by Christopher R. Agnew, Jonathan J. Carter, and Ledina Imami
Published onAug 01, 2022
Forming Meaningful Connections Between Strangers in Virtual Reality: Comparing Affiliative Outcomes by Interaction Modality
·
Abstract
Video Abstract
At present the extent to which close, meaningful connections can be formed in virtual reality (VR) between strangers is unknown, as is how the development of such connections compares to those formed in other established communication modalities. The present preregistered experiment examined the induction of closeness in VR using a validated social interaction task and how it compares with such induction in other interaction modalities with respect to affiliative outcomes (e.g., closeness toward and perceived responsiveness of interaction partner). Based on previous theory and research, we hypothesized that affiliative outcomes would be higher in interaction modalities featuring more versus less sensory input but would not differ significantly between the most sensory-rich modalities (video and VR). Two hundred seventy-two previously unacquainted undergraduate students were randomly assigned to interact with a partner (forming 136 dyads) using the “fast friends” procedure (Aron et al., 1997) via either text, audio chat, video chat, or VR. Results were consistent with hypotheses, suggesting that closeness and related outcomes can be generated and experienced in VR akin to levels experienced via other sensory-rich computer-mediated modalities. Exploratory analyses of possible individual difference moderator variables (Big Five personality dimensions and attachment orientations) yielded largely nonsignificant findings, suggesting the robustness of the obtained findings across individual differences.
Keywords: close relationships, closeness, affiliative outcomes, virtual reality, computer-mediated communication
Acknowledgments: This research would not have been possible without the efforts of an incredible team of research assistants, including Zoey Cao, Aneri Deshpande, Justin Ho, Kevin Hu, Jenna Kaminski, Anna Keeperman, Adam Lechowicz, Yuchen Li, Emily Myers, Estevan Rodriguez, Chris Song, Aria Sturmer, Ariel Velasco, Alex Yang, and Stella Yao. The authors also acknowledge helpful assistance with initial study setup provided by Elizabeth Crail, Juwon Lee, and Susan Sprecher, and statistical guidance from Ximena Arriaga, and Sean Lane.
Funding: This research was supported by a grant from Facebook (now Meta), for which the authors are thankful. Facebook had no role in this research other than providing financial support.
Disclosures: Beyond review and approval of study procedures by Purdue University’s institutional review board, we complied with all American Psychological Association ethical standards in the treatment of study participants. The authors also have no conflicts of interests to disclose.
Data Availability: The measures used to conduct the research, the data collected, and the code used in analyses are available to any researcher for purposes of reproducing the results or replicating the procedure. It can all be found here: https://osf.io/xsh4b/.
Open Access License: This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC-BY- NC-ND). This license permits copying and redistributing the work in any medium or format for noncommercial use provided the original authors and source are credited and a link to the license is included in attribution. No derivative works are permitted under this license.
Contact Information: Correspondence concerning this article should be addressed to Christopher R. Agnew, Department of Psychological Sciences, Purdue University, 703 Third Street, West Lafayette, IN 47907-2081, United States. Email: [email protected]
An increasing number of people are becoming acquainted with previously unknown others in virtual reality (VR) settings. Reduced barriers in recent years to using VR (e.g., less expensive and easier-to-use equipment, greater availability and selection of VR apps) has yielded significantly more users, with over a third of Americans now estimated to be using VR products at least once a month and more than 100 million additional Americans projected to become first-time VR users in 2022 (Bidar & Patterson, 2022). Although VR was initially dominated by gaming applications, there are an increasing number of apps that allow for open communication among previously unacquainted VR participants in nongaming venues, providing an opportunity for relationship development. With the rising use of VR comes rising interest among psychological researchers in understanding interpersonal relationship dynamics in VR settings. Currently, the extent to which close connections can be formed in VR between strangers is unknown as is how such close connections compare to those developed via other established communication modalities.
VR applications have proliferated in recent years and now include an array of nongaming virtual settings in which users find themselves interacting with previously unknown others. As one example, novel interaction opportunities occur within VR business apps such as Horizon Workrooms, in which coworkers can meet virtually and be afforded many of the meeting amenities previously available only via in-person gatherings (e.g., moving seat locations during a meeting, standing and walking to a whiteboard; Corazza, 2022). Participants in such meetings include new colleagues, many of whom are now onboarded at companies without having stepped foot in a physical workplace, despite the established importance of a positive onboarding experience for employee retention (Sibisi & Kappers, 2022). Instead, technological developments, including within the VR space, have allowed workers to initially meet and get to know new colleagues virtually. As such virtual initial social interaction opportunities increase, it is important to understand the extent to which basic relational dynamics that have been found to occur in person and/or in other interaction modalities operate similarly in VR. This is all the more critical given the known linkages among workplace relational dynamics, subjective well-being, and productivity (Eby & Allen, 2012).
Past research has examined how reciprocal interpersonal communication promotes affiliative outcomes (e.g., Sprecher et al., 2013) and how technology (including in the form of computer-mediated communication, or CMC) can be used to induce closeness among previously unacquainted people (Antheunis et al., 2007; Sprecher, 2021). However, no previous research has examined the generation of closeness in VR and no research to date has studied how initial social interactions that occur in VR compare to other interaction modalities so as to establish their relative effects on the induction of various affiliative outcomes. The present preregistered experiment fills these gaps by examining how the induction of closeness through a validated social interaction task affects affiliative outcomes among unacquainted participants interacting via text, audio chat, video chat, or VR.
Social interactions, particularly those that involve mutual and reciprocal self-disclosure (Collins & Miller, 1994; Willems et al., 2020), have been shown to increase positive affiliative outcomes (such as closeness and liking) among previously unacquainted individuals (Sprecher et al., 2013). Seminal work by Aron et al. (1997) introduced a methodological procedure for generating closeness between previously unacquainted people meeting in person within a laboratory setting. The procedure, sometimes referred to as “fast friends,” involves unacquainted participants discussing sets of questions designed to elicit increasingly deep self-disclosure. This procedure and its derivatives (e.g., Sedikides et al., 1999) have enabled subsequent researchers to manipulate experimentally the manner in which study participants relate to each other and the specific contexts in which those relationships develop (Sprecher, 2020).
The modality in which zero-acquaintance social interaction occurs has been found to influence social behaviors and outcomes, including nonverbal behaviors (Croes et al., 2019) and several affiliative outcomes (Croes et al., 2016; Sprecher, 2014). Explanations for modality-based variation in affiliative outcomes include an emphasis on the greater “richness” (vs. “leanness”) of nonverbal cues afforded by different communication modalities (Ramirez & Burgoon, 2004; Sprecher, 2014). Differences in affiliative outcomes between strangers have been found to be particularly substantial when comparing text-based interactions to interactions taking place via a richer sensory modality (Sprecher, 2014). As communication modalities increase in sensory richness, modality effects appear to attenuate. Indeed, research has demonstrated no significant differences in affiliative outcomes obtained between strangers involved in face-to-face (FTF) versus video-based (e.g., Skype video) interactions (e.g., Sprecher, 2021).
VR provides a sensory-rich environment, akin in many respects to that afforded by video-based interactions. Both include synchronous audio between interactants, and both involve a dynamic visual representation of interactants in real time. However, the visual representations differ with respect to both realism and ability to display nuanced nonverbal facial information. Video-based interactions involve synchronous moving images of interactants as captured by device-embedded cameras and delivered via computer-enabled device screens. VR interactions involve synchronous moving images of an avatar representation of an interactant (which may or may not attempt to mirror the interactant’s actual physical appearance) within a virtual space. VR interactions featuring avatars of humanoid form most closely parallel video-based interactions and were the focus of the current research. We posited that such interactions are particularly likely to cue established FTF interaction norms associated with both giving and receiving self-disclosures with another person in real time, leading interactants in the video and VR modalities to evidence similarly high levels of affiliative outcomes. Although the two modes provide differing levels of overall experiential immersion, both engage the same primary senses (hearing and seeing) and past research has demonstrated that the relative immersive quality of an interaction medium does not result in differing levels of reported closeness between interactants (Herrera et al., 2018).
Accordingly, we hypothesized that affiliative outcomes would be higher among dyads in interaction modalities featuring richer (audio, video, and VR) versus leaner (text) sensory input (Hypothesis 1), and higher in modalities featuring a visual representation of interactants (video and VR) than in modalities without such sensory input (audio and text; Hypothesis 2). Consistent with past findings of no differences in affiliative outcomes between those interacting FTF versus via video chat, we further hypothesized that there would not be significant differences in affiliative outcomes among dyads interacting in video versus VR modalities (Hypothesis 3).
In an exploratory vein, we also were interested in examining the possible moderating role of individual differences in affiliative outcomes by interaction modality, focusing on general personality dimensions (i.e., the “Big Five”; John & Srivastava, 1999) and attachment orientations (Mikulincer & Shaver, 2016). To our knowledge, no past research has examined possible differences in affiliative outcomes associated with individual differences by interaction modality. Thus, we were unable to ground our exploration in past findings. Previous research has found that the overarching Big Five personality traits predict relationship quality and ratings of closeness (Berry et al., 2000), but we did not believe they would do so differentially by modality. Regarding the interpersonal attachment dimensions of avoidance and anxiety, past research has revealed that these key interpersonal orientations can influence interaction dynamics and outcomes (e.g., Feeney, 2002; Fraley & Aron, 2004). We suspected that higher levels of attachment avoidance might dampen overall mean levels of affiliative variables, but we advanced no specific hypotheses with respect to any differential effects of either attachment orientation by modality. Thus, we tested for possible moderation of study hypotheses by these individual differences purely in an exploratory vein.
Method
Design
The present cross-sectional study featured one between-subjects independent variable (modality), with four levels. Participants were randomly assigned to one of the four modality conditions, where they were instructed to participate in a structured interaction with a previously unacquainted other via (a) text, (b) audio chat, (c) video chat, or (d) VR. Random assignment to condition was based on numbers generated from Research Randomizer (Urbaniak & Plous, 2013). Our hypothesis preregistration in Open Science Framework (OSF) can be found here: https://dx.doi.org/10.17605/OSF.IO/84RUQ.
Participants
A power analysis (conducted prior to data collection using G*Power; Faul et al., 2007) was computed based on the effect sizes reported in related past research (Sprecher, 2014; Table 1). It indicated that data from 212 individuals (forming 106 dyads) would be needed to achieve .90 power and 280 individuals (forming 140 dyads) would be needed to achieve .95 power. Inclusion criteria included those at least 18 years of age, fluent in conversational English, having no prior VR experience, and no prior acquaintance with their interaction partner as reported in the postinteraction survey. Exclusion criteria included sensitivity to flashing light or motion, a recent injury to the eyes, face, neck, or arms that may prevent the comfortable use of VR hardware or software, and/or a current diagnosis of epilepsy, dementia, or other neurological diseases that may prevent the safe use of VR technologies. Data meeting inclusion criteria and our predetermined data cleaning approach were collected from 274 individuals (forming 137 dyads), thus ensuring .90–.95 power.
Table 1
Means and Standard Deviations, Overall and by Experimental Condition
Study variable
Interaction modality
Overall (N=274) M (SD)
Text (N=72) M(SD)
Audio (N=70) M(SD)
Video (N=60) M(SD)
VR (N=72) M(SD)
Affiliative outcomes
Closeness
3.79 (1.18)
3.35 (1.24)
3.86 (1.12)
4.04 (1.15)
3.96 (1.10)
Liking
5.60 (1.17)
5.18 (1.02)
5.67 (1.03)
5.88 (1.01)
5.70 (1.00)
Responsiveness
5.84 (0.97)
5.42 (1.18)
5.92 (0.87)
6.10 (0.82)
5.95 (0.82)
Enjoyment
5.27 (1.01)
4.80 (1.09)
5.43 (0.83)
5.40 (0.98)
5.47 (0.98)
Potential moderators
Attachment anxiety
4.45 (1.28)
Attachment avoidance
5.22 (1.05)
Open-mindedness
3.71 (0.64)
Conscientiousness
3.61 (0.66)
Extraversion
3.42 (0.71)
Agreeableness
3.82 (0.56)
Negative emotionality
2.80 (0.77)
Control variables
Modality experience
1.83 (1.19)
2.37 (1.34)
2.85 (1.27)
1.24 (0.78)
Relationship status
1.71 (0.94)
Note. VR = virtual reality; N = number of individuals.
Participants (136 female and 138 male) were undergraduate students at a large U.S. state university. They were not paid for their participation; rather, they received research credit toward fulfillment of requirements in a course in which they were enrolled. With respect to race/ethnicity, participants (age M = 19.07, SD = 2.04) were mostly White (194 White or 70.8%, 50 Asian, 14 Hispanic, 8 Black, and 8 Mixed Race), reflecting the racial/ethnic composition of the university student population. Female–female (n = 34), male–male (n = 36), and female–male (n = 67) dyads were formed spontaneously based on participants’ individual scheduling selection. We did not advance any hypotheses regarding dyad sex composition. Each participant took part in one laboratory session in a campus building. The implemented procedures, including inclusion and exclusion criteria, were reviewed and approved by the university’s institutional review board.
Procedure
Participants used an online website (administered by SONA Systems, https://www.sona-systems.com/) to sign up for a laboratory session and were subsequently randomly assigned to interact with a partner using one of four modalities: text chat, audio chat, video chat, or VR. Data were collected from two participants (interacting with one another via the assigned modality) during each session. Before a scheduled session, one participant was instructed to arrive at a second-floor waiting room in a campus building and the other participant was instructed to arrive at a third-floor waiting room. The purpose of this was to help ensure that the members of each dyad had zero acquaintance with each other prior to the study; this was subsequently confirmed via direct questioning in the postinteraction questionnaire. To prevent transmission of COVID-19, all research assistants and participants were required to wear a face mask throughout each laboratory session and to maintain at least 6 feet of distance between one another during the session. Two research assistants, one on each floor of the building, met and escorted a participant to experimental rooms on their respective floors, where the participants were seated at a desk featuring a laptop. They were then presented with a consent form, given time to read and ask any questions about it, and instructed to sign if they agreed to participate.
Participants in the text, audio, and video conditions were then instructed on how to use the capabilities of Skype on the laptop to communicate with the other participant, while participants in the VR condition were instructed on how to use Oculus Quest VR headsets to interact in a custom virtual room (designed to mirror in appearance the physical lab room) created in the Rec Room app. We selected Rec Room to use for this study because it is a massively multiplayer online game that enables the creation and customization of user-generated content. It is also a popular app and an excellent illustration of VR’s potential for facilitating social connection. The laptop and headset in both lab rooms were thoroughly disinfected with alcohol wipes both before and after each session.
After being trained on how to use their respective communication modalities, participants engaged in three 5-min discussions, with sets of questions provided for each discussion block. The question sets were intended to elicit increasingly deep self-disclosure (following the “fast friends” procedure; Aron et al., 1997). For example, the question “How did you celebrate last Halloween?” was included in the first set, “Would you like to be famous? If so, in what way?” was included in the second set, and “What is one of your deepest fears?” was included in the third set. For each set of questions, both participants were instructed to take turns answering as many questions as they could during the 5-min discussion, as turn-taking self-disclosure (vs. sequential self-disclosure) has been found to yield higher affiliative outcomes (Sprecher & Treger, 2015). If a participant felt uncomfortable with any of the questions in the disclosure task, they had the option to “pass” and not answer those questions. The question sets were presented to participants in the text, audio, and video conditions on large boards set up on an easel in the experimental rooms directly in front of the lab table at which they were seated. In a similar manner, the question sets appeared on the wall in front of participants in the VR condition. Research assistants moved the question sets onto the “wall” of the virtual room and were visible to participants in avatar form during question set transitions (as they were in physical form in the other conditions). All participants in the VR condition were represented visually by the same basic human avatar (see Figure 1).
Figure 1
Virtual Experimental Room and Participant Avatar
Although participants were left alone in their respective lab rooms during their interaction, the research assistants reentered the rooms after each 5-min discussion to instruct the participants to advance to the next set of questions or to stop the task when time was up (after the third set). Participants did not meet or interact with the other member of their dyad after they completed the disclosure task. Following the interaction, participants used the laptop to complete an online survey that contained basic demographic questions and multiple affiliative and personality measures. Participants accessed the online survey through a secure webpage hosted by Qualtrics.com, and all collected data were stored in a password-secured online database. All participants were fully debriefed at the end of the study. Each participant was able to complete the study in less than 1 hr.
Measures
After the interaction, the following measures were collected from each participant (see Table 1, for descriptive statistics, overall and by condition as appropriate):
Affiliative Outcomes
Measures of affiliative outcomes largely paralleled those used by Sprecher (2014). Closeness toward interaction partner was assessed with two measures. Participants answered the question “How close do you feel toward your interaction partner?” using a Likert scale ranging from 1 (not at all) to 7 (a great deal). Participants also completed a modified version of the Inclusion of Other in the Self Scale (IOS; Aron et al., 1992). The IOS asks respondents to select a picture from a series of seven increasingly overlapping circles that correspond to different degrees of perceived closeness with their interaction partner. These two measures were significantly correlated (α = .79; Spearman–Brown r = .64; Eisinga et al., 2013) and were combined to serve as an index of closeness.
Liking of interaction partner was assessed with two measures. Participants were asked “How much did you like your interaction partner?” and provided with a Likert scale ranging from 1 (not at all) to 7 (a great deal). Participants also completed a question regarding their feelings about future interaction with their interaction partner, answering a 7-point item anchored by (1) I feel that I would never want to interact with this person again in the future and (7) I feel that I would very much want to interact with this person again in the future. These two measures were significantly correlated (α = .83; Spearman–Brown r = .69) and were combined to serve as an index of liking.
Perceived responsiveness of interaction partner was assessed with a four-item measure based on work by Reis et al. (2011) and Sprecher and Treger (2015), using a 7-point Likert scale with responses labeled at 1 (not at all true in this situation), 4 (somewhat true in this situation), and 7 (very true in this situation). Sample items include “My interaction partner seemed to really listen to me” and “My interaction partner was responsive to my questions/answers.” The four items were averaged and the average score was used in analyses (α = .84).
Enjoyment of interaction was assessed with a four-item measure, using a 7-point Likert scale ranging from 1 (not at all) to 7 (a great deal). Sample items include “How much did you enjoy the interaction?” and “How much fun was the interaction?” The four items were averaged and the average score was used in analyses (α = .82).
Individual Differences
Personality
To assess for the possibility that personality moderates the pattern of obtained results, we administered the Big Five Inventory-2 (BFI-2; Soto & John, 2017). The BFI-2 assesses all “Big Five” personality traits (60 items total, with 12 items each assessing open-mindedness: e.g., “Is curious about many different things,” α = .83; conscientiousness: e.g., “Keeps things neat and tidy,” α = .85; extraversion: e.g., “Is outgoing, sociable,” α = .86; agreeableness: e.g., “Is respectful, treats others with respect,” α = .80; and negative emotionality: e.g., “Is temperamental, gets emotional easily,” α = .88). Responses were provided using a Likert scale ranging from 1 (disagree strongly) to 5 (agree strongly).
Attachment
To assess for the possibility that interpersonal attachment orientations moderate the pattern of obtained results, we administered the Experiences in Close Relationships-12 Scale (ECR-12; Lafontaine et al., 2016). The ECR-12 consists of 12 items, six of which tap attachment anxiety and six of which tap attachment avoidance. Participants responded to the items regarding “How you generally feel in close relationships (e.g., with romantic partners, close friends, or family members),” using a 7-point rating scale ranging from 1 (strongly disagree) to 7 (strongly agree). We averaged across anxiety and avoidance items separately to obtain composite measures for each of these variables (sample item for attachment anxiety: “I worry about being abandoned,” α = .83; sample item for attachment avoidance: “I don’t feel comfortable opening up to close relationship partners,” α = .83).
Control Variables
To control for differences in experience with the communication modality to which they were randomly assigned, participants were asked “How often have you used [Skype text, Skype audio, Skype video, or Oculus Quest VR] before?” and provided with the following response options: 1 = never before today, 2 = once in my lifetime, 3 = once a year, 4 = a few times a year, 5 = once a month, 6 = a few times a month, 7 = a few times a week, and 8 = every day. To control for the possibility that differences in current relationship status (e.g., single vs. partnered) would influence interaction dynamics, participants were asked to “Please indicate your current relationship status” and provided with the following response options (n = frequency observed): 1 = single (n = 170), 2 = casually dating (n = 16), 3 = exclusively dating (n = 86), and 4 = married (n = 2). Responses to both of these ordinal variables were treated as continuous (Robitzsch, 2020) and used as controls in analyses.
Manipulation Check
To confirm participants’ awareness of the modality in which they communicated with their interaction partner, they were asked “How did you communicate with your interaction partner?” (response options: Skype Text, Skype Audio, Skype Video, or Oculus Quest VR). All participants answered consistently with their experimental condition.
Results
Analytic Approach
Because individuals who participated in the present study were nested within dyadic interactions, multilevel modeling (MLM) was used to account for this nonindependence (Scariano & Davenport, 1987). MLM is particularly useful for indistinguishable dyadic data (i.e., when there is not a distinction within dyads on a variable of interest) and it focuses on testing for mean differences in between-dyad conditions. Note that we conducted preliminary analyses to assess for the possibility of experimenter (n = 16), room location (n = 2), and/or dyad sex composition (matched vs. mismatched) effects. None were detected, so these variables were not included in subsequent analyses. We tested our hypotheses via planned contrasts, with results summarized in Table 2. Correlations between all study variables can be found here: https://osf.io/xsh4b/.
Table 2
MLM Condition Results for Planned Contrasts of Affiliative Outcomes
95 % UL
Variable
Estimate
SE
LL
UL
p
Closeness
Condition contrast (Hypothesis 1a): 1 versus 2, 3, 4
−1.75
.53
−2.79
−.70
<.001
Condition contrast (Hypothesis 2b): 1, 2 versus 3, 4
−.79
.31
−1.40
−.17
<.001
Condition contrast (Hypothesis 3c): 3 versus 4
−.01
.25
−.49
−.70
.98
Liking
Condition contrast (Hypothesis 1a): 1 versus 2, 3, 4
−1.69
.49
−2.67
−.71
<.001
Condition contrast (Hypothesis 2b): 1, 2 versus 3, 4
−.73
.29
−1.31
−.16
.01
Condition contrast (Hypothesis 3c): 3 versus 4
.14
.22
−.30
.58
.54
Responsiveness
Condition contrast (Hypothesis 1a): 1 versus 2, 3, 4
−1.78
.45
−2.66
−.89
<.001
Condition contrast (Hypothesis 2b): 1, 2 versus 3, 4
−.68
.26
−1.20
−.16
<.001
Condition contrast (Hypothesis 3c): 3 versus 4
.24
.20
−1.60
−.64
.24
Enjoyment
Condition contrast (Hypothesis 1a): 1 versus 2, 3, 4
−1.94
.46
−2.84
−1.04
<.001
Condition contrast (Hypothesis 2b): 1, 2 versus 3, 4
−.65
.27
−1.18
−.13
.02
Condition contrast (Hypothesis 3c): 3 versus 4
−.02
.21
−.43
.40
.94
Note. CI = confidence interval; LL = lower limit; UL = upper limit; MLM = multilevel modeling; VR = virtual reality. a Hypothesis 1 = contrasting text versus video/VR conditions. b Hypothesis 2 = contrasting text/audio versus video/VR conditions. contrasting video versus VR conditions.
Differences in Affiliative Outcomes
We began by conducting separate overall mixed model tests (using the restricted maximum likelihood [REML] method and Satterthwaite approximation of degree of freedom, including random intercepts and slopes for fixed effects, via SPSS MIXED Version 28) for each of the four outcome variables. Type III tests of fixed effects yielded a significant overall effect for condition (i.e., interaction modality, a Level 2 variable), controlling for modality experience and relationship status at the individual level (Level 1 variables). Consistent with the pattern of means displayed in Table 1 and Figure 2, closeness [F(3, 142.59) = .86, p = .01; modality experience: F(1, 241.06) = .74, p = .39; relationship status: F(1, 255.24) = .34, p = .56], liking [F(3, 141.81) = 4.11, p = .01; modality experience: F(1, 219.97) = .28, p = .60; relationship status: F(1, 235.95) = .41, p = .84], responsiveness [F(3, 142.35) = 5.57, p < .001; modality experience: F(1, 227.14) = 1.92, p = .17; relationship status: F(1, 242.93) = .24, p = .24], and enjoyment [F(3, 142.38) = 6.08, p < .01; modality experience: F(1, 234.73) = .37, p = .54; relationship status: F(1, 249.91) = 1.09, p = .30] all differed significantly by condition.
Figure 2
Affiliative Outcome Means With 95% Confidence Intervals by Interaction Modality Note. VR = virtual reality.
Testing Hypothesis 1
We hypothesized that affiliative outcomes would be higher in interaction modalities featuring sensory (audio, video, and VR) versus nonsensory input (text). Analyses of this planned contrast revealed that, consistent with Hypothesis 1, participants in the audio, video, and VR conditions reported significantly higher closeness (fixed effects coefficient estimate = −1.75, SE = .53, p < .001, estimated Cohen’s d = .51), liking (−1.69, SE = .49, p < .001, d = .56), responsiveness (−1.78, SE = .45, p < .001, d = .56), and enjoyment (−1.94, SE = .46, p < .001, d = .63) than did participants in the text condition.
Testing Hypothesis 2
We hypothesized that affiliative outcomes would be higher in interaction modalities featuring a visual representation of interactants (video and VR) than in modalities without such input (audio and text). Analyses of this planned contrast revealed that, consistent with Hypothesis 2, participants in the video and VR conditions reported significantly higher closeness (fixed effects coefficient estimate = −.79, SE = .31, p = .01, estimated Cohen’s d = .34), liking (−.73, SE = .29, p = .01, d = .34), responsiveness (−.68, SE = .26, p = .01, d = .39), and enjoyment (−.65, SE = .27, p = .02, d = .33) than did participants in the audio and text conditions.
Testing Hypothesis 3
We hypothesized that there would not be significant differences in affiliative outcomes among dyads interacting in video versus VR modalities. Analyses of this final planned contrast revealed that, consistent with Hypothesis 3, participants in the video and VR conditions did not differ significantly in levels of closeness (fixed effects coefficient estimate = −.01, SE = .24, p = .98), liking (−.14, SE = .22, p = .54), responsiveness (.24, SE = .20, p = .24), or enjoyment (−.02, SE = .21, p = .94).
Examining Possible Moderating Effects of Individual Differences
We tested for the possibility that the obtained effects would be moderated by individual differences in personality traits (Big Five) and/or attachment orientations (avoidance and anxiety). We noted one consistent main effect for the personality dimension of agreeableness, such that participants who were higher in this trait reported significantly higher levels of all affiliative outcomes irrespective of condition. However, as summarized by the Type III fixed effects tests presented in Tables 3 (personality) and 4 (attachment), no evidence for moderation of condition effects was found for any of the five personality dimensions or for attachment anxiety on any of the four affiliative outcomes. That is, the obtained affiliative outcome results by condition did not differ significantly as a function of participants’ personality traits or attachment anxiety. This same absence of moderation also emerged for attachment avoidance for three of the four affiliative outcomes. We observed some evidence for moderation of condition effects for attachment avoidance with respect to closeness (p = .04), but the meaningfulness of this finding is tempered by the total number of tests for moderation conducted.
Table 3
Analyses of Moderated Condition Effects for Affiliative Outcomes: Personality
Outcome
Big Five personality trait
df(num, den)
F
p
Closeness
Open-mindedness
1, 241.91
1.24
.27
Conscientiousness
1, 247.21
3.43
.07
Extraversion
1, 239.83
0.94
.33
Agreeableness
1, 246.02
10.54
<.001
Negative emotionality
1, 246.42
0.00
.95
Open-mindedness × Condition
3, 242.28
0.16
.92
Conscientiousness × Condition
3, 243.95
0.16
.92
Extraversion × Condition
3, 238.42
1.19
.31
Agreeableness × Condition
3, 242.36
0.20
.90
Negative emotionality × Condition
3, 245.22
0.77
.51
Liking
Open-mindedness
1, 224.94
0.13
.72
Conscientiousness
1, 236.64
1.14
.29
Extraversion
1, 220.03
0.01
.91
Agreeableness
1, 231.93
29.58
<.001
Negative emotionality
1, 245.78
0.89
.35
Open-mindedness × Condition
3, 225.74
0.35
.79
Conscientiousness × Condition
3, 231.74
0.18
.91
Extraversion × Condition
3, 218.32
0.76
.52
Agreeableness × Condition
3, 226.98
1.55
.20
Negative emotionality × Condition
3, 244.40
1.69
.17
Responsiveness
Open-mindedness
1, 232.57
0.04
.85
Conscientiousness
1, 242.58
0.04
.85
Extraversion
1, 228.64
0.73
.40
Agreeableness
1, 239.26
15.62
<.001
Negative emotionality
1, 247.97
0.83
.36
Open-mindedness × Condition
3, 233.28
0.94
.42
Conscientiousness × Condition
3, 237.89
0.56
.64
Extraversion × Condition
3, 226.88
1.11
.35
Agreeableness × Condition
3, 234.48
0.10
.96
Negative emotionality × Condition
3, 246.68
0.95
.42
Enjoyment
Open-mindedness
1, 235.64
0.18
.67
Conscientiousness
1, 243.98
0.38
.54
Extraversion
1, 232.37
0.44
.51
Agreeableness
1, 241.33
14.71
<.001
Negative emotionality
1, 248.00
0.58
.45
Open-mindedness × Condition
3, 236.24
0.57
.63
Conscientiousness × Condition
3, 240.00
1.44
.23
Extraversion × Condition
3, 230.83
0.15
.93
Agreeableness × Condition
3, 237.22
2.08
.10
Negative emotionality × Condition
3, 246.92
0.53
.67
Note. Type III tests of fixed effects, controlling for modality experience and relationship status; df(num, den) = degrees of freedom (numerator, denominator).
Table 4
Analyses of Moderated Condition Effects for Affiliative Outcomes: Attachment
Outcome
Attachment dimension
df(num, den)
F
p
Closeness
Attachment anxiety
1, 252.33
1.73
.19
Attachment avoidance
1, 254.83
0.08
.78
Attachment anxiety × Condition
3, 252.39
0.60
.62
Attachment avoidance × Condition
3, 250.75
2.86
.04
Liking
Attachment anxiety
1, 237.84
1.86
.17
Attachment avoidance
1, 241.76
0.77
.38
Attachment anxiety × Condition
3, 238.09
0.51
.68
Attachment avoidance × Condition
3, 237.40
0.66
.58
Responsiveness
Attachment anxiety
1, 243.18
0.67
.42
Attachment avoidance
1, 246.72
0.28
.60
Attachment anxiety × Condition
3, 243.39
0.34
.80
Attachment avoidance × Condition
3, 242.43
1.24
.30
Enjoyment
Attachment anxiety
1, 248.82
0.15
.70
Attachment avoidance
1, 251.85
0.04
.84
Attachment anxiety × Condition
3, 248.94
0.45
.72
Attachment avoidance × Condition
3, 247.55
0.52
.67
Note. Type III tests of fixed effects, controlling for modality experience and relationship status; df(num,den)= degrees of freedom (numerator, denominator).
Discussion
The present preregistered experiment examined how the induction of closeness through a validated social interaction task affects affiliative outcomes among unacquainted participants interacting via text, audio, video, or VR. Consistent with hypotheses, VR interactions were found to be comparable to both audio and video interactions in generating closeness, while text interactions led to significantly lower outcomes. These findings, with significant hypothesized effects in the medium size range (d = .33–.63) per Cohen (1988), parallel the pattern of results first observed by Sprecher (2014), with dyads in the text condition reporting the lowest affiliative outcomes and dyads in the audio and video conditions reporting higher (but not significantly different) affiliative outcomes.
Consistent with past findings of no differences in affiliative outcomes between those interacting FTF versus via video chat, we also found support for our hypothesis that there would be no significant differences in affiliative outcomes between the video and VR modalities. After all, both modalities include synchronous audio between interactants and both involve a dynamic visual representation of interactants in real time. These findings are consistent with our contention that such interactions cue established FTF interaction norms associated with both giving and receiving self-disclosures with another person in real time, leading interactants in both modalities to evidence similarly high levels of affiliative outcomes. In addition, no significant differences were found between the audio and VR conditions. Although no hypotheses were offered regarding a comparison between these two conditions, this finding suggests that the synchronous nature of social interaction that both of these modalities (along with video) feature may be a critical element influencing affiliative outcomes.
One notable strength of the present study is the inclusion of VR, for the first time to our knowledge, as an interaction modality for testing the possibility of closeness generation among previously unacquainted people. The application of Aron et al.'s (1997) “fast friends” procedure enabled us to use a well-established experimental paradigm to examine the generation of interpersonal closeness in VR and to directly compare VR to other interaction modalities to test relative effects on closeness and allied affiliative outcomes. An additional strength of the present research is the inclusion and analysis of important individual difference variables (the Big Five personality dimensions and the attachment orientations of anxiety and avoidance), as past research has not examined key personality variables that may influence the generation of closeness among strangers in various modalities. The general lack of effects observed for these individual differences suggests the robustness of the obtained findings across individual differences. The one significant finding among these individual differences, moderation of condition effects for attachment avoidance with respect to closeness, is intriguing but we are hesitant to interpret it given the sheer number of moderation tests conducted. When we decomposed this interaction, we found that those higher in avoidance reported lower closeness in the text condition relative to the other conditions. Whether this finding is robust or spurious will require future research.
One limitation of the present study is the lack of a FTF condition. This was largely due to the COVID-19 pandemic, in the interest of protecting the health of all participants and research assistants. Future research might directly compare a VR interaction condition with FTF interactions. Expanding upon past research findings (Sprecher, 2014), recent research by Sprecher (2021) found no significant differences in affiliative outcomes between video and FTF interactions. In addition, research examining conflict resolution found no meaningful advantage for FTF over texting (Pollmann et al., 2020). We predict that VR and FTF interactions may similarly produce statistically identical outcomes.
A second potential limitation of the current investigation is that, although all participants (and research assistants) were required to wear a face mask throughout their laboratory session, the avatars in the VR condition were not wearing face masks. When queried in the postinteraction survey as to their beliefs regarding what may have prevented the development of closeness with their interaction partner, several participants in the video condition reported that the use of face masks may have reduced closeness. In the video condition, partners’ facial expressions were partially obscured by their mask. Still, reduced facial cues due to masking in the video condition should be considered in tandem with the limitations in avatar facial expression in the VR condition, as both conditions presented participants with suboptimal facial cues. A replication of the present study should be conducted in the future once the COVID-19 pandemic has ended and research can be conducted safely without the use of personal protective equipment.
It is also important to note that participants in the present study had no previous experience using VR. This inclusion criterion was put in place to focus on novice users and their social interaction outcomes given the growing number of new VR users in the population. Accordingly, it is not known whether the current findings would be obtained among those who have prior experience using VR and this is an important direction for future research.
Another future direction for psychologists interested in examining the effects of VR on interpersonal relationships may be to gain a better understanding of the conditions under which VR may hinder versus enhance the development of intimacy and positive interpersonal outcomes. It is noteworthy that our results did not show a difference between the video and VR conditions. This finding is in line not only with research on communication modalities and interpersonal outcomes (Sprecher, 2014, 2021), but also with other research showing that visual cues may not be indispensable in guiding first impressions or promoting affiliative outcomes (Croes et al., 2020; Schroeder & Epley, 2015). For example, Schroeder and Epley (2015) found that participants were more likely to evaluate a job candidate more positively when presented with auditory versus text-based information and that adding visual cues to auditory information did not lead to more positive evaluations. These results suggest the interesting possibility that, by focusing one’s attention on auditory information, VR may, perhaps, promote affiliation compared to other modalities precisely because it can mask cues that are either unnecessary or a source of bias in impression formation (e.g., race and attractiveness, which influence impression formation through top–down processes such as stereotype activation; Connor et al., 2021; Fiske et al., 2007). Viewed in this light, VR may provide an avenue for forming and developing more accurate first impressions among members of social groups who may have few opportunities to interact FTF in everyday life, or for individuals who find it difficult to build relationships in this context.
Interest in remote connectivity between people has increased markedly since the beginning of the COVID-19 pandemic and there are no signs that it will recede any time soon (Bidar & Patterson, 2022). As excitement among consumers to experience VR increases, one question worth pondering is the “value added” of using VR to form interpersonal connections versus using other established communication modalities. Clearly, VR is particularly valuable in allowing people to experience places, situations, and events that they otherwise could not experience. Meeting a stranger, whether in real life or in VR, is not a particularly novel or unique experience; it can happen at almost any time in one’s social life. Perhaps, then, the value of VR is particularly seen when it provides opportunities not afforded readily by other modalities. That said, despite the lack of ultra-realism with respect to humanoid avatars currently afforded by the instantiation of VR provided by Rec Room, the affiliative outcomes found were of similar magnitude to those obtained via other sensory-rich modalities. However, it is entirely possible that more immersive VR applications would provide an enhanced interactive experience and generate different outcomes than those obtained in the present research. This possibility provides another fruitful avenue for future research.
Moreover, in discussing the personal experience of VR, Steuer (1992) refers to critical determinants of telepresence, “the experience of presence in an environment by means of a communication medium,” in which communication modalities vary: vividness and interactivity. As Steuer (1992) notes, vividness can be differentiated with respect to the breadth and depth of engagement with the experiencer’s senses. The present research varied the communication modality in which participants interacted, providing greater breadth with respect to vividness in some conditions than in others. Additional sensory engagement beyond that afforded in the present study, via the inclusion of haptics, for example, would enhance vividness and offer an opportunity to further test the extent to which affiliative outcomes may differ on this dimension.
This study contributes to the body of scientific knowledge regarding the effects of technological advances on developing relationships (Sprecher, 2014; Sprecher et al., 2013), especially concerning the degree to which VR technology impacts perceptions of closeness between strangers. This research and its implications are particularly relevant today, given the heightened use of CMC during the ongoing COVID-19 pandemic. Future research should further investigate social–psychological phenomena within virtual settings, as virtual venues for social interaction continue to increase.
Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC-BY-NC-ND 4.0)
