Examining the Effects of Gender Transfer in Virtual Reality on Implicit Gender Bias

INTRODUCTION

Gender bias, or showing preference for one gender over the other, is a type of discrimination based on a person’s gender which is prevalent in many aspects of society (Grogan, 2019; Lekchiri et al., 2019; Mengel et al., 2019). Although gender bias can be a threat to both men and women, a systematic literature review revealed that women are more likely to experience gender bias (Blau, 1998). Modern gender bias (e.g., sexism) is characterized by the denial of personal bias and prejudice against women, a general conscious belief in equality of the genders, but unconscious attitudes that foster nonsupport for programs and legislation helpful to women (Keller, R.M. and Galgay, 2010). Survey results have shown that gender bias and its experience are “common to almost all women” (Klonoff & Landrine, 1995). For instance, fewer women hold leadership roles in organizations as these roles are commonly seen to be more suited for men. This is revealed in recent data that shows only 21% of board members, 27% of executives, and 5% of chief executives are women among Fortune 500 firms (Barnes et al., 2020; Myers et al., 2020). Women also receive lower pay and are promoted less often than their male peers (Blau, 1998). This negative impact often leads to increased self-doubt and decreased self-esteem (Dardenne et al., 2007).

Sexual harassment is an overt and hostile form of gender bias (Barthelemy et al., 2016) which can cause anxiety, depression, and post-traumatic stress disorder (Berg, 2006; Landrine et al., 1995). Sexual harassment includes unwelcomed sexual advances, requests for sexual favors, and other verbal or physical conducts of a sexual nature (Terpstra & Baker, 1987). Reports of sexual harassment are also common in many working women (Baugh, 1997), especially in male-dominated field (Dresden et al., 2018). Sexual harassment is rooted in the power structure of society which frequently puts male employers in positions of authority over female employees (McLaughlin et al., 2012). Efforts have been made to address inappropriate sexual harassment behaviors, where an individual would bear an embodied perspective of a woman subject to experience sexual harassment and thereby change their attitudes and behaviors (Neyret et al., 2020). Taking on the role of a victim of sexual harassment in virtual reality (VR) has helped to increase empathy (Ventura et al., 2021) and reduce harm toward a potential victim at a later time (Neyret et al., 2020). While studies have provided supporting evidence that one’s attitudes and behaviors can be changed through virtual sexual harassment scenarios (Neyret et al., 2020; Ventura et al., 2021), the influence on one’s gender bias through embodiment has not been fully examined. We expected that by implementing affective processes, like experiencing sexual harassment scenarios from an embodied perspective in VR, could help modify participants’ gender bias.

Taking an embodied perspective may have an influence one’s gender bias can be rationalized by two theoretical concepts: virtual embodiment in embodied cognition theory and theory of mind. Virtual embodiment is stemmed from theories of embodied cognition (Moghimi et al., 2016). A user controls a co-located avatar to achieve virtual embodiment. Usually, the appearance of the avatar is reflected in a virtual mirror to suggest embodiment. By means of avatars, people take on another person’s perspective, while continuously interacting with other objects in the virtual environment (Pan & Hamilton, 2018; Ventura et al., 2021). Avatars can also help bridge a person in the physical environment with VR through virtual embodiment, which consists of three subcomponents: the sense of self-location, the sense of agency, and the sense of body ownership (Kilteni et al., 2013; Riva et al., 2019). Studies that examined the effect of virtual embodiment on perception suggested that multiple sensory experiences could influence cognition (Barsalou, 2008, 2010; Moghimi et al., 2016). In addition, the theory of mind (ToM), which involves mentalizing (Uta & Frith, 2003), is the ability to think about the mental states in oneself and others, and to use this information to understand what other people know and predict how they would act. VR can be a creative approach to enable individuals “to be in another person’s shoes” in a safe experiential setting (Neyret et al., 2020; Vass et al., 2018). The results from the immersive VR-based targeted ToM interventions confirmed that utilizing immersive VR technologies in the remediation of ToM deficits can be a promising direction in the development of interventions (Vass et al., 2018).

There are several technical approaches to enable virtual embodiment in VR. For instance, through real-time motion capture, an avatar can move in synchrony with a person’s real body movements. These movements can be reflected and seen in a virtual mirror, which further contributes to the perceptual illusion of ownership of the virtual body (González-Franco et al., 2010; Neyret et al., 2020). Taking the perspective of another person in VR can lead to changes in cognitive task performance, and one’s attitudes toward gender, race, age (Peck et al., 2013; Schulze et al., 2019; Seinfeld et al., 2018; Tajadura-Jiménez et al., 2017). Relevant studies where embodiment took place at a gender-related context provided insights for the present study (Jouriles et al., 2009; Lopez et al., 2019; Seinfeld et al., 2018). Jouriles et al. (2009) demonstrated that VR can heighten the realism of sexually threatening role plays and used VR-enhanced role plays to help college women resist sexual attacks. Seinfeld and colleagues (Seinfeld et al., 2018) embodied male domestic violence offenders in the virtual body of a woman who was then subjected to an abusive attack from a male virtual character. It was found that the offenders’ emotion recognition skills improved, especially in their ability to recognize fearful faces of females (Seinfeld et al., 2018). Peck et al. (2018, 2020) found that gender body-swap illusions induced stereotype threats. Lopez’s study embodied male participants in different genders of avatars while following the movements of a virtual Tai Chi teacher in a virtual environment, and they found that the participants’ gender bias increased after embodying a female avatar in contrast to a male avatar (Lopez et al., 2019). These findings suggested that embodying participants in avatars of different genders in a virtual environment where they encountered different sexual harassment scenarios could have effects on gender bias.

Researchers have also explored the immediate and long-term effects of VR exposure (not limited to bias-related studies). The duration of interest ranged from several weeks to 1 year after the VR experience (Anderson et al., 2017; Banakou et al., 2016; Cesa et al., 2013; Herrera et al., 2018; Lindner et al., 2021; Viziano et al., 2019). In most of the cases, the positive effects of VR-based interventions mitigated several weeks after the conclusion of VR exposure (Ahn et al., 2016; Herrera et al., 2018), while some studies reported continued improvement even after discontinued VR exposure (Anderson et al., 2017; Batson et al., 1997; Lindner et al., 2021; Viziano et al., 2019). However, studies on the long-term effects of VR exposure on implicit biases and prototypes were limited, especially implicit gender biases (Banakou et al., 2016; Lopez et al., 2019; Maister et al., 2015; Schulze et al., 2019). Given the limited research investigating the effects of VR exposure on implicit gender biases, we aim to contribute to the understanding of the immediate effects of VR exposure on implicit gender biases.

While previous research provided insights to the present study, it also revealed certain areas that were overlooked. While it has been demonstrated that experiencing virtual sexual harassment scenarios could change one’s attitudes, like empathy (Ventura et al., 2021), the influence of virtual harassment scenarios on implicit gender bias has not been covered. Implicit bias is viewed as an unobservable structure in the mind of an individual (e.g., an association in memory) that drives behaviors in an unconscious manner (Amodio & Mendoza, 2010; Gawronski & Bodenhausen, 2006; Sherman, 1996). Although there might be no significant explicit bias against women, there can be implicit biases, especially in male-dominated field (Dresden et al., 2018). Depending on the research objective, some research focused on identifying and characterizing the mental representation of implicit bias (Gawronski & Bodenhausen, 2006; Sherman, 1996), and other research focused on the role of implicit bias in behavior (Amodio & Devine, 2006; Dovidio et al., 2002; Payne, 2005). Individuals may unconsciously exhibit behaviors (of implicit bias) through tasks they perform. The Implicit Association Test (IAT; e.g., A. Greenwald et al., 1998) is a measurement tool that leverages the theoretical concept of implicit tasks to index the strength and nature of implicit bias (an unobservable construct). The IAT has been used to study implicit social cognition in part because of its ease of implementation, large effect sizes, and relatively good reliability (Cattaneo et al., 2011; Hansen et al., 2019; Lopez et al., 2019; Parker et al., 2018; Wang et al., 2019). The IAT can help assess implicit biases and has been used in several domains including race, gender, weight, and age (Carnes et al., 2015; Peck et al., 2013; Schulze et al., 2019; Wang et al., 2019). The gender IAT is sensitive in detecting underlying implicit social attitudes and stereotypes toward or against genders by measuring automatic concept-attribute associations (Cattaneo et al., 2011).

Assessing implicit biases after embodiment in VR has been examined, particularly in the context of racial bias (Banakou et al., 2016; Maister et al., 2013, 2015; Peck et al., 2013). Peck and colleagues (Peck et al., 2013) demonstrated that the embodiment of light-skinned participants in a dark-skinned avatar significantly reduced implicit racial bias against dark-skinned people. However, in a different study, Groom and colleagues (Groom et al., 2009) embodied light-skinned participants in a dark-skinned avatar during a virtual job interview, which exacerbated implicit biases against Black people. The different findings may imply that the plot of scenario itself might influence implicit bias, together with the embodiment experience in VR. Of the studies that examined the effect of virtual experience on implicit gender bias, the results have been inconclusive (Lopez et al., 2019; Schulze et al., 2019; Starr et al., 2019). It is therefore important to provide more evidence for the influence of the virtual embodiment experience on implicit gender bias.

The overarching goal of this study was to explore whether embodying participants in avatars of different genders during different sexual harassment scenarios would affect implicit gender bias. Participants’ sense of body ownership was measured after the virtual experience to show whether a body ownership illusion was elicited among different groups of participants. Implicit gender bias was assessed before and after the participants were embodied in avatars of different genders and experienced seven levels of sexual harassment scenarios. It was expected that experiencing sexual harassment scenarios while experiencing gender transfer (i.e., being embodied in an opposite-gender avatar) in VR would affect participants’ implicit gender bias. The long-term goal is to improve awareness of implicit gender bias through VR, with the ultimate vision of increasing implicit biases awareness for narrowing gender gap.

METHODS

Experimental Design

The experiment was a between-subjects 2 × 2 factorial design (two independent variables: participant gender and gender transfer) that yielded four distinct groups to which participant were evenly assigned, constrained to 10 of each gender in their respective groups (Table 1). Since sexual harassment can be a matter of perception and interpretation, different levels of sexual harassment (from mild to severe) were implemented. Throne (1995) posited that these scaling of various sexual harassment behaviors provided the ability to assess individuals’ judgement of the severity of different acts of sexual harassment. Table 2 lists the seven levels of scenarios: no harassing behavior, visual, verbal, touching, sexual touching, implied sexual request, and sexual request (Throne, 1995). Sample scenarios from Throne’s study are described in text, along with the scenarios used in the present study. Scenarios used in the present study were not identical to the those in Throne’s original work due to differences in the nature of scenarios presentation in text and in virtual environment. All participants experienced seven levels of scenarios that illustrated possible gender bias situations. Each level of scenario was repeated (i.e., two repetitions per level) to help increase the magnitude of the exposure effect (Melo et al., 2017, 2018; Van Dam & Stephens, 2018). For each group, the offender avatar had the opposite gender from the participant avatar. For instance, if a participant embodied in a female avatar, then the offender avatar was a male.

OPEN IN VIEWER

TABLE 1:

Four Groups of Participants With 10 Participants in Each Group

FEF: Female participant embodied in female avatar experiencing sexual harassment.

FEM: Female participant embodied in male avatar experiencing sexual harassment.

MEM: Male participant embodied in male avatar experiencing sexual harassment.

MEF: Male participant embodied in female avatar experiencing sexual harassment.

OPEN IN VIEWER

TABLE 2:

Sample Scenarios From Seven Levels of Sexual Harassment Scenarios in Throne’s Study and in VR. Virtual Scenarios Adapted From Sample Scenarios Due to Differences in the Nature of Scenarios Presentation in Text and in Virtual Environment

Level	Sample Scenarios in Throne’s Study	Virtual Scenarios
None	Sandy Phelps told Rick Matthews that she had enjoyed studying with him for class, and that she hoped they would have other classes together.	In a classroom, A said to B, “I had enjoyed studying with you for class. I hope we would have other classes together.”
Visual	When Amy Tomkins went up after class to ask her professor, Dr. Greg Billings, a question, she saw small drawings on his notes. She noticed one of the pictures was of a naked woman and he had written her name underneath the drawing.	In one of the studios in the gym, B walked toward A with B’s upper body naked, and said, “How do you like my body?”
Verbal	Dr. Mary Hart approached one of her students, Jason Carter, before class and asked if he had heard any good jokes lately. When Jason said no, Dr. Hart told him a dirty joke she had just heard.	In a workplace before a meeting A asked B, “Have you heard any good jokes lately. I have a good one. Why did the ketchup blush? Because he saw the salad dressing.”
Touching	When Mark Henderson introduced himself to a classmate, Tiffany Mattingly, they shook hands. Before Mark let go of her hand, he pulled her towards him and gave her a hug.	Before a meeting about an upcoming project, B walked toward A and gave A a hug.
Sexual touching	Jenny Oliver was helping her classmate Derek Pierce with their class assignment. As Jenny was explaining the problem, Derek got up and kissed her on the lips.	A and B were in a classroom. A walked toward B and tried to kiss B.
Implied sexual request	Cassie Barker is working with her supervisor, Mr. Burt Owens, on a joint project for work. Because it is taking more time than they expected, they have been meeting together after work to discuss the project. At their last meeting, Mr. Owens said, “You know Cassie, we’ve been meeting for three weeks now, and I can’t help but notice how nice you look in those skirts you wear to work. I was wondering why you don’t wear them when we meet together. I thought you knew how much they turned me on.”	A and B are in an office before a meeting. A said to B, “You know, we’ve been meeting for three weeks now, and I can’t help but notice how nice you look in those suits you wear to work. I was wondering why you don’t wear them when we meet together. I thought you knew how much they turned me on.”
Sexual request	Sandra Randolph has been getting together with her supervisor Mr. Ray Elliot for the past few weeks to go over some paperwork left over from her last assignment. Last time they met, Mr. Elliot told Sandra that he would love to sleep with her.	In a classroom, A walked toward B and said to B, “We have spent a lot of time doing homework. I really like you. I’d love to sleep with you.”
Physical assault	Ryan Duvall asked Marian Hackett, a classmate, to help him with some homework after class. After everyone else had left, Ryan pushed Marian against the wall and proceeded to remove her shirt.	This level (i.e., the most severe level) was omitted to decrease participants’ emotional risk.

Experiment Procedure and Task

Pre-experiment Assessment

Upon informed consent, the participant completed a self-reported demographic questionnaire on paper, which included questions on gender, age, and experiences using VR technology. Next, the participant completed a gender IAT (Greenwald et al., 2003, 2009; Nosek et al., 2005) on the lab computer, which produced the baseline IAT score that was denoted as preIAT. The gender IAT evaluated implicit biases by requiring people to quickly categorize faces (female or male) and words (positive or negative) into groups. The scores of the gender IAT test were calculated based on the differences in speed and accuracy in associating faces of different genders and words with positive and negative meanings. A positive IAT score indicated a preference for male and a negative value of IAT score indicated a preference for female. A larger absolute IAT score indicated greater gender bias. The participant then completed the Sensitivity Towards Sexual Harassment Questionnaire (SQ) on paper (Throne, 1995), which assessed the participant’s baseline sensitivity toward sexual harassment. The SQ included 15 items where each item was a written description of a potentially harassing scenario involving two individuals (one female and one male). The SQ was administered to understand whether the participants felt a scenario was appropriate or not. While the sensitivity toward sexual harassment was measured in the overall study and not immediately related to statistical analysis of IAT, the measure was part of the large project and was mentioned to make the study more complete and replicable. Next, 28 reflective motion tracking markers were attached onto the participant’s bony landmarks for capturing the real-time movement of the participant for driving the motion of the participant avatar.

Experimental Experience

Upon entering the virtual environment, the participant first appeared in a lobby area and stood in front of a mirror, which was referred to as the “mirror lobby.” The participant was embodied inside the participant avatar by assuming the same position and movement of the avatar. The participant was able to see the virtual arms and legs by looking at the virtual body segments directly as well as into the mirror. The mirror reflected the entire appearance of the participant avatar, which aimed to help the participant to recognize the gender of the avatar in which they were embodied. A 3-minute priming period was administered during which the participant explored the mirror lobby, observed the avatar reflected in the mirror, and described the appearance of the avatar and the surrounding environment, with the goal of acclimatizing the participant to the avatar and the virtual environment (Figure 2).OPEN IN VIEWER

Figure 2.

The 3D models of the virtual characters in which participants embodied: (a) female avatar front view, (b) female avatar left view, (c) male avatar front view, (d) male avatar left view.

After the priming period and before the participant entered the virtual harassing scenarios, the participant was verbally informed that there would be no intimate relationship between the participant avatar and the offender avatar. The experiment protocol about the context of the scenarios was read to the participant, by the experimenter, before entering the scenarios to make sure no participants were confused about the context. During the sexual harassment experience, the offender avatar was programmed to exhibit specific harassment behaviors while the participant avatar did not have to make any responses. For example, in the “sexual request” scenario, the participant avatar would appear standing in a classroom and the offender avatar walked toward the participant avatar and said, “We have spent a lot of time doing homework. I really like you. I’d love to sleep with you.” (Figure 3). The scenarios were randomized with constraints where the same scenario was not seen back-to-back and the scenarios were not in an increasing or decreasing levels of harassment as defined by Throne. There was a two-minute rest break between each scenario. During the rest break, the participant returned to the mirror lobby, which was intended to help increase the participant’s virtual body ownership while the next harassing scenario was loaded. The experiment task was completed when the participant experienced all seven harassing scenarios twice. The participant verbally rated the perceived appropriateness of the various harassment scenarios using the same scale as the SQ after experiencing each virtual sexual harassment scenario as soon as the participant returned to the mirror lobby. All scenarios lasted approximately 30 seconds. The participant was transferred to mirror lobby as soon as the scenario ended, and the way of transmission was fading to black and fading back to the next virtual scene.OPEN IN VIEWER

Figure 3.

Examples of harassing scenarios in VR: (a) level “no harassing behavior” from first-person perspective view, (b) level “no harassing behavior” from third-person perspective view, (c) level “touching” from first-person perspective view, (d) level “touching” from third-person perspective view.

Post-Experiment Assessment

After the experimental experience, the participant completed the gender IAT for acquiring the post-experiment IAT score (postIAT). The participant then completed the SQ again to measure post-experiment sensitivity toward sexual harassment after the virtual experience. The participant also answered two questions on the post-experiment questionnaire regarding the sense of body ownership (Dewez et al., 2019; Peck et al., 2013), denoted as postQ (Table 3). The postQ items were rated on a Likert-like scale from 1 (strongly disagree) to 5 (strongly agree).

OPEN IN VIEWER

TABLE 3:

Sense of Body Ownership Items From the Post-Experiment Questionnaire

Item	Statement
1	I felt as if the virtual body I saw when looking in the mirror was my own body.
2	I felt as if the virtual body I saw when looking in the mirror was another person.

RESULTS

DISCUSSION

This research was the first step to understanding the potential changes in implicit gender biases through embodiment and gender transfer using avatars in VR, with the long-term objective of increasing one’s awareness in implicit gender biases. Overall, the results show that experiencing sexual harassment scenarios with gender transfer (e.g., embodiment in an opposite-gender avatar) in VR can immediately reduce one’s implicit gender biases. The effect on implicit gender bias was most significant for the MEF group among the four groups.

The primary research objective was to investigate the effect of gender transfer in VR on implicit gender bias. The results revealed that both female and male participants who experienced gender transfer, namely the FEM and MEF groups, reported lessening in implicit gender bias immediately after the VR experience. In other words, gender transfer through body ownership of an avatar has helped shift the preference for the gender in which the participant embodied. Although the participants’ physical appearances were different from the avatars, the participants had developed a sense of body ownership and thereby shifted their implicit bias in favor of their avatar’s gender. The results shown that VR technologies can be a promising tool to enable embodied cognition through virtual embodiment where the cognition was influenced by the virtual experience. This finding is aligned with Peck and colleagues (2013) who demonstrated that the embodiment of light-skinned participants in a dark-skinned avatar significantly reduced implicit racial bias against dark-skinned people (Peck et al., 2013). While Peck and colleagues’ study focused on implicit racial bias and the present study investigated implicit gender bias, the outcomes from both studies suggested a tendency to increase bias in favor of the characteristics of the avatar in which individuals were embodied. The gender transfer effect further confirmed that VR-based interventions may support the remediation of ToM deficits (Vass et al., 2018).

There are many other forms of implicit biases and stereotypes in society, such as racism and ageism. These implicit biases and stereotypes can be difficult to avoid because they are what humans unconsciously believe and feel (Shulman, 2017). Yet, these are important societal challenges that are constantly being evaluated (Behrend et al., 2012; Gonzalez-Liencres et al., 2020; Schulze et al., 2019) and needing our attention. Although the present study has focused on implicit gender bias, the implications and methodology may help inform and devise strategies to mitigate and address prejudice and stereotypes, and acknowledge valid appraisal of individuals’ abilities instead of stereotypic decisions (Malos, 2015). As the malleability of implicit biases and stereotypes was shown in previous research (Peck et al., 2013; Schulze et al., 2019; Seinfeld et al., 2018; Tajadura-Jiménez et al., 2017), our results also demonstrated that implicit gender bias could be changed through affective processes such as embodiment in a specific avatar in a virtual environment. However, the present study employed primarily visual and auditory feedback, and no haptic feedback was provided to simulate touching. While the harassing scenarios might have been more impactful if the participants felt that they had been touched by a virtual harasser, the literature has demonstrated visual feedback alone provided sufficient cues for users to sense ownership of their avatars and to perceive body transfer (Ahn et al., 2016). Findings from the present study provided a promising approach in applying VR to reduce individuals’ implicit bias through modifying the appearance, movement, and actions of an avatar in addition to the depicted scenarios in VR.

Of the two groups that yielded significant change in implicit gender biases, namely the FEM and the MEF groups, the magnitude of change in implicit gender bias of the MEF group was greater than the FEM group, as indicated in the statistical interaction between gender transfer and participant gender. This may be due to that male participants were more likely to be influenced by the appearance of avatars than females. This phenomenon has been observed by Mello and colleagues that men were more susceptible to gender transfer illusion, for instance, bearing the perspective of an opposite-sex avatar (Mello et al., 2022). Another plausible explanation could be due to the fact that female victims make up a larger proportion in the reported sexual harassments cases, and women are more likely to be sexually harassed than men (Baugh, 1997; Dresden et al., 2018). Thus, at the subconscious level, embodying a female avatar while experiencing sexual harassment scenarios produced greater cognitive agreement in a harassment scenario. As modern sexism is characterized by the denial of personal bias and prejudice against women (Keller, R.M. and Galgay, 2010) and reports of sexual harassment were mostly common with women (Baugh, 1997), especially in male-dominated fields (Dresden et al., 2018), the virtual sexual harassment system could be considered as a potential tool of increasing implicit biases awareness for narrowing gender gap.

In terms of the two other groups, namely the FEF and MEM groups that shared the same gender as the avatars, they did not have statistically significant changes in implicit gender biases after the virtual harassment experience. This finding suggested that embodying in an avatar of the same gender while experiencing virtual harassment scenarios was unable to change implicit gender bias toward the other gender, which further strengthened the argument that the gender of an avatar has the potential to influence implicit gender biases. While the FEF group did not show a statistically significant change in the IAT score, a nominal increase in the IAT score was recorded in the FEF group suggesting a seemingly bias toward males. The statistical insignificance (p = .06) was likely due to the lack of statistical power, and therefore future work with a larger sample size is needed to further examine the relationship. Although the sample size (n = 40) was determined based on the power analysis (power = .9, alpha = .05, effect size = 1.45) from our initial work (Wu & Chen, 2021), the present study was a between-subject experiment with 10 participants in each group. The results should be interpreted within the context of the experiment.

Taken together, implicit gender biases can be modified, at least temporarily, by embodiment in specific avatars in a virtual environment. In general, there was a tendency for individuals to have increased bias in favor of the gender of the avatar in which they embodied. For the scope of this study, this finding supported the expectation that embodiment in an opposite-gender avatar was able to improve implicit gender bias, and this is in line with the embodied cognition theory suggesting that cognition could be influenced through embodiment in VR (Barsalou, 2008, 2010; Moghimi et al., 2016). Immersive virtual environment can be a promising tool to enable individuals “to be in another person’s shoes” toward the improvement of implicit bias which further expanded the findings utilizing VR-based interventions in the remediation of ToM (Vass et al., 2018). As for the virtual scenarios implemented in VR, there was a primary study conducted by the research team whose results supported that experiencing sexual harassment scenarios in VR could help increase the awareness of sexual harassment. The current research provided promising evidence that virtual environments can be used as a potential training tool to improve implicit gender bias and increase the awareness of sexual harassment.

CONCLUSION

Embodying in avatars of different genders in an immersive virtual environment and experiencing different degrees of sexually harassing scenarios, individuals showed changes in implicit gender bias. The findings revealed that implicit gender biases could be modified via gender transfer, at least temporarily. Specifically, embodying people in avatars of different genders to experience sexual harassment scenarios resulted in an improvement in implicit gender bias. The improvement of implicit gender bias was most prominent in male participants who were embodied as a female avatar, which suggested that having men experience the perspective of women could help reduce men’s implicit gender bias. It was also observed that there was a tendency for individuals to have increased bias for the gender of the avatar in which they embodied. The results suggest that gender transfer in VR has the potential to help reduce implicit gender bias and possibly narrow the gender gap.

KEY POINTS

The results indicated that embodying participants in avatars of different genders in a virtual environment where they encounter different sexual harassment scenarios could help reduce implicit gender bias.