for virtual characters in a manner somewhat similarly to that for humans.

Social interaction is a vital part of human everyday life. Having and maintaining social relationships is directly connected to a person’s wellbeing while loneliness has been found to have a negative effect on health, similar to that of smoking (Holt-Lundstad, Smith, Baker, Harris & Stephenson, 2015). Human interaction is complex; filled with social cues including verbal and non-verbal language (Isbister & Nass, 2000), and can be a source of distress when social validation and acceptance from one’s peers is not experienced (Masten et al., 2009). Today, technology has become the primary means of communication. Social exchange occurs constantly and continuously through the Internet, meaning we no longer need to go outside in order to connect socially with others. Access to social networking sites, virtual platforms and online role-playing games, means people can replace real-life social interaction with online and virtual social interaction – something that appears to be particularly common for people who are uncomfortable in social situations (Shepherd & Edelmann, 2005).

A study by Sioni, Burleson and Bekerian (2017), found excessive use of online role-playing games to be strongly correlated with high social anxiety and identifying with one’s online avatar. It was suggested that, through the avatar, the players experienced feelings of accomplishment and self-enhancement, which these players did not receive from face-to-face interactions. Namely, those uncomfortable with real-life social interaction tend to fulfill their social needs in other ways, in this case through their gaming avatar (Sioni et al., 2017). Yet, more studies are needed that investigate the positive outcomes virtual interactions could have on real-life social interactions, especially for those who are socially anxious. According to the Threshold Model of Social Influence (Blascovich, Loomis, Beall, Swinth, Hoyt & Bailenson, 2002), virtual environments influence its users socially

2

when social verification is experienced. Social verification occurs when social cues verify that the communication is meaningful, such as eye contact, gestures and facial expressions. This social exchange is mediated by the behavioral realism of the virtual human and by social presence, namely whether the virtual character is perceived to represent a real person. For example, a human controlled avatar supposedly generates social responses regardless of realistic behavior, while a computer controlled avatar needs to be highly realistic in order to evoke a social reaction (Blascovich et al., 2002).

However, despite the self-evident claim that computers are not human, studies show that sensing social presence is not always necessary in order for users to be socially influenced. A study by Kothgassner, Kafka, Rudyk and Beautl (2014) revealed feelings of social exclusion to be perceived as equally negative in virtual and face-to-face encounters, regardless of the participants thinking humans controlled the virtual characters or whether they were fully aware that the character was a computer (Kothgassner et al., 2014). A study by Nass and Moon (2000) further confirms this. They discovered that people tend to apply human categories to computers, such as gender-stereotypes and identifying ethnically with the computer-generated individual. People even engaged in social behaviors like showing politeness and reciprocity towards the computer. Furthermore, this was seen irrespective of the participants believing the characters were human or computer-controlled (Nass & Moon, 2000).

Regarding behavioral realism, a moderate level of realism might be favorable. Nearly 50 years ago, the Japanese robotics professor Masahiro Mori (1970) attempted to explain the “eeriness” we experience when a very lifelike robot does something we perceive as non-human. For example, imagine watching a very life-like robot speak. At first it looks human, but eventually one will realize because of the unresponsive eyes and strange facial expressions that it is not actually alive. This abrupt shift from empathy to revulsion is called the “Uncanny Valley” (Mori, 1970; Mori, MacDorman & Kageki, 2012). There are many different hypotheses as to why this happens, with one possible explanation being that it arises from a mismatch in perception, such as artificial eyes on a human-like face (Kätsyri, Förger, Mäkäräinen & Takala, 2015). A possible solution in order to prevent certain descent into the Uncanny Valley is to simply pursue a moderate degree of human likeness (Mori, 1970; Mori et al., 2012). For example, the animators of the PDI/Dreamworks movie Shrek had to make princess Fiona slightly less life-like because “she was beginning to look too real, and the effect was getting distinctly unpleasant” (Modesto, quoted in Weschler, 2002). What might be essential in order to be socially influenced by a virtual environment is the ability to feel empathy. Empathy relies on sharing another person’s affect as well as perspective (mentalizing) while maintaining a self-other distinction (Lamm, Rütgen & Wagner 2017). Empathy appears to be an important predictor for emotional understanding of narratives (Leite et al., 2017) and pro-social behavior towards a virtual character (Felnhofer, Kafka, Hlavacs, Beutl, Exner & Kothgassner, 2017). Several studies have confirmed a relationship between ability to see things from another person’s perspective and ability to see things from a character’s perspective (Chory-Assad & Cicchirillo 2005; Nomura & Akai, 2012).

3

Social Anxiety and Exposure Therapy

Those who are highly anxious in social situations could be diagnosed with Social Anxiety Disorder (SAD), which is a common anxiety disorder characterized by fear of social situations in which the individual experiences a risk of being judged by others (American Psychiatric Association, 2013). These situations include everyday conversations, meeting new people, being observed, and performing in front of others. This anxiety usually stems from the individual’s fear of being negatively evaluated and consequently rejected, either because of the way they act or because they are showing symptoms of anxiety that may be perceived as embarrassing (such as blushing and shaking). This causes the individual to either avoid these social situations or to withstand them with severe discomfort (American Psychiatric Association, 2013). Despite being one of the most common psychological disorders in the world, with a Swedish point prevalence estimated at 15.6% (Furmark, Tillfors, Everz, Marteinsdottir, Gefvert & Fredrikson, 1999), few receive help. The very symptoms experienced by SAD sufferers result in a reluctance to seek treatment since it requires them to interact with other people. Only one third of individuals with SAD seek treatment (Ruscio, Brown, Chiu, Sareen, Stein & Kessler, 2008) and the most commonly reported reasons for not seeking help are financial, uncertainty about where one can receive help, and fear of negative evaluation (Olfson, Guardino, Struening, Schneier, Hellman & Klein, 2000). There are three types of fears in SAD that have been identified through several studies; social interaction fears, performance fears, and observation fears (Bögels et al., 2010). Performance anxiety (or fear of public speaking) is the most common and has the most compelling evidence for creating a subtype (Bögels et al., 2010). Specifically, 77% of those diagnosed with SAD fear speaking or performing in front of an audience, and 20% of these report this being their only feared social situation (Tillfors & Furmark, 2007). SAD can therefore be specified as “Performance anxiety only” in the newest version of DSM-V (American Psychiatric Association, 2013). Although performance fears are common, it could be argued that social interaction fears leads to a more limited everyday life since social interactions are more frequent than public speaking. People with social interaction fears (or social interaction anxiety) tend to avoid certain situations or rely on maladaptive safety behaviors, e.g. drinking alcohol at a party (Abramowitz, Deacon & Whiteside, 2011). The fear of interactions can interfere with occupational, relational and other life goals, resulting in a lower quality of life (Safren, Heimberg, Brown & Holle, 1996).

Exposure therapy is a well-established treatment for anxiety disorders and consists of gradually exposing the patient to the feared stimulus (Abramowitz et al., 2011), which in turn leads to a decreased level of anxiety due to habituation and desensitization (McNally, 2007; Wolpe, 1958). Since people with SAD tend to overestimate the risk of being judged, exposure to the feared situations is thought to correct this distorted belief (Abramowitz et al., 2011; Deacon & Abramowitz, 2004). Exposure therapy is commonly used in clinical settings, treating conditions such as SAD, specific phobias, OCD, and PTSD (Abramowitz et al., 2011). However, with new technology comes the possibility of receiving exposure therapy through virtual environments (Emmelkamp, 2005). There are many advantages in using technology to deliver exposure therapy; it can be delivered anywhere, is practical and allows for more control over the played-out situation. Therapists can vary the situations and contexts endlessly, without ever leaving

4

the office (Bouchard et al., 2017; Emmelkamp, 2005). Additionally, virtual exposure is preferred by patients too anxious to undergo real-life exposure (Emmelkamp, 2005).

Plotagon and Previous Studies In 2011 a Swedish start-up company created the application Plotagon (Plotagon, 2018), a platform allowing its users to create animated characters and movies. The animations are so-called “low-polygon”, meaning they appear to be cartoons instead of real-life (see Figure 1). The application is easily operated and allows for full customization of everything from virtual characters (avatars), to locations, conversations, placement of the camera and sound effects. The program offers pre-recorded computer voices, or one can choose to record their own voice and the avatar will lip-sync the sentences. In short, Plotagon allows its users to create whatever scenes they want. The application can be used for educational purposes, marketing, and for individuals to express themselves (Plotagon, 2018). Consequently, it could prove to be a useful tool in Psychology as well.

Figure 1. Sample images from Plotagon

One possible use of Plotagon is in the treatment of SAD. It is believed to be a successful application in clinical settings as a tool delivering exposure therapy to individuals with interaction anxiety, as it offers a visual representation of social interactions and allows for customized scenarios. Customization is necessary for individuals with SAD since interaction fears can appear in many different situations depending on with whom one is interacting, where one is, and what the purpose of the interaction is (Ruscio et al., 2008). For some the anxiety only arises in a few contexts, such as dating situations, when interacting with an authority or when talking to a stranger. Others experience anxiety in many different situations (Ruscio et al., 2008). Through Plotagon, one can play out the feared situation and even control the full content of the conversation. A “worst-possible-scenario” can therefore be customized, resulting in more effective exposure. Moreover, customization allows the user to create an avatar similar to oneself. In a study by Aymerich-Franch, Kizilcec and Bailenson (2014), participants were able to choose whether they wanted an avatar dissimilar or similar to themselves in a virtual public speaking task. Social anxiety correlated positively with a preference for embodying a dissimilar avatar and participants assigned to a similar avatar experienced 14% higher levels of anxiety during the task (Aymerich-Franch et al., 2014). Hence, avatar similarity and identification could be an important factor in generating anxiety. Many studies have already examined the use of Virtual Reality Exposure Therapy (VRET) for performance fears (such as public speaking) and other anxiety disorders

5

like phobias and PTSD (Emmelkamp, 2005). Several meta-analyses confirm it being as effective as real-life exposure therapy (Emmelkamp, 2005; Powers & Emmelkamp 2008). Only a few studies have examined the effectiveness of virtual exposure therapy on social interaction anxiety, see Table 1. Table 1. Summary of studies investigating virtual exposure therapy for social interaction anxiety, using either HMD (Head Mounted Display), computer-screens or one-screen projection.

Study Participants Objective Main findings

Klinger et al., 2005

n=18 Virtual exposure n=18 group CBT (clinical sample)

To examine the effectiveness of exposure therapy through virtual situations on a computer in treatment of SAD, compared to Group CBT

Treatment was equally effective in both groups (Klinger, Bouchard, Légeron, Roy, Lauer, Chemin & Neugues, 2005).

Morina et al., 2014

n=38 (non-clinical sample)

Whether interactions between individuals and virtual humans can produce levels of anxiety, along with a comparison of HMD and one-screen projection.

The interactions with virtual humans caused anxiety. Levels of anxiety did not differ between groups; those using HMD had similar anxiety levels as those using one-screen projections (Morina, Brinkman, Hartanto & Emmelkamp, 2014)

Morina et al., 2015

High SAD n=16, low SAD n=18 (non-clinical sample)

Examining levels of anxiety and the effectiveness of exposure therapy when interacting with virtual humans through HMD or one-screen projection

Individuals with high SAD scores reported lower levels of anxiety post-treatment compared to pre-treatment (Morina, Brinkman, Hartanto, Kampmann & Emmelkamp, 2015)

Kampmann et al., 2016

n=20 VRET n=20 real-life exposure n=20 waitlist (clinical sample)

Comparing VRET with real-life exposure therapy in treatment of SAD through HMD

Traditional exposure was more effective than VRET, both conditions reduced social anxiety symptoms post-treatment (Kampmann, Emmelkamp, Hartanto, Brinkman, Zijlstra & Morina, 2016).

6

Bouchard et al., 2017

n=17 VRET n=22 real-life exposure n=20 waitlist (clinical sample)

Investigating whether VRET for SAD-patients is effective, using HMD

VRET was more effective than traditional exposure and therapists reported VRET to be more practical (Bouchard et al., 2017)

Purpose Studies have already examined the effectiveness of exposure therapy for many anxiety disorders, such as performance anxiety and phobias. Yet, only a few have studied the effectiveness of virtual exposure therapy on social interaction anxiety. While the results of these few studies are promising, the limited number of participants and the use of HMDs in most of the studies, calls for research including more participants and a tool that allows for full customization. Plotagon has the potential to provide an easy way of creating scenes that could be useful as exposure to anxiety-generating situations.

Before examining whether Plotagon can be used for exposure therapy, a study is needed to validate that people are able to react emotionally to scenes depicting interactions between Plotagon avatars. Since Plotagon creates scenes with virtual humans in different scenarios, this study compares scenes created through Plotagon with scenes created with real human actors in order to evaluate whether emotional involvement is possible despite the low-polygon graphics. This pilot study aims to investigate i) whether scenes created through Plotagon are equally effective as scenes with real humans in eliciting an emotional response, ii) if it is possible to empathize with a virtual character as one would with a human character and iii) whether participants with high levels of social anxiety report more negative emotions when watching the scenes.

Method Participants

The participants were mainly students at Stockholm University and individuals acquainted with the author. Out of the 137 persons who initiated the online survey, 103 completed it. One participant was excluded from the analysis for responding one of the control questions incorrectly, resulting in a total of 102 participants between 18-58 years old (M=27.5, SD=8.64), 21 were male and one participant responded “other/do not want to respond” when stating their gender. 87 participants reported studying to be their main occupation, while 13 were working, one was on sick leave and one on parental leave. 73 of the students received course credit from Stockholm University for partaking in the survey.

Apparatus and Material

The Social Interaction Anxiety Scale (SIAS) and Social Phobia Scale (SPS) are two companion scales developed to measure social interaction anxiety and social performance anxiety. Originally created by Mattick and Clarke (1998), the measures have been shown to successfully measure SAD symptoms separated from other anxiety disorders (Peters, 2000). This study used the short versions of these measures, SIAS-6 and SPS-6, developed by Peters, Sunderland, Andrews, Rapee and Mattick (2012) in order to reduce the time of administration. The correlations between the short and

7

original versions of the scales are high (r>0.89, p<0.001) and the short versions are equally effective in discriminating between those with and without SAD (Peters et al., 2012). In the current study SIAS-6 and SPS-6 presented high reliability (α=0.897 for all 12 items). Examples of statements are “I feel tense if I am alone with just one person” and “I worry about shaking or trembling when I’m watched by other people”. Each assertion is rated along a five-point Likert scale where 0=Not at all characteristic of me, and 4=Extremely characteristic of me. The Swedish translation of the scales has been validated by Carlbring, Andersson, Cuijpers, Riper and Hedman-Lagerlöf (2017). Questionnaire of Cognitive and Affective Empathy (QCAE) was developed by Reniers, Corcoran, Drake, Shryane and Völlm (2011) and includes 31 items measuring cognitive and affective empathy derived from the commonly used questionnaires IRI (Davis, 1983), IVE (Eysenck & Eysenck, 1978), EQ (Baron-Cohen, Richler, Bisarya, Guruanathan & Wheelwright, 2003) and HES (Hogan, 1969). The QCAE has demonstrated strong convergent and construct validity and is negatively correlated to psychopathic personality traits (Reniers et al., 2011). Examples of statements are “I find it easy to put myself in somebody else’s shoes” and “I get very upset when I see someone cry”. Each statement is rated along a four-point Likert scale where 0=Strongly Disagree and 3=Strongly Agree. The scale was translated into Swedish and then validated by translating back to English. In this study the scale had a high internal consistency (α=0.82). The State Empathy Scale (SES), developed by Shen in 2010, measures affective, cognitive, and associative empathy during message processing. In the current study, reliability was high with a Cronbach’s alpha of α=0.92. The following three items were selected to be included in this survey: “I can feel the character’s emotions”, “The character’s reaction to the situations are understandable” and “I can relate to what the character was going through”. Each assertion represented a dimension of affective, cognitive, or associative empathy and was selected because of the high correlation with Trait Empathy (Shen, 2010). The scale was translated into Swedish and then validated by translating back to English.

Two Visual Analogue Scales (VAS) were included, where the participants were asked to score how strongly they experienced a negative or positive emotion when watching the scene (Experienced Emotional Response), and how strongly negative or positive the emotion was that they perceived the protagonist had experienced in the scene (Perceived Emotional Response). The participants were asked to score the emotion with a slider ranging from -100 to 100 where -100 represented a strongly negative emotion, 0 a neutral emotion and 100 a strongly positive emotion. This separation of emotional measures was made in order to investigate how successful the scenes were in eliciting emotions while also measuring how well the character portrayed the emotions.

Preparation of scenario stimuli

In order to ensure that the results did not rely solely on one type of scenario, two different types of scenes were created: a classroom scene and a café scene. The negative versions were created first and displayed situations that were expected to be perceived as humiliating and embarrassing for the protagonist. Since individuals with SAD have an intense fear of negative evaluation (American Psychiatric Association, 2013), the negative scenes aimed to provoke a negative reaction. The main character of all scenes was called “Emma”. The negative classroom scene showed Emma being laughed at for

8

answering a question incorrectly. This scenario was hoped to portray a common fear of individuals with SAD, namely fear of speaking in class (Cox, Clara, Sareen & Stein, 2008). The negative café scenes instead displayed a dating situation, where Emma was cruelly rejected by her date for “not being what he expected”. Fear of dating situations is also a common fear for individuals with SAD (Cox et al., 2008), especially for males (American Psychiatric Association, 2013). Each type of scene also had a positive and neutral version that included the same characters but had a plot with the aim of eliciting either positive or neutral emotions. The scenarios were created with the negative version as a starting point where the positive scenes essentially displayed scenarios that were the opposite of the negative scenes. The neutral scenes were instead made to fit the same context as the positive and negative scenes while being as neutral as possible without being too dull. The narrative of the total of six scenes is described in Table 2.

Table 2. Storyline of all six scenes that had either a positive, neutral or negative valence and took place in either a Classroom or Café setting.

Classroom scenes Café scenes

Positive valence

Emma got the highest grade on a test and received praise from teacher and a classmate

Emma had a good date where they both expressed they “really like each other” and would like to meet again

Neutral valence

Emma was bored during a history lesson Emma asked a stranger (a man) for the time and directions

Negative valence

Emma answered a question incorrectly and was laughed at by teacher and corrected by a classmate

Emma was rejected by her date with the motivation that he was expecting something different

Each scenario was first created through Plotagon and then replicated using a video camera creating six Plotagon scenes and six Live scenes. The Live scenes had the same script as the Plotagon scenes but displayed real actors who played out the same situation as the avatars. The surrounding environment and camera angles were as similar as possible to the Plotagon equivalent of the scene. Both the Plotagon scenes and Live scenes were then edited to look as alike as conceivable; camera angles, appearances and facial expressions were matched and the actors in the Live version recorded the voices of the avatars in the Plotagon version. Each scene was between 22-32 seconds long and the versions did not differ with more than one second.

The characters were the same in all the scenes, including Emma (the protagonist in all scenes), Emma’s classmate, the teacher in the classroom and the man at the café who played either her date or a stranger. The actress who played Emma was an acquaintance of the author of the current study; she was 22 years old at the time and had some previous experience as an actress. The author instead acted as the teacher in the classroom scenes, was 25 years old and had no previous experience acting. Emma’s classmate was a psychology student at Stockholm University and an acquaintance of the author; she had no previous experience as an actress and was 26 years old. The man at the café was recruited through this study’s supervisor and was a doctoral student at

9

Stockholm University. He was 32 years old and had no previous experience as an actor. The avatars were made to look as similar to their human counterparts as possible (see Figure 2 for example pictures) and the human actors recorded the voices of the avatars.

Figure 2. Plotagon and Live versions of the Classroom and Café scenes.

Procedure

The online survey was in Swedish and started with a few background questions regarding age, gender and primary occupation. The participants were then presented with the statements of the scales SIAS-6, SPS-6 and QCAE and asked to assess how closely each statement described them. Before commencing the movie part of the survey, the participants were introduced to the character “Emma” followed by two pictures showing avatar Emma and human Emma. This introduction was made in order to make the participants less confused by who the protagonist was in all the scenes, since Emma was not always the most active character. All twelve scenes were then shown in a random order. After each clip the participants were asked a control question (“What was the movie about?”), followed by the two VAS (measuring Experienced and Perceived Emotional Response) and the three items selected from SES (measuring State Empathy).

Ethics

This study was approved by the Swedish Central Ethical Review Board (DNR: 2018/599-32). The participants were informed of the purpose of the study, that their participation was anonymous, voluntary and that the survey could be terminated at any time.

Data analysis Three separate Repeated Measure Analyses of Variance were carried out with Emotional Response, Perceived Emotional Response (of the protagonist) and State Empathy Score as dependent variables, and Version (Live and Plotagon), Valence

10

(positive, neutral and negative) and Scene (Classroom and Café) as independent variables. Correlational analyses were conducted between the sum of QCAE (M=63.9, SD=8.32), sum of SES (M=32.8, SD=8.04), sum of SIAS/SPS (M=8.75, SD=7.45) and the mean values of Experienced Emotional Response (M=-0.292, SD=7.10) and Perceived Emotional Response (M=-0.399, SD=7.56).

Alpha level was set at α=0.05 for all primary tests of significance. In post-hoc pairwise tests, the alpha level was set to α<0.00833 (0.05/6) in order to adjust for the six pairwise comparisons of interest (the three by two scenes). Power

Calculated with a power table, a number of n>90 participants was estimated to be adequate in order to achieve a within-group effect of d>0.3 and 80% power.

Results

Experienced Emotional Response

A three-way RM-ANOVA with Experienced Emotional Response as a dependent variable, revealed a significant main effect for Scene (F1,101=7.9, p=0.006, η2=0.073), Valence (F2,202=220, p<0.001, η2=0.115) and Version (F1,101=13.14, p<0.001, η2=0.115). Significant interaction effects were found for Scene*Valence (F2,202=68.8, p<0.001, η2=0.405) and Version*Scene*Valence (F2,202=6.9, p=0.001, η2=0.064). As seen in Figures 3 A and B, all positive scenes were experienced as positive, all negative scenes as negative and the neutral scenes differed between versions in whether they were experienced as slightly positive or negative. The Classroom scenes were experienced slightly more positive than the Café scenes, except for in the neutral conditions. Additionally, the Positive Classroom scene (Figure 3A) and Neutral Café scene (Figure 3B) appears to have been experienced as more positive in the Live version compared to the Plotagon version.

Figures 3 A&B. Experienced Emotional Response scored -100 to 100 (and 95% confidence interval) for scenes with positive, neutral and negative valence split by Live and Plotagon version. Figure A (left) displaying Classroom scenes and Figure B (right) Café scenes.

11

A Bonferroni Post Hoc test comparing each Live version with its Plotagon counterpart found the Live versions of the Positive Classroom scene and Neutral Café scene to be experienced significantly more positive (see Table 3). No significant differences between versions were revealed in the Neutral and Negative Classroom scene or the Positive and Negative Café scene. Table 3. Mean values (standard deviations), Post Hoc pairwise tests and Cohen’s d effect sizes of Emotional Response in each scene. Live M (SD) Plotagon M (SD) Post Hoc Cohen’s d Positive Classroom 40.8 (31.2) 32.1 (31.2) p<0.001* d=0.279 Neutral Classroom -13.0 (19.7) -11.4 (15.7) p=0.506 Negative Classroom -31.7 (31.1) -31.0 (28.0) p=0.792 Positive Café 34.7 (31.0) 33.1 (28.0) p=0.529 Neutral Café 17.0 (21.1) 6.85 (16.2) p<0.001* d=0.540 Negative Café -39.6 (33.5) -41.2 (29.5) p=0.542 *Significant after Bonferroni correction Perceived Emotional Response

In a three-way RM-ANOVA with Perceived Emotional Response of the protagonist as a dependent variable, a main effect was found for Version (F1,101=22.66, p<0.001, η2=0.183), Valence (F2,202=919.74, p<0.001, η2=0.901) and Scene (F1,101=14.99, p<0.001, η2=1.129). Interaction effects were found for Version*Scene (F1,101=8.30, p=0.005, η2=0.076), Scene*Valence (F2,202=158.01, p<0.001, η2=0.61) and for Version*Valence*Scene (F2,202=13.88, p<0.001, η2=0.121). As seen in Figures 4 A and B, the positive scenes were indeed perceived as positive, the negative scenes as negative and the neutral scenes as more or less neutral depending on the type of scene. Similarly as in the Experienced Emotional Response, the Classroom scenes were perceived more positively than the Café scenes, except for in the neutral situations. The Perceived Emotional Response additionally appear to have been appraised as more positive in the Live versions of the Positive Classroom scene (Figure 4A) and the Neutral Café scene (Figure 4B).

Figures 4 A&B. Perceived Emotional Response of the protagonist scored -100 to 100 (and 95% confidence interval) for positive, neutral and negative valence according to

12

Live and Plotagon version. Figure A (left) displays the Classroom scenes and Figure B (right) the Café scenes. A Bonferroni Post Hoc test comparing each Live version with its Plotagon equivalent found the Live versions of the Positive Classroom scene and Neutral Café scene to be experienced significantly more positive (see Table 4). In the Negative Café scene, the protagonist’s emotional response was perceived significantly more negative in the Plotagon version. No significant differences between versions were revealed in the Neutral and Negative Classroom scene or the Positive Café scene. Table 4. Mean values (standard deviations), Post Hoc pairwise tests and Cohen’s d effect sizes of Perceived Emotional Response for each scene. Live M (SD) Plotagon M (SD) Post Hoc Cohen’s d Positive Classroom 76.2 (24.4) 64.5 (30.2) p<0.001* d=0.426 Neutral Classroom -23.6 (19.4) -17.7 (16.9) p=0.021 Negative Classroom -59.6 (27.3) -59.2 (26) p=0.900 Positive Café 63.0 (29.6) 60.7 (26.7) p=0.383 Neutral Café 23.2 (24.8) 10.4 (23.5) p<0.001* d=0.530 Negative Café -67.7 (34.1) -75.0 (26.1) p=0.004* d=0.240 *Significant after Bonferroni correction

Empathic response A RM-ANOVA with sum of State Empathy score as a dependent variable, presented a significant main effect of Version (F1,101=15.727, p<0.001, η2=0.135), Scene (F1,101=14.034, p<0.001, η2=0.122) and Valence (F2,202=24.711, p<0.001, η2=0.197). A significant interaction effect was also found for Version*Scene (F1,101=12.156, p<0.001, η2=0.107) and Version*Valence (F2,202=21.692, p<0.001, η2=0.177). Figure 5 A and B reveals State Empathy was appraised higher in the positive and negative conditions than in the neutral conditions. This was found in both versions, however State Empathy was lower in the Plotagon versions of the positive and neutral scenes, but not in the negative scenes where it appears to be similar (Figure 5A) or even higher (Figure 5B). Overall, the Classroom scenes appear to have elicited a slightly higher level of State Empathy compared to the Café scenes.

Figures 5 A&B. State Empathy scores (and 95% confidence interval) for Positive, Neutral and Negative valence, according to Live and Plotagon version. Figure A (left) showing Classroom scenes and Figure B (right) displaying Café scenes.

13

A Bonferroni Post Hoc pairwise comparison revealed the Positive and Neutral Classroom scenes and the Neutral Café scene to provoke significantly more empathy in the Live version than in the Plotagon version, see Table 5. Table 5. Mean values (standard deviations), Post Hoc pairwise tests and Cohen’s d effect sizes of State Empathy score in each scene. Live M (SD) Plotagon M (SD) Post Hoc Cohen’s d Positive Classroom 9.42 (2.40) 8.26 (3.01) p<0.001* d=0.426 Neutral Classroom 8.40 (2.72) 6.95 (3.06) p<0.001* d=0.501 Negative Classroom 8.78 (2.57) 8.75 (2.71) p<0.904 Positive Café 8.79 (2.67) 8.26 (2.67) p<0.030 Neutral Café 7.75 (2.85) 6.80 (2.89) p<0.001* d=0.331 Negative Café 7.87 (2.91) 8.45 (2.50) p<0.018 *Significant after Bonferroni correction

Correlations The correlation between QCAE and SES was positive and significant r=0.319, p<0.001. The correlation between QCAE and SES when only including the positive scenes was r=0.42, p<0.001 and when only including the negative scenes r=0.342, p<0.001. By comparison the correlation between QCAE and SES for the neutral scenes was r=0.148, p=0.137. The correlations between SIAS/SPS and QCAE and SIAS/SPS and SES were small and non-significant (r=0.016, p=0.875 and r=-0.103, p=0.302 respectively). A small but significant negative correlation was found between SIAS/SPS and Emotional Response, r=-0.221, p=0.025. The correlation between SIAS/SPS and Emotional Response when only including the Live scenes was significant (r=-0.246, p=0.013), while when only including the Plotagon scenes it was not (r=-0.110, p=0.270). No significant correlations were found between SIAS/SPS and Perceived Emotional Response (p>0.05).

Discussion The purpose of this study was to investigate whether scenes created through Plotagon and scenes with real humans were equally effective in eliciting an emotional response and empathy for the virtual characters. The relationship between social anxiety and the evaluations of the scenes were also examined. Most of the Plotagon scenes were experienced and perceived slightly less positive compared to the Live versions. The Live version of the Positive Classroom scene and Neutral Café scene were experienced and perceived significantly more positive. The Perceived Emotional Response of the protagonist was significantly more negative in the Plotagon Negative Café scene, however since the objective in this scene was to elicit a negative emotion, this indicates the Plotagon version succeeded better in portraying emotions in this particular scene. Additionally, three scenes (the Positive Café scene, Neutral Classroom scene and Negative Classroom scene) were assessed to produce and portray similar emotions in both versions.

14

Level of state empathy for the human and virtual protagonist revealed conflicting results; in half of the scenes the Live versions produced significantly higher levels of empathy and overall state empathy was lower for the Plotagon version in the positive and neutral scenes. However in the negative Plotagon scenes level of empathy was assessed similarly or even higher than in the Live versions, although the difference was non-significant. This appears to be in line with the previous results; some of the positive and neutral scenes were slightly superior in the Live versions compared to the Plotagon versions, while the negative scenes can be concluded to be equally effective in both versions in eliciting an emotional response, in portraying the emotional response as well producing levels of empathy. Since no correlations were found between SAD and Perceived Emotional Response or SAD and State Empathy, it appears that social anxiety is not a predictor for better emotional understanding of the scenes. However, the results revealed an association between high levels of SAD and evaluating their Experienced Emotional Response more negatively in all the scenes. Additionally, this correlation was significant when only including the Live scenes but not when only including the Plotagon scenes. Firstly, this indicates that individuals with high SAD tended to perceive situations more negatively, which is in line with previous studies reporting individuals with SAD are biased to interpret situations more negatively and as more threatening (Beard & Amir, 2009). Secondly, this suggests the Live scenes were in fact superior in producing these negative emotions for those with high levels of social anxiety.

This does not necessarily mean Plotagon cannot be used for delivering exposure therapy to persons with SAD. Most importantly, the pairwise comparisons of the scenes revealed many of them to be assessed equally, including the negative scenes. Additionally, this study had a non-clinical sample and the scenes were only mildly unpleasant. Scenes depicting more anxiety-generating situations might have a different outcome. Moreover, in studies such as Bouchard et al. (2016) or Morina et al. (2015) where virtual environments were equally effective in producing anxiety as real-life situations, they primarily used HMDs. Since we did not use Plotagon as an immersive tool in this study, avatar identification could be more important. Identifying with a human character is naturally easier than it is with a virtual character, yet this could perhaps be overcome if one created an avatar representing oneself. Additionally, since Plotagon allows its users to customize each scenario, the outcome might be different if individuals with social anxiety were to create a situation representing their social fears.

Limitations and strengths The main limitation of the current study is that the Live and Plotagon versions of the scenes cannot be guaranteed to be identical. The perspective in the Live scenes was slightly further away, the volume was lower, and the surroundings were not identical to the Plotagon scenes. Although the human actors attempted to act similarly to the avatars, they sometimes added actions and movements unconsciously; such as looking at the wristwatch when asked for the time or pointing out the window when explaining the direction of the bus. In general, the human actors tended to move according to the context in order to make the situation more realistic, and these movements could not always be replicated in Plotagon. Likewise, some differences were found between the depictions of emotions. In the positive scenes the human actors tended to smile more

15

(showing teeth), while the avatars could only smile with their mouths closed. One additional difference was found in the negative café scene, where the male actor acted skeptical when rejecting Emma, whereas in the Plotagon version, the male avatar appeared more disgusted. This could have affected the validity of the study. Yet, since this study compared humans with avatars, these differences are not surprising. Human actors will depict a situation differently from avatars, simply because they are human. Since Plotagon portrays low-polygon environments with a low level of realism, it is encouraging that emotional responses and levels of empathy were elicited at all. Additionally, some scenes were evaluated similarly despite the large difference in realism. One additional limitation regards the issue of self-reported emotions. Since the participants assessed their emotional responses and level of empathy subjectively, there is a risk the responses were influenced by for example social desirability and demand characteristics, namely answering what the participants think makes them seem like a good person (Nederhof, 1985) and answering what they think they “should” answer in order to please and help the researchers (Nichols & Maner, 2008). Since this study was anonymous it is hoped that social desirability was not a large problem. However, the participants might have realized the purpose of the study (comparing each Plotagon scene with its Live version), especially since many of them were psychology students. If that was the case, they might have tried to remember what they responded in one version when they assessed the other. One possible solution to this problem could be measuring physiological responses instead, which would be applicable in case this study had measured for example anxiety instead of emotions. Considering all twelve scenes were shown in a random order, and the many different measures, these biases hopefully did not influence the results much. Self-report was nevertheless best suited for a pilot study such as this one. Another possible problem is the separation of emotional response in two different assessments: Experienced Emotional Response and Perceived Emotional Response. There is a possibility the participants did not understand the difference between the two measures, namely the valence of the emotion they experienced themselves, and the valence of the emotion they perceived the protagonist to feel. The reason for this separation was to measure both the difference in how the avatars and humans portrayed the emotions and how the participants in turn experienced these emotions. The results were similar in both measures, except the overall difference of Perceived Emotional Response being assessed as even more positive and more negative than Experienced Emotional Response. This difference is not surprising; understanding the emotions of the characters should be easier than actually experiencing the same emotion, yet it confirms a relationship between the two. Despite the possible confusion that could have arisen because of the separation of these two measures, it resulted in the interesting finding that avatar Emma portrayed the negative emotions “better” than human Emma in the negative Café scene (see Table 4), but that this did not result in a large difference in how the participants experienced this emotion (see Table 3).

Conclusion Although the emotions generated and portrayed in the Plotagon scenes were sometimes significantly different from the scenes with human actors, this study still confirms that

16

people are able to feel positive, neutral and negative emotions when watching avatars interacting. Two scenes differed significantly between versions on all three measures, in two scenes no significant difference was found and in two scenes the versions only differed on one measure. Plotagon scenes are not perceived or experienced as positively as the Live scenes, however in the negative scenes this could be argued to be an advantage. Additionally, participants showed ability to empathize similarly with the virtual characters and the human characters in half of the scenes, more importantly in both negative scenes. This could imply that the human actors were better at producing and portraying positive emotions, while the avatars were equally successful or sometimes even slightly superior in the negative conditions. However, those with high levels of social anxiety tended to experience the Live scenes more negatively, but not the Plotagon scenes. This demonstrates a disadvantage for the Plotagon scenes when it comes to experiencing the portrayed emotions and is most likely due to the low level of realism. Nevertheless, this difference could be overcome thanks to of the customization Plotagon offers. Playing out a scenario one is afraid of, with an avatar similar to oneself or using immersive technology such as HMDs, could bring about a different outcome.

As previously stated, there are many advantages to using a virtual tool to deliver exposure therapy. It might be preferred by patients anxious to undergo real-life exposure, as well as have more practicality for both therapists and patients. Considering people with interaction fears are reluctant to seek treatment, virtual exposure therapy could encourage these individuals to seek help and thus result in an improved quality of life. Further research is needed in order to determine whether scenes created through Plotagon can successfully generate levels of anxiety. Future studies should include a clinical sample and use a virtual reality version of Plotagon where one can interact with and be observed by the avatars from a first-person perspective.

17

References

Abramowitz, J.S., Deacon, B.J., & Whiteside, S.P.H. (2011). Exposure therapy for anxiety: Principles and practice. NY: The Guilford Press.

American Psychiatric Association (2013). Diagnostic and statistical manual of mental disorders: DSM-5.Washington, DC: American Psychiatric Publishing.

Aymerich-Franch, L., Kizilcec, R. F., & Bailenson, J. N. (2014). The relationship between virtual self similarity and social anxiety. Frontiers in Human Neuroscience, 8. doi:10.3389/fnhum.2014.00944

Baron-Cohen, S., Richler, J., Bisarya, D., Gurunathan, N., & Wheelwright, S. (2003). The systemizing quotient: An investigation of adults with Asperger syndrome or high-functioning autism, and normal sex differences. Philosophical Transactions of the Royal Society of London, Series B: Biological Sciences, 358(1430), 361–374. doi: 10.1098/rstb.2002.1206

Beard, C., & Amir, N. (2009). Interpretation in social anxiety: When meaning precedes ambiguity. Cognitive Therapy and Research, 33(4), 406-415. doi:10.1007/s10608-009-9235-0

Blascovich, J., Loomis, J., Beall, A.C., Swinth, K.R., Hoyt, C.L., & Bailenson, J.N. (2002). Immersive virtual environment technology as a methodological tool for social psychology. Psychological Inquiry, 13(2), 103-124. doi:10.1207/S15327965PLI1302_01

Bouchard, S., Dumoulin, S., Robillard, G., Guitard, T., Klinger, É, Forget, H., … Roucaut, F.X. (2017). Virtual reality compared with in vivo exposure in the treatment of social anxiety disorder: A three-arm randomised controlled trial. British Journal of Psychiatry, 210(4), 276-283. doi:10.1192/bjp.bp.116.184234

Bögels, S. M., Alden, L., Beidel, D. C., Clark, L. A., Pine, D. S., Stein, M. B., & Voncken, M. (2010). Social anxiety disorder: Questions and answers for the DSM-V. Depression and Anxiety,27(2), 168-189. doi:10.1002/da.20670

Carlbring, P., Andersson, G., Cuijpers, P., Riper, H., & Hedman-Lagerlöf, E. (2017). Internet-based vs. face-to-face cognitive behavior therapy for psychiatric and somatic disorders: an updated systematic review and meta-analysis. Cognitive Behaviour Therapy, 47(1), 1–18. doi:10.1080/16506073.2017.1401115

Chory-Assad, R.M., & Cicchirillo, V. (2005). Empathy and affective orientation as predictors of identification with television characters. Communication Research Reports, 22(2), 151-156. doi: 10.1080/00036810500130786

Cox, B.J., Clara, I.P., Sareen, J., & Stein, M.B. (2008). The structure of feared social situations among individuals with a lifetime diagnosis of social anxiety disorder in two independent nationally representative mental health surveys. Behaviour Research and Therapy, 46, 477-486. doi:10.1016/j.brat.2008.01.011

Davis, M. H. (1983). Measuring individual differences in empathy: Evidence for a multidimensional approach. Journal of Personality and Social Psychology, 44(1), 113-126. doi:10.1037//0022-3514.44.1.113

Deacon, B. J., & Abramowitz, J. S. (2004). Cognitive and behavioral treatments for anxiety disorders: A review of meta-analytic findings. Journal of Clinical Psychology, 60(4), 429-441. doi:10.1002/jclp.10255

Emmelkamp, P. M. (2005). Technological innovations in clinical assessment and psychotherapy. Psychotherapy and Psychosomatics, 74(6), 336-343. doi:10.1159/000087780

18

Eysenck, S. B., & Eysenck, H. J. (1978). Impulsiveness and venturesomeness: Their position in a dimensional system of personality description. Psychological Reports, 43, 1247–1255.

Felnhofer, A., Kafka, J.X., Hlavacs, H., Beutl, L., Kryspin-Exner, I., & Kothgassner, D. (2018). Meeting others virtually in a day-to-day setting: Investigating social avoidance and prosocial behavior towards avatars and agents. Computers in Human Behavior, 80, 399-406. doi:10.1016/j.chb.2017.11.031

Furmark, T., Tillfors, M., Everz, P.O., Marteinsdottir, I., Gefvert, O., & Fredrikson, M. (1999). Social Phobia in the general population: prevalence and sociodemographic profile. Social Psychiatry and Psychiatric Empidemiology, 34(8), 416-424. doi: 10.1007/s001270050163

Hogan, R. (1969). Development of an empathy scale. Journal of Consulting and Clinical Psychology, 33, 307–316.

Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T., & Stephenson, D. (2015). Loneliness and social isolation as risk factors for mortality. Perspectives on Psychological Science,10(2), 227-237. doi:10.1177/1745691614568352

Isbister, K., & Nass, C. (2000). Consistency of personality in interactive characters: verbal cues, non-verbal cues and user characteristics. International Journal of Human-Computer Studies, 53(10), 251-267. doi:10.1006/ijhc.2000.0368

Kampmann, I. L., Emmelkamp, P. M., Hartanto, D., Brinkman, W., Zijlstra, B. J., & Morina, N. (2016). Exposure to virtual social interactions in the treatment of social anxiety disorder: A randomized controlled trial. Behaviour Research and Therapy, 77, 147-156. doi:10.1016/j.brat.2015.12.016

Klinger, E., Bouchard, S., Légeron, P., Roy, S., Lauer, F., Chemin, I., & Nugues, P. (2005). Virtual reality therapy versus cognitive behavior therapy for social phobia: A preliminary controlled study. CyberPsychology & Behavior, 8(1), 76-88. doi:10.1089/cpb.2005.8.76

Kothgassner, O.D., Kafka, J.X., Rudyk, J., & Beautl, L. (2014). Does social exclusion hurt virtually like it hurts in real life? The role of agency and social presence in the perception and experience of social exclusion. Proceedings of the International Society for Presence Research 2014, 45-56.

Kätsyri, J., Förger, K., Mäkäräinen, M., & Takala, T. (2015). A review of empirical evidence on different uncanny valley hypotheses: Support for perceptual mismatch as one road to the valley of eeriness. Frontiers in Psychology,6(390). doi:10.3389/fpsyg.2015.00390

Lamm, C., Rütgen, M., & Wagner, I.C. (2017). Imaging empathy and prosocial emotions. Neuroscience Letters. doi:10.1016/j.neulet.2017.06.054.

Leite, I., Mccoy, M., Lohani, M., Ullman, D., Salomons, N., Stokes, C., ... Scassellati, B. (2017). Narratives with robots: The impact of interaction context and individual differences on story recall and emotional understanding. Frontiers in Robotics and AI, 4. doi:10.3389/frobt.2017.00029

Mattick, R. P., & Clarke, J. C. (1998). Development and validation of measures of social phobia scrutiny fear and social interaction anxiety. Behaviour Research and Therapy, 36(4), 455–470. doi:10.1016/S0005- 7967(97)10031-6

Masten, C.L., Eisenberger, N.I., Borofsky, L.A., Pfeifer, J.H., McNealy, K., Mazziotta, J.C., & Dapretto, M. (2009). Neural correlates of social exclusion during adolescence: Understanding the distress of peer rejection. Social Cognitive and Affective Neuroscience, 4(2), 143-157. doi:10.1093/scan/nsp007

McNally, R.J. (2007). Mechanisms of exposure therapy: How neuroscience can improve psychological treatments for anxiety disorders. Clinical Psychology Review, 27(6), 750-759. doi:10.1016/j.cpr.2007 .01.003

19

Mori, M. (1970). The uncanny valley. Energy, 7(4), 33-35.

Mori, M., MacDorman, K.F., & Kageki, N. (2012). The uncanny valley [From the field]. IEEE Robotics & Automation Magazine, 19(2), 98-100. doi:10.1109/MRA.2012.2192811

Morina, N., Brinkman, W., Hartanto, D., & Emmelkamp, P. M. (2014). Sense of presence and anxiety during virtual social interactions between a human and virtual humans. PeerJ, 2. doi:10.7717/peerj.337

Morina, N., Brinkman, W., Hartanto, D., Kampmann, I. L., & Emmelkamp, P. M. (2015). Social interactions in virtual reality exposure therapy: A proof-of-concept pilot study. Technology and Health Care, 23(5), 581-589. doi:10.3233/thc-151014

Nass, C., & Moon, Y. (2000). Machines and mindlessness: Social responses to computers. Journal of Social Issues, 56(1), 81-103. doi:10.1111/0022-4537.00153

Nederhof, A.J. (1985). Methods of coping with social desirability bias: a review. European Journal of Social Psychology, 15, 263-280.

Nichols, A.L. & Maner, J.K. (2008). The good-subject effect: Investigating participant demand characteristics. Journal of General Psychology, 135(2), 151-166. doi:0.3200/GENP.135.2.151-166

Nomura, K., & Akai, S. (2012). Empathy with fictional stories: reconsideration of the fantasy scale of the interpersonal reactivity index. Psychological Reports, 101(1), 304-314. doi:10.2466/02.07.09.11.PR0.110.1.304-314

Olfson, M., Guardino, M., Streuning, E., Schneier, F.R., Hellman, F., & Klein, D.F. (2000). Barriers to the treatment of social anxiety. American Journal of Psychiatry, 157(4), 521-527. doi:10.1176/appi.ajp.157.4.521

Peters, L. (2000). Discriminant validity of the Social Phobia and Anxiety Inventory (SPAI), the Social Phobia Scale (SPS) and the Social Interaction Anxiety Scale (SIAS). Behaviour Research and Therapy, 38, 943–950. doi:10.1016/S0005-7967(99)00131-X

Peters, L., Sunderland, M., Andrews, G., Rapee, R. M., & Mattick, R. P. (2012). Development of a short form Social Interaction Anxiety (SIAS) and Social Phobia Scale (SPS) using nonparametric item response theory: The SIAS-6 and the SPS-6. Psychological Assessment, 24(1), 66-76. doi:10.1037/a0024544

Plotagon (2018). Retrieved 09/05/18 https://plotagon.com/about/

Powers, M. B., & Emmelkamp, P. M. (2008). Virtual reality exposure therapy for anxiety disorders: A meta-analysis. Journal of Anxiety Disorders, 22(3), 561-569. doi:10.1016/j.janxdis.2007.04.006

Reniers, R., Corcoran, R., Drake, R., Shryane, N., & Völlm, B. (2011). The QCAE: A questionnaire of cognitive and affective empathy. Journal of Personality Assessment, 93(1), 84-95. doi:10.1080/00223891.2010.528484 Ruscio, A., Brown, T., Chiu, W., Sareen, J., Stein, M., & Kessler, R. (2008). Social fears and social phobia in the USA: Results from the National Comorbidity Survey Replication. Psychological Medicine,38(1), 15-28. doi:10.1017/s0033291707001699

Safren, S. A., Heimberg, R. G., Brown, E. J., & Holle, C. (1996). Quality of life in social phobia. Depression and Anxiety,4(3), 126-133. doi:10.1002/(sici)1520-6394(1996)4:33.3.co;2-e

Shen, L. (2010). On a scale of state empathy during message processing. Western Journal of Communication, 74(5), 504-524. doi:10.1080/10570314.2010.512278

20

Shepherd, R. & Edelmann, R. J., (2005). Reasons for Internet use and Social Anxiety. Personality and Individual Differences, 39(5), 949-958. doi:10.1016/j.paid.2005.04.001

Sioni, S. R., Burleson, M. H., & Bekerian, D. A. (2017). Internet gaming disorder: Social phobia and identifying with your virtual self. Computers in Human Behavior, 71, 11-15. doi:10.1016/j.chb.2017.01.044

Tillfors, M., & Furmark, T. (2007). Social phobia in Swedish university students: Prevalence, subgroups and avoidant behavior. Social Psychiatry and Psychiatric Epidemiology, 42(1), 79–86. doi:10.1007/s00127-006-0143-2

Weschler, L. (2002). Why is this man smiling? Wired. Retrieved 09/05-18 from http://www.wired.com/wired/archive/10.06/face_pr.html Wolpe, J.: Psychotherapy by reciprocal inhibition. Stanford, CA: Stanford University Press, 1958.