An Exploration of Virtual Reality User Profiles and Their Relationship with Symptoms of Addiction

Abstract

Previous research on digital media and well-being showed that the impact of media technologies on users largely depends on specific usage characteristics, context, and individual user factors. This research report explores the existence of distinct user profiles in virtual reality (VR), using an existing dataset and applying latent class analysis. The results point to the existence of three classes: two of them primarily include VR game users, who differ in the intensity of their psychological engagement (sense of presence) during gameplay, and a third class that includes users of social VR platforms, who report a relatively very high level of engagement with these platforms. The three groups differ in symptoms of addiction to the technology. These results provide valuable proof of concept for the usefulness of identifying usage profiles as a way to reveal how the relationship between VR and well-being might vary across different usage situations.

Introduction

With the popularization of virtual reality (VR) devices and applications in recent years, concerns are growing about the potential negative effects that VR may have on users, ¹ particularly regarding their impact on users’ mental health.^2,3 To develop an effective research program on this issue and avoid “reinventing the wheel,” ⁴ it is essential to consider previous experiences in other areas of research on digital media (such as social media and video games) and well-being.

In this respect, numerous scholars argued that, after important research efforts leading only to mixed and inconclusive results, it has become clear that the simplest models of social media effects (e.g., dose–response models, which simply propose effects of usage time on users in general) are limited in their ability to adequately capture the complexity of social media impact.^4–6 Indeed, different social media applications have distinct structural features and can be used for various and disparate activities (e.g., browsing, social interactions—with friends or strangers ⁶ ), which may, in turn, have different impacts on users depending on their individual characteristics or vulnerabilities. ⁷ At the same time, users of different digital applications (e.g., social media, video games) may be exposed to different online contents and risks.^8–10

There is no reason to assume that these arguments would not also apply to VR applications. Hence, when analyzing the impact of VR on users’ mental health, a logical starting point may be—rather than just assuming average effects across diverse users and usages—to identify which user subpopulations and types of use may pose a greater risk. While some preliminary research ¹¹ has explored response profiles to VR in laboratory settings, to the best of our knowledge, there are no published studies that examine the existing profiles among VR users in everyday life. Therefore, the aim of this study is to take a first step in this direction by exploring the VR user profiles that emerge from a previously existing dataset ³ using latent class analysis (LCA). ¹² Additionally, we examine the relationship between those profiles and one of the indicators of digital mental health most commonly analyzed, the risk of addictive use. Regarding this point, we follow the perspective adopted in the original article from which our dataset is drawn, ³ where addictive use of VR is regarded in line with the concept of (video)game addiction as in Lemmens et al., ¹³ (understood, in general terms, as a compulsive engagement with the technology that leads to long-term harm or distress).

Within this context, the goal of this short report is to serve as an initial proof of concept for the relevance of establishing user profiles and to contribute to a better understanding of which groups may be at greater risk of addictive use of VR.

Users and usage profiling

Previous research in the field of digital media has pointed to various types of factors that can determine the impact of digital media on its users. First, sociodemographic factors are key in shaping this impact: age and age-related developmental aspects may explain vulnerabilities to certain effects, while gender may also be associated with differences in how technology is experienced ¹⁴ or with a higher likelihood of victimization, ¹⁵ among others. Another type of factor that can clearly influence media impact is the specific characteristics of the application used, ⁶ as well as the intensity of its use. ³ Last, the cognitive and emotional states experienced by the individual during media exposure are also often considered crucial in explaining its effects.^16,17 In this regard, in the case of VR, feelings of presence—the sensation of “being there,” including both physical presence and social presence (“being there with others”)—as well as the sense of embodiment in an avatar, are key to explaining both the positive and negative effects of using VR applications.^3,18

The multiple possible combinations of all these factors, and the fact that their effects are not necessarily additive or linear, make it very difficult to analyze their influence in isolation. Moreover, in practice, only certain combinations of these factors may occur in real-world settings. This underscores the usefulness of a person-oriented research approach focused on the identification of distinct subgroups (i.e., profiles or classes) of individuals, based on a pattern of factors (and in contrast with variable-oriented research approaches, which emphasize specific variables or characteristics considered in isolation). ¹⁹ Therefore, user profiling techniques can be particularly advantageous in this context, as they help reduce variability and heterogeneity in the data and identify patterns shared by many users, allowing for simple and reliable user classification. User profiles could also be instrumental in explaining certain media effects that may be facilitated or hindered by specific combinations of individual, usage-related, and cognitive–emotional response factors. ¹⁷

Thus, in this short report, we explore the following two research questions (RQs):

RQ1. Can VR user profiles be defined based on sociodemographic factors (age, gender), patterns of VR use (type of application, time spent using it, previous use), and user responses to VR (feelings of presence and embodiment)?

RQ2. Are such user profiles associated with a greater susceptibility to experiencing symptoms of addiction?

Method

Participants and procedure

The present study applies LCA to an existing dataset collected for a previous research on the use of VR and addiction. ³ It includes data from a sample of 754 participants. Participants ranged from 18 to 86 years old (M = 30.12, SD = 9.89) with 81.70% (n = 616) self-identified as male, 11.54% (n = 87) as female, 4.38% (n = 33) as of other gender (i.e., nonbinary, third gender, other) while 2.39% (n = 18) preferred not to say. The majority of participants were from the United States (59.55%, n = 449), followed by Canada (9.02%, n = 68), and the United Kingdom (6.76%, n = 51), with the remaining participants from a wide variety of countries spread across several continents (for additional details, refer to the Supplementary Data).

Measures

Variables used for LCA

Since LCA in R requires categorical variables, we transformed some of the variables in the original dataset. Age was categorized into tertiles, including emerging adults (M = 21.06, SD = 1.77), young adults (M = 28.43, SD = 2.2), and adults (M = 41.79, SD = 7.93), with around one-third of the participants in each tertile. Self-identified gender was categorized as male, female, other, and prefer not to say. Regarding VR previous use, participants were categorized either as novice users (1 month of use or less), experienced users (between 1 and 6 months), and experts (6 months or more). Based on the hours of use of VR per week, participants were categorized into light (33.42% of the sample; M = 3.77, SD = 1.69), moderate (34.35%; M = 9.99, SD = 2.31), and heavy use (32.23%; M = 26.91, SD = 13.15). The type of VR application most used by each participant was included in three possible categories: VR games (i.e., VR experiences with gamelike dynamics such as goals, rules, progression; e.g., Beat Saber), social VR platforms (VR experiences without gamelike dynamics and focused on real-time social interaction; e.g., VRChat), and other VR, as coded in the original dataset. Feelings of spatial presence and embodiment, initially measured using participant’s responses to one item for each variable (“Usually, when I use VR, I feel like I am actually there in the virtual environment” and “Usually, when I use VR, I feel as if my virtual body is my body”) on a 5-point Likert-type scale (from “I do not agree at all” to “I totally agree”) and were recategorized in disagreement (those scoring 1 or 2 in the scale), being neutral (scoring 3), and agreement (scoring 4 or 5).

Addiction to VR

An adapted version of the 7-item Game Addiction Scale (GAS-7) ¹³ was used to measure addiction to VR applications, with the items modified to refer specifically to the use of VR rather than video games. ³

Analysis

LCA was employed to identify distinct subgroups (i.e., classes) of VR users based on the seven indicators described above. Being a latent model-based method, LCA was preferred to other clustering techniques (e.g., cluster analysis) because its focus on the latent statuses conceptually matches the view of individuals as a totality. It also provides model fit to data, allowing for comparisons across different models. To find the best model and the optimal class solution, we run a series of models starting with a one-class model and adding one additional class at a time until reaching a nine-class model (estimating each latent class model multiple times, i.e., nrep = 30). Following existing recommendations, ¹² to choose the best model, we used multiple fit statistics giving priority to the Bayesian information criterion (BIC), but also considered theoretical interpretability, entropy diagnostic statistic, and class size.

Chi-squared test was used to reveal differences between the classes in the distribution of the independent categorical variables or indicators. Standardized residuals of the cells were considered to quantify the specific contribution to the results of the chi-squared test in case it was significant. ²⁰ Finally, we examined the association between class membership and symptoms of VR addiction—as a potential validator of the selected class solution using a linear regression model. All statistical analyses were conducted in RStudio using the “psych,” ²¹ “dplyr,” ²² “coefficientalpha,” ²³ and “poLCA” ²⁴ packages.

Results

Supplementary Table S1 (Supplementary Data) reports the characteristics of the sample. Most participants were males, expert users, played VR games, and reported experiencing spatial presence but were neutral about feelings of embodiment.

LCA of VR users

The results from the LCA show the existence of distinct classes of VR users (see Table 1). We selected a three-class model as the optimal solution, taking into account both BIC and aBIC, and considering that entropy did not improve with model 4.

Table 1.

Model Fit and Diagnostic Criteria

Model	Log-likelihood	Residual df	BIC	aBIC	cAIC	Likelihood ratio	Entropy
Model 1	−4,631.787	739	9,362.955	9,315.324	9,377.955	1,240.154	—
Model 2	−4,533.143	723	9,271.673	9,173.236	9,302.673	1,042.867	0.586
Model 3	−4,476.220	707	9,263.834	9,114.590	9,310.834	929.021	0.657
Model 4	−4,452.017	691	9,321.434	9,121.384	9,384.434	880.615	0.653
Model 5	−4,436.037	675	9,395.480	9,144.623	9,474.480	848.655	0.657
Model 6	−4,419.568	659	9,468.549	9,166.885	9,563.549	815.717	0.634

aBIC, sample-size-adjusted BIC; BIC, Bayesian information criterion; cAIC, consistent Akaike information criterion; df, degree of freedom.

Class 1 (“Gamers experiencing Low Presence,” GLP; estimated class population share = 28% of the sample) members were more likely male, showing low weekly hours of use, primarily using games and not reporting spatial presence and embodiment. Class 2 (“Social VR Users,” SVRU; 29%) included participants who were more likely female (and other gender), emerging adults, heavy users, mostly of social VR platforms, and experiencing higher levels of spatial presence and embodiment. Finally, Class 3 (“Gamers experiencing High Presence,” GHP; 43%) were less likely emerging adults, less likely spending high amount of weekly hours, more likely male, using games in VR, and reporting spatial presence and embodiment. Figure 1 shows the probability of indicator categories for each class (the results of chi-squared test are reported in the Supplementary Material, Supplementary Table S2 and Supplementary Figure S1).

FIG. 1.

Probability of indicator categories for each class.

VR classes and symptoms of VR addiction

The results of the regression model with adapted GAS-7 total score as the dependent variable indicated that class membership explained 12.8% (adjusted R²) of the score variation in the model, F(2, 751) = 56.04, p < 0.001. The SVRU class was associated with an increase of 4.1 points on the adapted GAS-7 compared with that of GLP (Table 2), and the effect size of this association was large. ²⁵ Similarly, being member of the GHP class was associated with an increase of 1.7 points on the adapted GAS-7 compared with LPG, and the effect size of the difference was small to moderate.

Table 2.

Linear Regression Model of Association Between Classes and Virtual Reality Addiction Symptoms

Independent variable	SMD (SE)	Unstandardized estimate (SE)	t value	p value
Intercept	−0.399 (0.058)	11.678 (0.258)	45.290	<0.001
Gamers experiencing Low Presence	reference
Social VR Users	0.925 (0.088)	4.078 (0.386)	10.574	<0.001
Gamers experiencing High Presence	0.377 (0.080)	1.663 (0.353)	4.715	<0.001

N = 754.

SE, standard error; SMD, standardized mean difference.

Discussion

Our results point to the existence of three distinct groups of VR users. Two groups were predominantly constituted by users of VR for gaming activity: those who engaged in more casual use and those with moderate usage patterns, who experienced higher levels of immersion and engagement in the game, and who reported slightly higher symptoms of addiction. What seems particularly interesting is that the results also suggest the existence of a third group of users—those who engaged mainly with social VR, with particularly high intensity, with a higher likelihood of including women and individuals of genders other than male, and who reported even higher levels of immersion and symptoms of addiction. These distinctions between light and more involved users and between primary activities of involvement (i.e., between social media and gaming) were previously observed in the use of other types of media technologies. ²⁶

The more intensive use by those engaging with social VR platforms may be explained by the wide range of activities available on these platforms, which go beyond pure gameplay and include various forms of social interaction. Previous research has highlighted the multiple benefits these platforms can offer to users: from meeting like-minded people in an immersive way that resembles physical reality,^18,27 to providing particularly valuable spaces for members of marginalized communities, ²⁸ or enabling users to explore their identities in a safe environment. ²⁹ Qualitative evidence suggests that social VR platforms may provide LGBTQ users with an emotionally supportive space to come out and build close relationships, ²⁸ which may foster self-esteem through experimenting and affirming their own identity. ³⁰ The sense of social presence may also boost social support. ³¹ The downside of these benefits is that the wide range of psychological rewards involved might also increase the addictive potential of these platforms, especially compared to other VR activities. Furthermore, experiences of harassment risks (which often affect women and LGBTQ users) in social VR necessitate additional examination. ³²

Recent approaches to studying media effects have emphasized the need to focus not on technologies or applications as a whole but on the specific technical features of platforms, and what users actually do with them. ⁶ Our findings align with this perspective, revealing clear differences between groups of users who engage with VR games and those who use social VR applications. Thus, one of the contributions of this research report is to begin establishing—albeit in an exploratory way—different categories of VR usage situations, which, as our findings show, may be differently associated with addiction symptoms. This represents a first step toward addressing future research on the addictive potential of VR in an effective way, one that acknowledges the complexity of social uses of these technologies and moves beyond overly simplistic approaches that fail to capture the nuance of the phenomenon.^4,6

One of the limitations of the present analysis is that it relied on a dataset collected previously, hence not tailored to uncover distinct subgroups of VR users. Furthermore, data were collected via self-report measures only and may have been influenced by social desirability and recall biases. Of importance, considering the study’s cross-sectional nature, our findings should be interpreted in terms of concurrent association rather than prediction. The present analysis relied on data primarily collected from male respondents from Western countries, recruited through VR-related Facebook groups and Reddit subreddits. Consequently, our findings cannot be generalized to the entire population of VR users, and further research should consider including a better representation of non-male users. Furthermore, we used a specific measure of addictive use, ¹³ which is grounded in a particular conceptualization based on pathological gambling criteria. When interpreting our results, it is important to bear in mind not only that there are alternative or complementary perspectives on technology addiction but also that these are part of an ongoing debate around the concept itself and the potential risk of pathologizing certain behaviors.^33,34

Despite these limitations, this research report represents a first proof of concept, not addressed before in the literature, regarding the usefulness of establishing user profiles as an initial step toward a person-oriented perspective in the study of the impact of VR on users’ mental well-being. It thus paves the way for future research aimed at understanding not only general or assumed effects of the technology on users as a whole but also the role of individual, social, and situational factors that determine the variability of these effects. Longitudinal and experimental studies, as well as mixed-methods research, are needed to better understand socio-affective dimensions and the impact of immersive technology use.

Declaration of Generative Artificial Intelligence and Artificial Intelligence-Assisted Technologies in the Writing Process

The text of this article was originally written by the authors; artificial intelligence tools were only used occasionally to improve and polish the language.

Authors’ Contributions

S.A.: Conceptualization, formal analysis, visualization, writing—original draft, and writing—review and editing. M.B.-Á.: Conceptualization, supervision, writing—original draft, and writing—review and editing.