Assessing Students’ Personality Traits: A Study of Virtual Reality-Based Educational Practices

Abstract

Personality, as a crucial foundation for assessing human psychology and behavior, stands as a significant subject of interest among psychology researchers. Tailoring education to the needs of the student is likewise an important topic in the field of education, where the personality traits of students play a crucial role in their future professional and personal development. Presently, experts in the field predominantly employ questionnaires to evaluate personality traits. However, this approach has limitations, particularly for younger students, whose developing cognitive abilities might lead to inaccuracies in conveying information, thus impacting their performance in predictive assessments. Moreover, the questionnaire’s quantitative nature could inadvertently affect the respondents’ psychological responses. To address these challenges, we incorporate VR technology. Leveraging immersive and highly controllable features of VR, this study introduces a personality assessment framework tailored for students and develops a VR prototype system based on this framework for future performance evaluation. In our experimental evaluation, we engaged 96 students, ranging in age from 10 to 22, to participate in the testing process. The results of this evaluation indicate that our personality assessment framework performs effectively across four dimensions of personality evaluation. However, there remains a need for further analysis and enhancement in areas such as classification accuracy and the logical structure of scenario design. Additionally, it is essential to continue seeking more objective methods for personality assessment.

1. Introduction

Contemporarily, customizing education to suit each child’s needs has been a significant concern in the field of education, as the individual traits of students play a crucial role in their future both professionally and personally. Examining these personality traits assists parents and schools in understanding children’s behavior in specific environments, along with their ways of perceiving, processing, and responding to information in critical situations [1]. This understanding enables parents and schools to design education that aligns with their children’s personality traits, equipping them with skills, competences, and mindsets that are advantageous in various environments [2]. Thus, the analysis of students’ personality traits is a vital aspect of school education. Moreover, these traits not only shape individuals’ emotions and life experiences, but they also define their interactions with the world, greatly influencing their socialization [3]. Therefore, understanding the development of personality traits is essential for both societal and individual well-being.

The field of personality assessment, a key component of personality research, has seen significant advancements by psychologists over recent decades. Contemporary methods primarily utilize paper–pencil and computer-based questionnaires. However, it is noted that these methods, conducted in controlled environments, are prone to information transmission errors, which can impede their effectiveness in predictive assessment [4]. In the 1960s, a study by Cornell University highlighted a discrepancy in students’ self-reported SAT scores compared to their actual scores in surveys. This variance was attributed to the tendency of individuals to align with social norms and present themselves in a more favorable light. In addition, personality assessment questionnaires can sometimes be psychologically leading (individuals may aspire to certain traits, influencing their choices). Concurrently, a review of most personality assessment questionnaires demonstrated that they predominantly employ rating scale questions. This question format can yield unhelpful data, particularly when respondents, unsure or unable to express their views aptly, resort to neutral responses [5]. VR-based scenarios for implicit measurement in psychological testing can be introduced, avoiding the need for users to report subjective assessments. Analysis of user interactions with the virtual environment, including their interactions, heart rate variability, and skin conductance responses [6], can be used to infer personality traits. This approach effectively mitigates the aforementioned risks. Moreover, the variability in cognitive abilities across different age groups, particularly among younger participants, can affect the accuracy of test results in personality assessments. Younger students may lack self-awareness, leading to results that do not entirely capture their true personality traits [7].

With the advent of a new wave of technological revolution and industrial transformation, novel information and communication technologies (ICTs) are increasingly being integrated into psychological evaluation. Notably, VR technology is emerging as one of the most promising and innovative tools in psychological assessments [8]. This technology, due to its enhanced sensitivity and objectivity, is gaining significant importance in psychological evaluations [9]. Traditional personality assessment tests that rely on questionnaires can only capture information that individuals are able and willing to report, failing to reflect genuine responses indicative of cognitive control. This method lacks objective, multi-dimensional, and automated measurement. To address these limitations, the field of computational personality assessment has emerged [10], utilizing computational techniques to predict personality traits more comprehensively and objectively. Among computational methods, Virtual Reality (VR)-based approaches have recently garnered significant interest in the psychology community. VR technology offers an immersive interactive environment. This experience, referred to as “presence,” creates an illusion of physical existence in a digital world for the user. It allows individuals to explore their mental perceptions in the environment without “realizing” that their interaction is mediated by technology [11,12]. Essentially, VR enables the replication of behaviors, emotions, and subjective experiences akin to real-life scenarios. This advantage significantly minimizes errors in information transmission during the assessment process. Moreover, virtual reality technology offers a controlled experience in a virtual setting, where the assessment program can be adjusted dynamically to cater to the unique requirements of different individuals, based on their specific circumstances [13]. Utilizing VR for the personality assessment of student groups appears to be a promising method.

In this research, we introduce an educational-focused personality assessment framework, which is grounded in the Myers–Briggs Type Indicator (MBTI) theory. This framework involves the development of a VR prototype system for personality assessment, where we shift away from the traditional questionnaire-based quantitative evaluation. Instead, we integrate the assessment mechanism directly into the framework to minimize the effect of psychological suggestions during testing. Our framework primarily targets the personality assessment of students, and it is crafted to suit various age groups. The narrative design and interaction dynamics are tailored to accommodate differing cognitive abilities, thereby significantly reducing potential errors in evaluation. Moreover, the immersive scenarios in this framework are designed to elicit specific personality traits from the users. Essential data such as user interaction responses and physiological characteristics are compiled and utilized as foundational elements for the assessment model, resulting in a comprehensive personality evaluation.

In summary, our main contributions include the following:

• Based on the MBTI theory, which is a well-recognized approach in personality assessment, this paper introduces a tailored framework for evaluating the personalities of students.
• The research has successfully established a relationship between VR scenarios and the mapping of personality traits, leading to the creation of specific assessment environments.
• We have developed a dataset that is optimally suited for our personality assessment framework. This dataset is the foundation for the personality assessment model.
• A prototype system using VR, crafted around our personality assessment framework, has been developed. This system is geared towards enhancing personality assessments in the educational sector.

The structure of this paper is as follows: Section 2 delves into the application of VR in psychological assessment, discussing previous research on the relationship between VR environments and personality traits. Section 3 details both the architecture of the personality assessment framework and the model itself. The design aspects of the prototype system are explored in Section 4. Section 5 presents our experiments on personality assessment and the evaluation of their results. The final section offers a conclusion.

2. Related Work

In recent years, the advancement and improvement of Virtual Reality technology have led to its integration in various fields, such as skills training and virtual simulation, significantly reducing costs and improving efficiency [14,15]. Similarly, VR has played a critical role in the fields of medicine and mental health [16]. Personality, serving as a crucial foundation for the exploration of human psychology and behavior, has garnered significant attention among psychological researchers [17]. In personality assessment, the MBTI self-assessment tool has gained widespread popularity as a means for understanding and distinguishing the personalities of ordinary individuals. It has received significant recognition from numerous global enterprises since its introduction, and is now a commonplace instrument in human resources. In the United States alone, more than two million MBTI questionnaires are utilized annually, establishing MBTI as one of the most renowned and authoritative personality assessment tools worldwide [18]. The fundamental concept underlying MBTI theory is directed towards the description and explication of individual differences in personality and behavioral preferences. This system primarily comprises four binary dimensions of indicators: “Extraversion (E) and Introversion (I)”, “Sensing (S) and Intuition (N)”, “Thinking (T) and Feeling (F)”, and “Judgment (J) and Perception (P)”. These four dimensions are employed to characterize individuals, resulting in sixteen unique combinations.

In this paper, our primary focus is on the MBTI theoretical framework, which we apply in conjunction with virtual reality technology to develop a framework for personality assessment tailored to educational purposes. In this process, we leverage the superior capabilities of VR technology, including its immersive nature, interactive features, ability to engage multiple senses, capacity for conceptualization, and user autonomy [19]. Importantly, VR technology is not constrained to mimicking physical reality in scenario creation. Instead, it offers the flexibility to generate unique, imaginative worlds [20]. These attributes of VR technology facilitate the provision of a secure, “no loss” immersive environment for personality assessment [21]. In this virtual setting, an individual’s personality traits are activated through a carefully crafted storyline. The evaluation of these traits is based on an integrated assessment model in the system.

Rosenthal et al. [22] conducted a study in the medical field to assess personality traits. This research compared the personality characteristics of surgical residents to those of the general population, utilizing a virtual reality (VR) laparoscopic simulator to evaluate any correlation between personality traits and technical performance. The findings strongly validated the effectiveness and feasibility of VR technology for personality assessment in professional settings [23]. In addition, the results of this study found that surgical residents showed personality traits different from those of the general population, demonstrating lower neuroticism, and higher extraversion and conscientiousness. Furthering this exploration, Lu Juan et al. conducted a study using the MBTI test on a group of naval officers and soldiers. Their research focused on the effect of occupational differences among various soldier types on personality traits. They particularly analyzed the effect of factors such as unit type, operational departments, and soldier categories, while noting that gender had minimal effect on these traits [24]. Therefore, the design of VR-based assessments must consider the varying emphasis placed on different personality dimensions across different occupations and job roles.

In another study, Roberts, Adam C. and co-researchers developed a custom, open-source VR system named PSY-VR, specifically for psychological assessments in a modifiable virtual environment [8]. This study also included an evaluation of the system’s performance by comparing responses in a Flanker task conducted in both a real laboratory and a similar virtual laboratory setting. The findings indicate that responses in VR settings align closely with those from real-life assessments, emphasizing VR’s effectiveness in psychological research. Ju et al. demonstrated that psychopathology was negatively correlated to prosocial personality This finding suggests that personality differences can predict intuitive decision-making, and this process can be investigated under specific controlled conditions. Observing decision choices within a specific environment can serve as a robust method for personality assessment [25]. The swift advancement of VR immersive technologies, coupled with their integration with physiological measures, suggests that VR-based evaluations will likely become the norm in future psychological assessments.

Markus W and team developed a VR tool specifically for assessing personality traits in team sports (e.g., football) [26]. This tool addresses the limitations of current leading methods in sports personality assessment, which primarily rely on validated questionnaires. These questionnaires often yield non-sport-specific, subjective self-reports, and fail to measure trait manifestation in context-specific performance. In another development, Zeng Yingchun et al. created a VR-based cognitive-assessment and immersive-nostalgia therapy system for elderly cancer patients [27]. This system utilizes VR to evaluate cognition in patients with mild dementia. Through analyzing receiver operating characteristic (ROC) curves, it was established that a VR cognitive assessment score of 9.5 or lower indicates a risk of mild dementia. The application of VR technology in assessing mild dementia risk among elderly cancer patients has proven both feasible and safe.

Overall, VR technology has undergone extensive research in psychological assessment, covering aspects such as assessment identification and early mental health warnings. However, the application of VR in personality assessment is still emerging, with current assessments largely confined to paper and computerized questionnaires. This study aims to identify which personality traits are activated by specific VR scenarios and episodes, and to develop a corresponding assessment algorithm. This approach enhances the objectivity of the VR personality assessment system, enabling more accurate conclusions. Accordingly, a more sophisticated and objective VR-based system for student personality assessment is proposed. This system could significantly aid schools in conducting more effective personality assessments of their students.

3. Methodology

In this section, we describe the methodology of this paper, which focuses on the structured logic of the personality assessment framework, detailing the specifics of each layer.

3.1. Framework

Addressing the aforementioned research objectives and content, the organized logic of the personality assessment framework is depicted in an ascending manner, as illustrated in the accompanying Figure 1. This representation not only demonstrates the placement and function, but also the interconnectivity of critical technologies with in the entire research framework’s structure.

The personality assessment framework is constructed with four modular layers. At its heart lies the personality assessment model, which conveys the final personality assessment results to the user through a tripartite process: data input, personality evaluation, and the revelation of type results. Encircling the personality assessment model, the framework employs a VR device, serving as the hardware facilitator for scene portrayal and user interaction. Concurrently, it logs the user’s heart rate fluctuations throughout the interaction, through a portable monitor. Additionally, the framework engineers an evolving scenario aimed at evoking the four facets of personality traits, drawing upon the correlation between VR scenarios and personality trait mapping. Following the completion of tests, the critical data are inscribed into a database for retention. This repository aids in algorithm optimization and performance appraisal [28]. A data exchange occurs across the framework’s layers, whereby raw scenario data traverse through the foundational virtual engine to the user, for evaluation. The user-generated evaluation data in the scene are instantaneously relayed to the personality evaluation model for processing. Meanwhile, both the evaluative and analytical data, alongside the raw data, are archived in the database. The evaluative results are thus conveyed to the user through the VR device at the evaluation’s conclusion.

3.2. Personality Scene Mapping

Addressing this, this subsection delves into the design of the relationship between VR scenarios in the framework and the mapping of personality traits. It also involves the creation of recursive narrative scenarios intended to draw out the four dimensions of the user’s personality traits.

The MBTI test traditionally involves participants responding to a series of questions grounded in MBTI theory. These questions aim to decipher an individual’s behavioral tendencies and personality attributes [29]. This approach is instrumental in determining their preferences across four dimensions, thus leading to the identification of their specific MBTI type. The test culminates in providing comprehensive insights, including information regarding their characteristic behaviors, preferred communication styles, and potential career paths.

However, to reduce the effect of psychological biases and errors in data transmission in this conventional method, our team has innovated by integrating a VR scenario, moving away from the typical questionnaire-based quantitative scoring [30]. In this immersive VR setting, participants engage with scenarios that prompt the expression of their personality traits. Their choices during these interactions, along with their reflective responses and physiological responses, serve as critical factors in determining their personality type. This method employs personality assessment algorithms, transitioning from the traditional scoring scales. By dramatizing the MBTI personality test, we craft an assessment environment tailored to the traits of the four dimensions, with the following points guiding our scenario design.

• E or I.

Extroverts generally exhibit a preference for engaging with the external world, emphasizing social interactions and external stimuli. In contrast, introverts are more inclined towards introspection and independent thought, often favoring individual work and contemplation [31]. Our designed scenarios thus hinge on social dynamics and teamwork, assessing personality tendencies based on how individuals navigate social contexts and integrate into teams.

2.

S or N.

Sensing types focus more on practical details and concrete facts, leveraging their past experiences effectively. On the other hand, intuitive types focus more on overall understanding and future possibilities, and are adept at uncovering underlying meanings and patterns [32]. In designing scenarios for these dispositions, our emphasis is on strategic decision-making, exemplified by how participants react when faced with disorienting situations.

3.

T or F.

Thinking individuals prioritize logic and objective principles, often exhibiting more analytical and rational approaches in decision-making. Conversely, feeling-driven individuals focus more on values and the feelings of others, incorporating human and emotional aspects into their decisions. This is particularly evident in situations such as conflict resolution and patience evaluation, where one’s emphasis on logical correctness or the emotions of others becomes apparent.

4.

J or P.

The Judging type prefers structure and planning, with a preference for certainty and decisiveness. In contrast, the Perceiving type prefers flexibility and openness, with a preference for exploring and adapting to new information and situations. By employing adventure-themed scenarios, we can effectively assess these personality traits by observing the individual’s preparatory actions and decision-making processes.

3.3. Core Evaluation Indicators

Creating appropriate scenes and episodes is crucial for eliciting these personality traits [33]. To capture them accurately, we require specific parameter indicators. These indicators are vital components of our personality assessment framework, feeding into the model as key data for evaluation. In this section, we focus on the conceptualization of these three essential assessment parameters.

Cognitive behavioral theory posits a close interconnection among thoughts, emotions, and behaviors. In virtual reality settings, a user’s interactive choices (behaviors) offer insights into their thought processes and personality preferences. Moreover, drawing from Kolb’s experiential learning theory, the scenario-based MBTI test allows users to learn through active engagement rather than traditional questionnaire methods. This approach enhances the alignment of user choices with their personality traits, thereby increasing the accuracy of personality assessments.

Meanwhile, the concept of reaction time is more about reflecting a user’s speed of decision-making and accuracy of responses, as well as their performance under various contextual factors such as environment, stress, and emotional state. Individuals generally react quicker when they can efficiently and accurately identify their genuine thoughts in a brief time span [34,35]. This metric helps us distinguish between intentional and “unintentional actions” during the assessment process. For instance, an introverted person in a virtual reality setting might experience a psychological push towards extroversion, prompting them to make choices contrary to their actual tendencies. Such decisions could lead to inaccuracies in our personality evaluation. Hence, we incorporate reaction time to enhance the accuracy of the assessments.

The heart rate fluctuation metric employs objective physiological signals to respond to human emotions. Heart rate, skin conductance, and psychological states are interlinked, as evidenced by psychophysiological research, which demonstrates that emotional states can intensify or reduce the body’s bioelectrical activity [36]. We utilize this metric to connect a user’s physiological responses with situational stimuli. For instance, during an endurance test, heart rate fluctuations can reveal the user’s emotional condition at that moment.

Therefore, we have introduced interaction response, reaction time, and heart rate fluctuation as key evaluation indicators, symbolized by R, T, and H, respectively. The dataset for these evaluation criteria, i.e., factor set A = {R, T, H}, is derived.

3.4. Personality Evaluation Model

This subsection is an introduction to the core algorithm of the personality assessment model, which is categorized into two segments: the computational flow of the algorithm, and its computational methodology.

3.4.1. Evaluation Procedure

In this paper, the four-dimensional indicators of personality assessment are transformed into structured hierarchical and mathematical models, employing the design principles of the TOPSIS method [37]. This approach begins with forming an original data matrix by normalizing the assessment dataset. We then identify the extremes of each dimension using the cosine method. Following this, we calculate the distance between each evaluation subject and these extremes. This calculation helps in determining the relative closeness of each subject to these extremes, and forms the core of this study. A visual representation of the model’s structure is provided as Figure 2.

3.4.2. Evaluation Algorithm

Interaction response data, reaction time and heart rate fluctuation are the key data of the model. To integrate these elements, the study first employs the entropy method to assign appropriate weights to each type of data. These weights are crucial for calculating the proximity of different dimensions. The formula for calculating information entropy is as follows:

$$P_{ij}={\displaystyle \frac{N_{ij}}{{\sum }_{i=1}^n{N}_{ij}}}$$

(1)

where i = (1, 2, 3, …), j = (1, 2, 3), N_ij denotes the {R, T, H} three metrics of each interaction episode, and P_ij is the information entropy of the three metrics of each interaction episode; the weights are then calculated by substituting the information entropy for each episode, respectively, as Equation (2):

$$w_{ij}={\displaystyle \frac{1+\frac{1}{lnn}{\sum }_{i=1}^n(P_{ij}\times ln{P}_{ij})}{{\sum }_{j=1}^3[1+\frac{1}{lnn}{\sum }_{i=1}^n(P_{ij}\times ln{P}_{ij}\left)\right]}}$$

(2)

The construction of standardized matrices was carried out separately for different dimensions of data:

$$Z_{ij}={\displaystyle \frac{X_{ij}}{\sqrt{{\sum }_{k=1}^n{\left(X_{ij}\right)}^2}}}$$

(3)

where the number of episodes belonging to the dimension i = (1, 2, 3, …), j = (1, 2, 3), and X_ij denotes the three metrics of the interaction episodes {R,T,H} to which the dimension belongs. Then, the distance from the poles of the dimension is

$$D_i^{+}=\sqrt{\sum _{j=1}^m{w}_{ij}(Z_{ij}^{+}-z_{ij}{)}^2}$$

(4)

$$D_i^{-}=\sqrt{\sum _{j=1}^m{w}_{ij}(Z_{ij}^{-}-z_{ij}{)}^2}$$

(5)

where w_ij represents the weight (importance) of the jth attribute, and Z⁺ (Z⁻) is defined as the distance to the maximum. To understand the practical meaning of the D⁺ and D⁻ values, consider them as representing the distance between the evaluation subject and the two extremes. A larger value indicates a greater distance. For instance, in the E/I dimension, a larger D+ value for a subject suggests a closer alignment with the I pole, indicating a more introverted personality. Conversely, a larger D− value indicates a closer alignment with the E pole, signifying a more extroverted personality.

4. Prototype Design

In this section, we develop a VR prototype system dedicated to personality assessment, as outlined in Section 3. This development concurrently emphasizes the crafting of the assessment scenarios and interaction plot in the system. Illustrated in Figure 3, the plot scenes, initially conceptualized in the personality assessment framework, are concretely integrated into the prototype. This system is structured around three evolving virtual reality scenarios. Participants are immersed in these scenarios, engaging in various tasks that garner interactive feedback, thus facilitating the assessment process. Additionally, the system incorporates the HTC-Vive device. This inclusion leverages the handheld controller’s motor vibration, enhancing the experience by synergizing visual and tactile stimuli. The integration of Kinect motion capture, particularly in the fire scenario, allows users to more authentically replicate fire extinguishing maneuvers and escape actions.

4.1. Prom Image Matching

The way individuals process information through perception, memory, thinking, judgment, and problem-solving (i.e., cognitive processes) directly impacts their behavioral choices. These consistent patterns of behavior, shaped by cognitive processes, contribute to the development of individual personality traits [38]. Clothing functions as a symbolic medium, offering insights into the wearer’s inner psyche [39]. Self-assessed clothing measures have a significant correlation with numerous personality facets [40]. For instance, a keen interest in clothing is correlated with extroversion. Likewise, high levels of clothing satisfaction typically suggest that the individual is extroverted, content, and adept in social interactions.

We establish a pre-prom personal dress-up scenario, allowing participants to select their preferred attire while simultaneously choosing a compatible partner from a range of model characters we provide. Participants have the option of selecting from thirty model characters. Each of these models is correlated with unique selection values, closely correlated with four key personality traits. This segment of the design utilizes the selection of dance partners as an implicit proxy for an individual’s choice of a future romantic partner, aiming to reveal underlying personality traits. Additionally, we offer various customization options, including hairstyles, body types, and attire, enabling participants to tailor their choices freely [41]. This diverse range of choices allows the assessment system to evaluate personality traits related to the highly correlated dimensions of sensation seeking and intuition, as well as extraversion and introversion. As illustrated in Figure 4, participants manipulate controls to select their attire, with each choice reflecting specific personality traits, such as a gentle style being associated with the ISFJ personality type. From a clothing consistency perspective, extraverted individuals with an intuitive preference may exhibit higher scores on clothing consistency than extraverted individuals with a sensing preference. They are more likely to prioritize matching their attire to the occasion and their dance partner’s clothing. This approach to clothing conformity can help explain participants’ personality types. Moreover, we monitor and utilize changes in reaction time and heart rate during these selections as critical data for our personality assessment model.

For this scenario, we developed a mapping relationship between MBTI personality types and specific clothing characteristics. For example, in terms of color preferences, participants’ color choices were interwoven with their daily lives, with each color representing a particular trait. Individuals who prefer red tend to be strong leaders, fast-paced thinkers, and risk-takers, demonstrating high energy levels and a strong sense of responsibility. They often lack patience and struggle with ambiguity. Those who favor yellow are sociable, expressive, and passionate. They tend to be informal and energetic, prioritizing their goals and building strong relationships. Green personalities are friendly and casual in their communication. They may appear emotional in their relationships and social interactions. Blue-personality types are deep thinkers and analytical. While they are meticulous and detail-oriented, they often struggle with the challenges they face [42,43]. In addition to color, we also assigned weights to different clothing styles, shoe preferences, and fashion philosophies, based on MBTI characteristics. As shown in

Table 1. MBTI clothing-style weights.

	Classic	Sporty	Maximalism	Minimalism	Modern	Vintage	Street	Casual
ISTJ	0.0	0.0	15.9	15.9	15.9	0.0	16.7	36.3
ISFJ	12.2	0.0	0.0	12.2	16.7	12.2	0.0	46.6
INFJ	15.1	0.0	15.1	15.1	0.0	0.0	0.0	54.7
INTJ	0.0	0.0	0.0	36.5	0.0	27.0	0.0	36.5
ISTP	8.7	0.0	19.3	33.3	0.0	19.3	0.0	19.3
ISFP	0.0	0.0	0.0	0.0	100.0	0.0	0.0	0.0
INFP	49.6	0.0	14.5	0.0	14.5	0.0	0.0	21.4
INTP	0.0	0.0	0.0	37.5	25.0	0.0	25.0	37.5
ESTP	0.0	0.0	0.0	36.3	0.0	0.0	25.0	63.7
ESFP	0.0	0.0	33.3	0.0	0.0	33.3	0.0	33.3
ENFP	0.0	0.0	25.0	25.0	0.0	0.0	0.0	25.0
ENTP	0.0	25.0	0.0	25.0	0.0	0.0	0.0	25.0
ESTJ	0.0	0.0	0.0	0.0	0.0	0.0	0.0	100.0
ESFJ ENFJ ENTJ	100.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
ENFJ	0.0	0.0	50.0	0.0	0.0	0.0	0.0	50.0
ENTJ	0.0	0.0	0.0	33.3	33.3	0.0	0.0	33.3

, this table illustrates the varying preferences for specific styles associated with different MBTI personality types.

4.2. Social Ball

We set up a prom setting, designed to simulate a social environment that presents participants with challenges such as selecting a dance partner and performing onstage. These activities allow us to evaluate social characteristics and decision-making skills, thereby providing insights into traits such as extroversion, introversion, and judgmental versus perceptual tendencies. Participants interact with NPCs in this environment. This interaction, including dialogue choices and the manner of inviting a dance partner, subtly reflects the participant’s internal responses in various challenging scenarios. Within the conversational design, key evaluation points such as Favorite Music Genre, Hobby, and Favorite Entertainment Genre were strategically integrated into the dialogue flow. A feature mapping relationship, formatted as shown in Table 1, was established to link these attributes. We also incorporate a stage-performance segment, where participants are invited to perform onstage. Post invitation, the system records a range of choices and physiological responses in real-time. This phase of the research utilizes VR technology to investigate cognitive control during complex everyday tasks, aiming to identify personality-driven behavioral patterns. Findings indicate that distinct personality types manifest in observable differences in spatial behavior, such as standing position, crowd interaction, and gaze direction [44]. Furthermore, during social interactions like dance-partner invitations, self-introductions, and public performances, individuals exhibit unique psychological states, reflected in their eye movements, body language, and vocal responses. For external device utilization, we use Kinect devices to track and analyze participants’ movement patterns. This comprehensive evaluation, which includes assessing the completeness of movements, reaction times, and heart rate fluctuations during the performance, aids in the accurate assessment of personality dimensions. Figure 5 demonstrates the system’s capability to capture and replicate the participant’s movements in a VR setting.

4.3. Fire Escape

During the prom, an unexpected fire scenario is introduced to assess the personal decision-making abilities and logical reasoning of the students when faced with a crisis [45]. The system incorporates the cognitive abilities, behavioral patterns, and life experiences of the student users. We employ somatosensory interaction technology to control the Kinect device, transforming it into a hands-free gaming platform. We create a 3D simulation of a fire scene, featuring fire extinguishers and other firefighting equipment, along with escape routes, to replicate fire evacuation scenarios. Users utilize left- and right-hand gestures to command their character’s movement, while stationary gestures serve as commands for understanding and selecting items. This intuitive approach allows users to control their character’s actions. As depicted in Figure 6, they can place their hands on their chests to switch roles, extend their arms to mimic using fire extinguishers, and adjust their aim by moving left and right to extinguish flames.

If a user fails to extinguish the fire, they are prompted to escape from the scene and call for assistance. The virtual fire scenario also includes paths that are unexpectedly blocked by poisonous smoke, oxygen burning at a high temperature, or falling objects produced by fire. During the escape, users must make on-the-spot decisions. For instance, in the presence of heavy smoke, they need to cover their heads, crouch down, and move slowly along the walls. The user’s behavior during the escape is directly tied to their personality traits. For instance, we introduce a fork in the escape route to assess whether the user can select a rational path. Additionally, we monitor reaction times and heart rate fluctuations throughout the process, providing valuable insights for evaluating sensory and intuitive dimensions. Once trapped in the scene of a fire, participants are required to first identify their location and the surroundings, calmly. Since the most dangerous aspect of fire is not the flame but the poisonous smoke, participants cover their mouth and nose with handkerchiefs to avoid inhaling smoke and gases as quickly as possible, and then keep their body as low as possible to look for a way out. As depicted in Figure 7, a user is actively engaged in a fire escape, keeps a low posture, and covers his mouth and nose following the game instruction.

5. Experiment

5.1. Experimental Design

• A. Participants

The study enrolled 96 participants (50 males and 46 females), ranging in age from 10 to 22 years, who were recruited and successfully passed physical and mental-health assessments. Prior to the commencement of the experiment, all participants received a detailed briefing and provided their consent by signing a form. In line with Erikson’s theory of psychosocial development, the participants were categorized into three age-based groups: the children’s group (ages 10–13, totaling 34 participants), the adolescents’ group (ages 13–17, with 30 participants), and the youth group (ages 18–22, comprising 32 participants).

• B. Experimental process

Our research utilized a prototype system, as described in this paper, to evaluate the three groups over a three-week period. During this period, participants completed a personality test once a week. We requested that participants ensure adequate sleep prior to each test and remain seated calmly for 10 min before commencing the VR personality test. This preparatory period aimed to help them adjust to the setting and retain a clear mind. Our staff provided thorough guidance on the usage of VR equipment and the system’s operational instructions before the test. Participants then performed an immersive, 20 min personality test in VR, under staff supervision. Thereafter, they completed the standard MBTI questionnaire (a different version each week). Post testing, we collected data on user satisfaction, marking the conclusion of the testing procedure, as depicted in Figure 8.

Following the three-week testing phase, we conducted an analysis comparing the VR-based personality assessments with traditional MBTI questionnaires. This comparison, focusing on the data from the three age groups, aimed to evaluate the suitability and accuracy of VR scenario-based personality assessments in a student demographic.

5.2. Experimental Data Analysis

After conducting three experiments, we obtained experimental data and test results for 96 participants. These were utilized to perform an in-depth analysis of the personality test results in the VR setting.

Table 2. Task data statistics of first experiment.

Serial Number	Task A			Task B			Task C
	R	T(s)	H(bpm)	R	T(s)	H(bpm)	R	T(s)	H(bpm)
C1	44	193	106	78	355	86	91	321	84
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮
C34	65	176	76	81	362	74	71	362	91
A1	84	163	86	55	387	82	85	335	84
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮
A30	32	178	93	72	373	88	77	359	89
Y1	92	176	73	87	343	76	93	316	79
⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮	⋮
Y32	76	183	98	43	391	92	61	371	102

displays the first experimental data from three groups of participants: children (C1–C34), adolescents (A1–A30), and young adults (Y1–Y32). These groups were assessed in three different tasks (Task A\B\C). In this table, R stands for the score achieved by the participants in each scenario, T represents their cumulative reaction time, and H indicates the average variability in heart rate during the tasks. These three metrics form the cornerstone of our personality assessment algorithm for character evaluation.

Following the obtaining of this preliminary data, we processed it through a personality assessment model to derive MBTI personality profiles for each participant. To provide a comparison group, a traditional MBTI questionnaire was administered to all 96 participants after the VR experiment. This allowed for a direct comparison between the traditional questionnaire group and the VR test group. This comparison aimed to measure the suitability of our personality test for the student demographic in VR settings.

Table 3. Comparison of data between VR group and Questionnaire group.

Group	E/I	E/I(Q)	S/N	S/N(Q)	T/F	T/F(Q)	J/P	J/P(Q)
C-1	21/12	17/16	18/15	18/15	9/24	10/23	19/14	20/13
C-2	21/12	22/11	17/16	19/14	10/23	12/21	19/14	19/14
C-3	21/12	19/14	17/16	17/16	11/22	8/25	19/14	19/14
A-1	19/11	21/13	12/18	10/20	16/14	15/15	20/10	19/11
A-2	19/11	21/13	12/18	9/21	16/14	17/13	20/10	19/11
A-3	19/11	21/13	12/18	12/18	16/14	17/13	20/10	19/11
Y-1	13/19	14/18	18/14	18/14	20/12	20/12	9/23	10/24
Y-2	13/19	14/18	18/14	18/14	20/12	20/12	9/23	10/24
Y-3	13/19	14/18	18/14	18/14	20/12	20/12	9/23	10/24

shows the distribution of participants across four personality dimensions for three age groups in the personality assessment test. C-1, C-2, and C-3 represent the three test sessions for the children’s group, with 21/12 representing the number division of participants with respect to the Extrovert/Introvert dimension. We analyzed the changes in the three sets of experimental data between the experimental and questionnaire groups, based on the assessment results presented in the table. Figure 9 visualizes data for the three age groups, with the left side representing the traditional questionnaire group and the right side representing the VR test group. The x-axis represents the number of participants (with a scale of 5 participants per unit), and the y-axis represents the three test sessions for the four dimensions. Specifically, Figure 9b,c illustrate the comparative analysis of these two groups’ assessment data. This analysis highlighted the fact that the experimental and questionnaire data for the youth group were the most consistent, suggesting a higher overall cognitive capability and a more developed personality in this group. However, the questionnaire results for the youth group displayed minor discrepancies in specific dimensions. In contrast, the children’s group demonstrated slight changes in certain dimensions in the VR assessment. As depicted in Figure 9a, the questionnaire results for this group were more inconsistent and varied significantly, potentially due to cognitive limitations in the children, which hindered their ability to accurately interpret and respond to the textual information. This issue likely contributed to the larger discrepancies observed in their results.

Upon comparing the assessment results of the two groups, we notice a minor variance, without significant deviation across various dimensions. This suggests that employing VR technology for personality evaluations is a viable approach. VR technology primarily relies on situational interactions for these assessments, and demands less cognitive effort from participants, making it especially accessible to younger demographics. In addition, the implementation of personality tests in VR environments cleverly conceals the assessment mechanism. This approach effectively circumvents the psychological effect that typically accompanies the numerical scoring of traditional questionnaires, thus enabling users to make more authentic choices. Our findings suggest that VR technology offers a potential advantage over traditional questionnaires tests when assessing personality in younger students through scenario-based testing.

5.3. Model Accuracy Analysis

This section employs the “confusion matrix” as a model evaluation metric, providing insights into the correspondence between predicted and actual personality assessment results. Based on the confusion matrix, accuracy (ACC) can be calculated as a key metric for objectively evaluating the model’s predictive power [46]. This experiment calculated confusion matrices for four dimensions across three experimental groups. When ACC > 0.6, it is indicative of a reliable prediction for the dimension. The formula for calculating accuracy is as follows:

$$ACC={\displaystyle \frac{TP+TN}{TP+TN+FP+FN}}$$

(6)

Using the Extrovert/Introvert dimension as an example, TP (True Positive) represents instances where the model correctly predicts an individual as an “Introvert.” FP (False Positive) represents instances where the model incorrectly predicts an individual as an “Introvert” when they are actually an “Extrovert”. TN (True Negative) represents instances where the model correctly predicts an individual as an “Extrovert”. FN (False Negative) represents instances where the model incorrectly predicts an individual as an “Extrovert” when they are actually an “Introvert”. As depicted in Figure 10, the diagonal entries of the confusion matrix represent TP and TN for each personality dimension. Notably, the adolescent and young adult groups exhibited superior predictive accuracies compared to the children’s group, with the assessment model demonstrating an impressive performance, exceeding 80% accuracy in both the Extraversion and Introversion dimensions. However, the performance in the Thinking and Feeling dimensions was less satisfactory, barely meeting the expected standards. In addition, the children’s group demonstrated weak model predictions in the Judging and Perceptive dimensions. These findings highlight the necessity to optimize our VR scenarios and assessment models, particularly in the Thinking and Feeling dimension, by incorporating scenes with more accurate metrics. Simultaneously, there is a need to explore more appropriate scenarios that resonate with children, to better elicit and assess their personality traits.

5.4. Prototype System Evaluation

We integrated both the VR joystick and the Kinect device as the foundational interaction elements in our prototype. It is significant that most users were previously unfamiliar with these devices, potentially leading to a significant number of operational errors. These errors can lead to misinterpretations of participants’ behavior, resulting in false positives (incorrectly identifying a trait) or false negatives (failing to detect a trait) [47]. For instance, operational errors can distort the overall pattern of participant behavior, making it difficult to draw accurate conclusions about their personality traits. If a child consistently encounters technical difficulties, their frustration might be misconstrued as a lack of resilience or adaptability, when the true issue is simply a lack of familiarity with the system. Furthermore, high rates of operational errors can easily introduce bias into the data, potentially favoring certain personality traits over others, leading to an imbalanced assessment. To address this, we introduced the ‘operation error rate’ as a metric to measure the effectiveness of the system’s designed interactions [48]. This rate is calculated as the ratio of invalid or incorrect actions during the test. Children, in particular, might struggle with maintaining focus and could be easily distracted, contributing to operational errors [49]. A low operation error rate suggests that users are comfortable with the system and find the scenarios engaging. The method for calculating this rate is outlined in Equation (7).

$$E_M={\displaystyle \frac{(M_{max}-{\mu }_m)}{(M_{max}-M_{min})}}\times 100\%$$

(7)

where EM is the operation error rate, and μ_m represents the average number of operative errors on each experiment date. M_max and M_min are the maximum and minimum numbers of operative errors. Figure 11 presents a statistical curve representing the operation error rates observed in the experimental group. At first, these rates were high, indicating challenges in user–system interaction. However, post one evaluation test, there was a noticeable reduction in error rates. For instance, during the initial phase of the experiment, the children’s group had an operation error rate of 43.6%, which notably decreased to 13.6% in the final phase. This decline is indicative of growing user familiarity with the system’s operations and its reduced complexity.

We selected users from the experimental group and invited a panel of three psychologists to engage in a survey assessing user satisfaction. Satisfaction scores of the training items, emotional response and their own opinions were calculated. This survey utilized a 5-point scale, where scores from 1 to 5 correspond to varying degrees of satisfaction: very dissatisfied, dissatisfied, neutral, satisfied, and very satisfied, respectively. To ensure objective analysis, we utilized the survey data as a foundational metric and calculated the user-satisfaction conversion rate according to Equation (8).

$$S=0.25\times ({\mu }_s-1)\times 100\%$$

(8)

where S is the conversion value of users’ satisfaction, and μ_s represents the average of users’ satisfaction scores. The statistical results displayed in

Table 4. Statistics of users’ satisfaction.

Judges’ Number	Scene A	Scene B	Scene C
1	4	5	5
2	3	5	5
3	4	4	5
4	5	4	3
5	5	4	5
6	4	3	4
7	5	3	4
8	4	5	5
μs	4.25	4.125	4.5
S	81.25%	78.13%	87.50%

indicate that the satisfaction degree of each scene has reached 75%. This indicates that the VR-based personality assessment framework we developed aligns well with user expectations and demonstrates significant practicality.

6. Conclusions

This paper integrated the fundamental principles of the MBTI with VR technology, to create a VR-based personality assessment framework. This research found that the VR technology employed is versatile, catering to students of various ages and significantly minimizing the chances of inaccurate assessments due to limited cognitive abilities. In comparison to traditional MBTI questionnaire methods, our framework introduces novel assessment metrics and embeds the personality assessment in VR scenarios. This approach effectively reduces the psychological bias typically associated with quantitative evaluations. Overall, our framework addresses certain limitations in existing personality assessments and exhibits robust predictive accuracy.

While our framework enhances certain aspects of traditional questionnaires, further research is necessary to optimize the accuracy of categorization and the logic behind scenario design. It is important to acknowledge the limitations of this study. Firstly, the participant pool was relatively small, indicating a need for broader participation and extended observations to reinforce the study’s conclusions. This aspect forms a key avenue for future research. Secondly, the assessment criteria and algorithms require ongoing refinement to ensure more objective evaluations. To enhance the algorithm’s accuracy and robustness, we will incorporate more objective physiological metrics, such as EEG data and eye-tracking signals. This will involve reconstructing the evaluation algorithm to improve its resistance to interference. Concurrently, we will optimize the scene design by integrating scenarios tailored to student groups, enhancing the immersive experience and interactivity of the VR environment. This will facilitate the exploration of deeper personality traits among participants.