Interaction modality trade-offs in XR training: evidence from PC, virtual reality, and mixed reality in esports production

1 Introduction

Extended reality (XR), encompassing virtual reality (VR) and mixed reality (MR), has emerged as a promising approach for skills training due to its ability to provide repeatable practice, controllable scenarios, and safe exposure to complex or high-stakes procedures. Across domains such as clinical training, emergency response, and technical skill development, XR-based systems have been widely evaluated using performance outcomes, user experience, and usability measures (Tusher et al., 2024; Tokuno et al., 2025; Alruwaili et al., 2025). In parallel, research on interaction design highlights that modalities such as hand tracking, haptic feedback, and passthrough-based MR significantly influence presence, cognitive workload, and overall learning experience, positioning interaction modality as a key factor in training effectiveness (Kim et al., 2017; Kim, Rhiu, et al., 2019; Kojic et al., 2025). In XR research, immersion and presence are conceptually related but distinct constructs. Immersion generally refers to the technological characteristics of a system, whereas presence refers to the user’s subjective psychological sensation of “being there” within the mediated environment. Accordingly, immersion in this study is primarily discussed as an interface characteristic associated with different interaction modalities rather than as a directly measured psychological construct. From an interaction design perspective, these systems differ not only in immersion but also in how users interact with digital content, ranging from embodied interaction in VR, situated interaction in MR, to symbolic interaction in PC-based systems.

Despite these advances, existing studies often examine XR modalities in isolation or within domain-specific implementations, making it difficult to determine how different interface conditions compare under equivalent training scenarios. In particular, few studies provide a controlled comparison between PC-based systems, fully immersive VR, and mixed reality (MR) while maintaining consistent task structures and system functionalities. Furthermore, prior research typically focuses on a limited set of evaluation metrics, such as performance or user experience alone, rather than integrating cognitive (knowledge acquisition), behavioral (task performance), and experiential (engagement, usability, workload) measures within a single experimental framework. These limitations are particularly critical in esports event production training, where learners must coordinate multiple technical subsystems and execute procedural tasks under time pressure. In such contexts, the benefits of immersive interaction (e.g., increased engagement and presence) may be offset by usability challenges or increased cognitive workload. As a result, understanding the trade-offs among different interface modalities becomes essential for designing effective training environments.

To address these gaps, this study presents a controlled comparison of three interface modalities—PC (Personal Computer), VR, and MR—within a unified training system for esports event production. The study integrates multiple evaluation dimensions, including knowledge acquisition, task performance, user engagement (UES-SF), system usability (SUS), and cognitive workload (NASA-TLX), to provide a comprehensive assessment of training effectiveness across modalities. Accordingly, this study aims to answer the following research questions:

RQ1: How do different interface modalities (PC, VR, and MR) affect knowledge acquisition in esports event production training?

RQ2: How do interface modalities influence user engagement and cognitive workload during training?

RQ3: How do interface modalities affect task performance and perceived system usability during training?

RQ4: How can the observed differences in performance, engagement, usability, and workload inform the selection of interface modalities for real-world esports training environments?

2 Related work

2.1 XR-based training systems

XR training systems have been investigated in multiple professional contexts where repeated practice and controlled scenario design are critical. Early applied AR work demonstrated that overlaying step-by-step guidance on real assembly contexts can improve task completion time compared to classic methods (Hořejší, 2015). In health and emergency skill development, hybrid physical–digital setups, such as combining 3D-printed instruments with mobile AR, have been used to create sandbox environments for practicing procedural skills while retaining tangible affordances (Moreta-Martinez et al., 2020). More recent MR platforms extend this direction by enabling remote guidance, assessment, and “in situ” procedural rehearsal, which is particularly relevant for distributed training models (Tokuno et al., 2025; Sasaki and Lee, 2025).

Evidence from healthcare-oriented VR/MR training also highlights how system evaluation often includes not only objective performance but also usability, perceived usefulness, and readiness outcomes. For example, VR simulation has been used to train high-risk neonatal emergency response with competency outcomes reported via mixed methods designs (Alruwaili et al., 2025). MR training platforms have also been developed and piloted for nursing education, aiming to simulate complex clinical contexts with interactive supports (Sepanloo et al., 2025). In high-pressure team settings, mixed-reality training has been paired with data-driven assessment techniques (e.g., interaction analysis and action recognition) to evaluate trainee behavior at a fine-grained level (Mereddy et al., 2025). Similarly, MR interventions have been applied to specific technical procedures, showing improvements in learning outcomes and retention compared with traditional methods in controlled studies (Wang et al., 2025). Collectively, these findings position XR training as a credible approach for skill acquisition where realism, feedback, and measurable performance matter (Uhl et al., 2025).

2.2 Immersion and interaction modalities in XR

A consistent theme in XR research is that training outcomes depend heavily on immersion (e.g., presence) and the interaction modalities provided to users. Reviews from an ergonomics perspective indicate that user experience (UX) evaluation in VR has matured into a structured area, commonly incorporating measures of presence, workload, usability, and simulator sickness alongside performance metrics (Kim, Rhiu, et al., 2019). Within immersive environments, adding wearable haptics can extend sensory feedback channels and potentially improve skill learning by aligning perception with virtual task demands (Frisoli and Leonardis, 2024). Empirical studies also show that hand-haptic systems can increase perceived presence relative to hand-tracking-only interactions, emphasizing that interaction fidelity can directly affect immersion and user engagement (Kim et al., 2017).

Beyond haptics, multimodal interaction is increasingly positioned as a practical path to robust XR training experiences. Studies of XR user experience with combinations of hand tracking, voice input, and passthrough interaction suggest that modality combinations can shift both pragmatic usability and hedonic experience, with certain configurations preferred by users depending on task characteristics (Kojic et al., 2025). At the task level, “virtual grasping” research illustrates that design involves trade-offs: higher interaction fidelity does not automatically guarantee better performance across object affordances and task goals (Lee et al., 2025). In collaborative contexts, whole-hand interaction and vibrotactile feedback can influence presence, engagement, and workload compared to controller-based baselines (Moon et al., 2023). In parallel, VR storytelling and 360° video work continue to offer insights into engagement and presence constructs that inform how immersion can be designed and measured (Škola et al., 2020). Finally, learning analytics in immersive VR is emerging as a complementary lens, enabling multimodal evidence (behavioral and physiological) to connect interaction traces with learning indicators and evaluation frameworks (Tao et al., 2025; Zhang et al., 2025).

2.3 PC vs. VR vs. MR in training environments

Comparative research suggests that more immersive is not always better, and the appropriate interface often depends on locomotion constraints, task structure, and evaluation criteria. For spatial learning, desktop VR can outperform non-ambulatory HMD VR in certain conditions, underscoring that restrictions on movement and interaction can offset the benefits of immersive display (Srivastava et al., 2019). Similarly, user background factors can interact with interface choice; for example, studies comparing HMD and PC immersive training have examined how language proficiency may influence perceived UX and task outcomes across display modes (Pleva et al., 2020). These findings are particularly relevant for esports production education, where training cohorts can be diverse and where PC tool fluency may itself be an intended learning outcome.

In applied training contexts, evidence across healthcare and rehabilitation further indicates nuanced trade-offs among VR, non-immersive systems, and traditional simulation. Randomized and quasi-experimental comparisons show that VR may enrich the learning environment but should often complement rather than replace hands-on or mannequin-based methods for certain procedural skills (Setiawan et al., 2025; Agasthya et al., 2020). In rehabilitation and motor training, non-immersive VR additions have demonstrated practical effectiveness and accessibility, including reduced fall rates when combined with treadmill training (Mirelman et al., 2016), while immersive VR and other interventions can show superior gains in mobility outcomes depending on design and population (Lee et al., 2024). For technical and industrial-like training, VR simulations have been evaluated for learning gains and learner satisfaction in manufacturing education (Upadhyay et al., 2021), and workload prediction research has compared VR and PC flight-simulation conditions using physiological features (Verkennis et al., 2025). Together, these studies justify direct comparison among PC, VR, and MR interfaces when designing training for complex operational tasks (Alotaibi et al., 2015).

2.4 XR applications in esports and event production

While esports-specific XR training literature is still emerging, adjacent application domains provide actionable evidence for event-like production workflows, where scene understanding, procedural sequencing, and time-critical decision-making are central. Mixed reality (MR) systems designed for training complex investigations, such as scenario-based crime scene workflows, demonstrate how hybrid approaches can integrate physical context with interactive digital evidence and guidance. This is similar to the necessity for “in situ” cueing and situational awareness in production rehearsal (Albeedan et al., 2024). Industrial MR systems for ergonomic training further show that head-worn MR can support mobile, hands-free evaluation and real-time feedback, characteristics that are potentially transferable to production settings where operators interact with physical rigs, control surfaces, and spatially distributed stations (Hu et al., 2023).

At the broader education and training level, systematic reviews reinforce that immersive platforms can contribute to competence development in higher education and professional training, while also identifying design considerations for effective VR learning environments (Korbel et al., 2025). Reviews of immersive VR in education and training synthesize evidence about learning effectiveness and boundary conditions, supporting the rationale for using XR as a training medium when designed with appropriate instructional and evaluation strategies (Karachristos et al., 2025). In parallel, domain reviews comparing immersive VR with other digital interventions (e.g., serious games) highlight that “effectiveness” is multi-dimensional and may depend on accessibility, user comfort, and outcome measures (Zhao et al., 2025). Finally, as XR research continues to explore immersive and interactive learning across contexts, including primary and professional education, the design patterns and evaluation methods can inform esports event production training research, particularly when defining learning objectives, selecting interface modality, and measuring transfer to real operations (Quah et al., 2025). Building on this foundation, esports-oriented work that explicitly addresses immersive broadcasting and production training can serve as a direct conceptual bridge for structuring comparative studies of desktop, VR, and MR training interfaces (Margetis et al., 2024).

3 System design and training implementation

3.1 Operational context and system overview

The training system was developed based on the operational environment of the Northern Regional Esports Learning Center (NREL), an official esports training and research facility established under Chiang Mai University with support from the Digital Economy Promotion Agency (DEPA). The center aims to strengthen the regional esports ecosystem by developing human resources in esports event production and digital media operations. NREL serves as both a training venue and a live event space, equipped with a competition stage, large-scale display systems, audience seating, and production control stations for lighting, audio, and broadcast management. As such, it provides an authentic operational context where students and trainees can observe and participate in real esports production workflows, as illustrated in Figure 1, which shows both the live esports event setup and the broadcast monitoring interface used in real production. The spatial configuration and system structure of NREL were used as the reference model for designing the training environment in this study.

Building on this operational context, the system was developed as a simulation-based training environment to support structured learning of esports event production tasks. The environment was implemented across three interface conditions—PC (Personal Computer), Virtual Reality (VR), and Mixed Reality (MR)—to enable a controlled comparison of interaction modalities. All conditions deliver identical training content, task sequences, and system behaviors, differing only in how users interact with and perceive the environment. The system simulates key production subsystems, including lighting control, audio management, and broadcast monitoring, which are derived from real operational systems observed at NREL (see Figure 2). These subsystems are integrated into scenario-based tasks that reflect real operational workflows. This design allows participants to engage in task-oriented training while enabling the study to examine how different interface modalities influence learning outcomes, user experience, usability, and cognitive workload.

3.2 Training design principles

The training environment was designed to support experiential and interaction-driven learning through the integration of three core principles: spatial representation, workflow-based task design, and interactive feedback. These principles were derived from the operational characteristics of esports event production, where effective performance depends on understanding spatial configurations, coordinating actions across multiple subsystems, and responding to dynamic system states in real time. Spatial representation was implemented to reflect the layout of the esports venue, enabling participants to perceive how production components such as lighting, audio, and broadcast systems are distributed and related within the environment. This supports the development of contextual awareness, particularly in immersive conditions where users interact with spatially organized elements. In parallel, a workflow-based structure was applied to organize tasks as sequences of interdependent actions that reflect real operational procedures, requiring participants to monitor system states, identify issues, and execute corrective actions in a logically consistent order. This relationship is illustrated in Figure 3, which presents the conceptual integration of spatial representation, workflow logic, user interaction, and system feedback in supporting learning outcomes.

3.3 Training tasks and scenario structure

The training content was designed as a set of scenario-based tasks representing core operational activities in esports event production, with a focus on developing both conceptual understanding and procedural competence. Participants were required to engage with three primary subsystems—lighting control, audio management, and broadcast monitoring—through structured tasks that involve monitoring system states, identifying operational issues, and executing appropriate actions to achieve desired outcomes. Each scenario was constructed to reflect real-world production workflows, where actions must be performed in a logically consistent sequence and often require coordination across multiple subsystems. The tasks were organized with increasing complexity, beginning with basic system interaction and progressing toward multi-step problem-solving situations that require participants to interpret system feedback and adjust their actions iteratively. The detailed structure of the tasks, scenarios, and associated learning objectives is presented in Table 1. All tasks and scenarios were identical across the PC, VR, and MR conditions to ensure comparability, allowing the study to isolate the effect of interaction modality on learning outcomes, performance, and user experience.

TABLE 1. Training tasks and scenario structure for esports event production.

3.4 Interface modalities and interaction design

To examine the effect of interaction modality on training outcomes, the system was implemented across three interface conditions with equivalent functional capabilities but distinct interaction mechanisms. In the MR condition, virtual system elements were overlaid onto the physical environment using a video passthrough-based mixed reality headset, enabling users to interact with virtual controls anchored in real-world space while maintaining awareness of their physical surroundings (see Figure 4). In the VR condition, participants engaged with a fully immersive 3D environment using a head-mounted display and motion controllers, allowing direct manipulation of spatially distributed system components through embodied interaction (see Figure 5). In the PC condition, participants interacted with the system through a conventional 2D interface using mouse and keyboard input, where system information was presented via panels and control dashboards, requiring users to navigate between interface elements and interpret symbolic representations of system states (see Figure 6). Across all conditions, the underlying system logic, task structure, and functional mappings were kept consistent, with differences limited to how users accessed, navigated, and manipulated system elements. This design enables a controlled comparison in which interaction modality serves as the primary independent variable influencing learning outcomes, user experience, usability, and cognitive workload.

4 Research methodology

4.1 Research design

This study employed a controlled experimental design to examine the effect of interaction modality on training outcomes in esports event production. A between-subjects design was used, in which participants were randomly assigned to one of three interface conditions: PC-based, VR, and MR. The experimental design ensured that all participants received identical training content, task structures, and duration, with differences limited to interaction modality. The study aimed to evaluate how different interface conditions influence learning effectiveness, task performance, user experience, usability, and cognitive workload. To achieve this, multiple dependent variables were measured following the training intervention, including knowledge acquisition, task performance metrics, user engagement, perceived usability, and workload.

4.2 Participants

A total of 90 undergraduate students participated in this study. Participants were recruited from programs related to digital media, information technology, and communication studies at the university. All participants were novice learners with no prior knowledge or experience in esports event production, including lighting control, audio management, and broadcast monitoring systems. Participants were randomly assigned to one of three groups corresponding to the interface conditions: PC, VR, and MR, with 30 participants in each group. The sample size was balanced across conditions to support comparability in statistical analysis. The demographic characteristics of participants, including age, gender distribution, prior VR experience, and gaming background, are presented in Table 2. All participants provided informed consent prior to participation.

TABLE 2. Demographic characteristics of participants across interface conditions.

4.3 Measures

4.3.1 Knowledge of esports event production

Participants’ conceptual knowledge was assessed using a 20-item multiple-choice questionnaire covering esports event structures, production roles, and core systems (lighting, audio, and broadcast monitoring). The instrument was reviewed by three domain experts, achieving an Index of Item–Objective Congruence (IOC) of 0.82, indicating satisfactory content validity. The same questionnaire was administered as both a pre-test and post-test to measure learning gains. Each correct response was awarded one point, with higher scores indicating greater knowledge.

4.3.2 Task performance

Task performance was evaluated based on participants’ execution of nine scenario-based tasks across lighting, audio, and broadcast subsystems within a simulated esports production environment based on Arena of Valor (ROV). Performance was measured using three system-recorded indicators: task completion score, completion time, and error count. Task completion score reflected the final correctness of task outcomes, with each task receiving credit only if it was successfully completed by the end of the scenario. Scores were aggregated and converted to a percentage scale.

Error count captured the number of incorrect actions during task execution, including wrong selections, incorrect sequences, or suboptimal operations, and was recorded regardless of whether the participant eventually corrected the mistake. Examples of errors included activating incorrect lighting presets, selecting the wrong broadcast feed, adjusting inappropriate audio channels, or performing operational steps in an incorrect sequence during task execution. Thus, error count reflects the interaction process, whereas completion score reflects the final outcome. Completion time represented the total duration required to complete all tasks; however, as the tasks were embedded within fixed ROV gameplay sequences, time was largely consistent across conditions and was treated as a contextual rather than primary performance measure.

4.3.3 User engagement

User engagement was measured using the User Engagement Scale–Short Form (UES-SF), which assesses engagement across four dimensions: focused attention, perceived usability, aesthetic appeal, and reward (O’Brien et al., 2018). Participants rated each item on a Likert scale, and both subscale and overall engagement scores were computed by averaging across items. The UES-SF was selected to capture the quality of the training experience beyond learning outcomes, particularly in relation to how different interface modalities influence users’ attention, involvement, and perceived value of the activity. This is important in comparing PC, VR, and MR conditions, where differences may emerge in experiential engagement rather than performance alone.

4.3.4 System usability

System usability was assessed using the System Usability Scale (SUS), a widely used 10-item questionnaire that provides a global measure of perceived usability, including ease of use, learnability, and overall system effectiveness (Brooke, 1996). Responses were collected on a Likert scale and converted into a composite score ranging from 0 to 100, with higher scores indicating better usability. The SUS was included to evaluate how effectively each interface condition supports interaction with the training system. Since esports production tasks require users to navigate interfaces, interpret system states, and execute actions efficiently, usability is a critical factor that may influence both task performance and user experience.

4.3.5 Cognitive workload

Perceived workload was measured using the NASA Task Load Index (NASA-TLX), a widely used multidimensional instrument that evaluates workload across six dimensions: mental demand, physical demand, temporal demand, performance, effort, and frustration (Hart and Staveland, 1988). Participants rated each dimension after completing the training session, and an overall workload score was calculated. The NASA-TLX was used to assess the cognitive and interactional demands associated with each interface condition. This is particularly relevant in comparing PC, VR, and MR environments, as more immersive interfaces may increase engagement while also imposing greater cognitive or physical effort. Measuring workload therefore enables the examination of trade-offs between effectiveness, usability, and effort across different interaction modalities.

4.4 Experimental procedure

The experimental procedure was structured into three main phases to ensure consistency across all interface conditions, as illustrated in Figure 7. First, all participants attended a group-based orientation session lasting approximately 15 min. During this session, participants were introduced to the esports event production environment at the venue, including real equipment and system components such as lighting, audio, and broadcast monitoring. An overview of the operational workflow was provided to establish a common baseline understanding of the system context prior to the training.

Second, participants completed a pre-test assessing their conceptual knowledge of esports event production, as described in Section 4.3. Following this, participants were assigned to their respective interface conditions (PC, VR, or MR) and proceeded to the training phase. During training, participants were required to complete tasks across three production subsystems: lighting control, audio management, and broadcast monitoring. Each subsystem consisted of three scenario-based tasks, resulting in a total of nine tasks per participant. The tasks were completed sequentially, and the total training duration was approximately 2 hours. The content, task structure, and duration were identical across all conditions to ensure comparability. The VR and MR conditions were implemented using Meta Quest 3 head-mounted devices. In the VR condition, participants interacted with a fully immersive virtual environment using handheld motion controllers, whereas the MR condition used the passthrough mixed reality capability of the Meta Quest 3 to overlay virtual control elements onto the physical workspace. The PC condition was implemented using desktop computers equipped with Intel Core i7 processors, 16 GB RAM, and NVIDIA RTX 3060 graphics cards, with interaction performed via keyboard and mouse.

Third, upon completion of the training tasks, participants proceeded to the evaluation phase. Task performance data, including completion score, completion time, and error count, were automatically recorded during task execution. Participants then completed a post-test knowledge assessment and a set of questionnaires measuring UES-SF, SUS, and NASA-TLX, as described in Section 4.3. This structured procedure ensured that all participants were exposed to the same training content and evaluation process, with interaction modality as the primary independent variable.

4.5 Data analysis

The collected data were analyzed using both descriptive and inferential statistical methods. Descriptive statistics, including means and standard deviations, were used to summarize all dependent variables, including knowledge scores, task performance metrics, engagement, usability, and workload. To examine differences across the three interface conditions (PC, VR, and MR), one-way analysis of variance (ANOVA) was conducted for knowledge gain, task performance (completion score and error count), user engagement, and cognitive workload, with knowledge gain calculated as the difference between post-test and pre-test scores. Where significant effects were identified, post hoc pairwise comparisons with Bonferroni correction were performed to determine differences between conditions. Completion time was treated as a contextual variable due to the fixed duration of the gameplay scenarios and was therefore analyzed descriptively. SUS was also analyzed using descriptive statistics to provide an overall comparison across conditions. Baseline equivalence across groups was examined using pre-test knowledge scores. Statistical significance was set at p < 0.05, and effect sizes (η²) were reported for all ANOVA results. All analyses were conducted using standard statistical software.

5 Results

5.1 Results of knowledge gain

As presented in Table 3, the paired-samples t-test results indicated that all groups demonstrated statistically significant improvements in knowledge scores from pre-test to post-test (p < 0.001). The mean gain scores were 7.96 (SD = 4.37) for the PC group, 8.76 (SD = 5.49) for the VR group, and 7.43 (SD = 4.59) for the MR group, with all conditions showing large effect sizes.

TABLE 3. Paired-samples t-test results for within-group knowledge gain across conditions.

However, as shown in Table 4, the one-way ANOVA results revealed no statistically significant differences in knowledge gain among the three conditions, F (2,87) = 0.575, p = 0.565. This indicates that the magnitude of improvement was comparable across PC, VR, and MR groups. Consistent with this result, the post hoc pairwise comparisons reported in Table 5 confirmed that no significant differences were observed between any pair of conditions (p > 0.05).

TABLE 4. One-way ANOVA results for knowledge gain scores.

TABLE 5. Post hoc pairwise comparisons of knowledge gain scores.

5.2 Results of user engagement and cognitive workload

As illustrated in Figure 8, the mean user engagement scores differed across conditions, with the MR group reporting the highest engagement (M = 3.97, SD = 0.85), followed by the VR group (M = 3.77, SD = 0.74), and the PC group (M = 3.10, SD = 0.55). The one-way ANOVA results presented in Table 6 indicated a significant effect of interface modality on user engagement, F (2,87) = 11.935, p < 0.001. Post hoc pairwise comparisons in Table 7 revealed that both VR and MR groups demonstrated significantly higher engagement than the PC group (p = 0.002 and p < 0.001, respectively), while no significant difference was found between VR and MR (p = 0.531).

TABLE 6. One-way ANOVA results for UES-SF and NASA-TLX across interface conditions.

TABLE 7. Post hoc pairwise comparisons of UES-SF and NASA-TLX scores across interface conditions.

For cognitive workload, as shown in Figure 8, the VR condition yielded the highest workload (M = 3.17, SD = 0.60), followed by MR (M = 2.57, SD = 0.91) and PC (M = 2.43, SD = 0.99). The ANOVA results in Table 6 indicated a significant difference across conditions, F (2,87) = 6.540, p = 0.002. Post hoc comparisons further demonstrated that VR imposed significantly higher workload than both PC (p = 0.003) and MR (p = 0.018), whereas no significant difference was observed between PC and MR (p = 0.811), as reported in Table 7.

5.3 Results of task performance

As presented in Table 8, differences in task performance were observed across the three interface conditions. The VR group achieved the highest completion scores (M = 81.70, SD = 7.63), followed by the MR group (M = 81.36, SD = 5.34) and the PC group (M = 78.20, SD = 6.27). Completion time remained comparable across conditions (approximately 118–120 min), reflecting the fixed duration of the esports training scenarios. In contrast, clear differences were observed in error count, with the PC group recording substantially more errors (M = 10.80, SD = 2.36) than the MR (M = 5.80, SD = 1.88) and VR groups (M = 4.90, SD = 1.58). As illustrated in Figure 9, the PC condition also exhibited greater variability, indicating less efficient task execution.

TABLE 8. Mean (SD) of task performance metrics across conditions (PC, VR, MR). Lower values indicate better performance for time and error.

To further examine these differences, one-way ANOVA results (Table 9) showed that the effect of interface modality on completion score was not statistically significant, F (2, 87) = 2.661, p = 0.076. Post hoc comparisons (Table 10) confirmed that no pairwise differences between conditions were statistically significant. In contrast, a significant effect of interface modality was found for error count, F (2, 87) = 77.984, p < 0.001. Post hoc analysis revealed that the PC condition produced significantly higher error counts than both VR (MD = 5.90, p < 0.001) and MR (MD = 5.00, p < 0.001), while no significant difference was observed between VR and MR (MD = −0.90, p = 0.314). The effect size for error count was large (η² ≈ 0.64), indicating a substantial impact of interface modality on task accuracy.

TABLE 9. One-way ANOVA results for completion score and error count across interface conditions.

TABLE 10. Post hoc pairwise comparisons of completion score and error count scores across interface conditions.

5.4 Results of system usability

As illustrated in Figure 10, system usability scores varied across interface conditions, with the PC condition achieving the highest SUS score (77), followed by MR (75) and VR (68). All conditions fell within the “acceptable” usability range; however, PC and MR were positioned closer to the “good” category, while VR remained at the lower bound of acceptability. These results suggest that although immersive interfaces provide enhanced interaction experiences, they may introduce additional complexity that affects perceived usability. In contrast, the PC interface remains more familiar and easier to operate, resulting in higher usability ratings. Notably, the MR condition demonstrates a comparable level of usability to PC, indicating its potential to balance intuitive interaction with system efficiency.

6 Discussion

6.1 Effects of interface modality on knowledge acquisition

The results demonstrate that all interface modalities—PC, VR, and MR—produced statistically significant improvements in participants’ knowledge of esports event production, with no significant differences observed across conditions. In addressing RQ1, these findings suggest that variation in interaction modality does not inherently lead to differences in conceptual knowledge gains when instructional content, task structure, and exposure time are systematically controlled. From the perspective of this study, knowledge acquisition appears to be primarily driven by structured task engagement and the alignment of instructional design with learning objectives, rather than by the degree of immersion afforded by the interface. Although immersive modalities such as VR and MR provide enhanced spatial representation and embodied interaction, these affordances are more closely associated with procedural and experiential learning processes than with declarative knowledge acquisition. Accordingly, immersion should be understood as a mediating factor whose effectiveness depends on its alignment with the cognitive demands of the learning task.

This interpretation is consistent with prior research in immersive learning. Conrad et al. (2024) report that immersive VR can support learning but does not consistently outperform less immersive approaches in conceptual tasks. Similarly, Poupard et al. (2025) highlight that learning outcomes are influenced by factors such as cognitive load, task complexity, and instructional design, while De Witte et al. (2025) demonstrate that immersive environments can introduce additional cognitive demands that may constrain learning effectiveness under sustained task conditions. Taken together, these findings suggest that the relationship between immersion and knowledge acquisition is not linear, but contingent upon how well the interface supports the cognitive processes required by the task. In this context, conceptual learning relies more on the processing of structured information than on the specific interaction paradigm, whether embodied, situated, or symbolic.

6.2 Engagement and cognitive workload dynamics

The findings reveal that interface modality has a significant effect on both user engagement and cognitive workload. In addressing RQ2, the results indicate that immersive conditions (VR and MR) led to significantly higher levels of engagement compared to the PC condition, while VR also imposed the highest level of cognitive workload among the three modalities. These results suggest that increased immersion may contribute to higher experiential engagement during training activities; however, this benefit is accompanied by elevated cognitive demands, particularly in fully immersive environments. From the perspective of this study, engagement and cognitive workload appear to be inherently interrelated, reflecting a trade-off in which richer interaction and sensory immersion may simultaneously intensify user involvement and mental effort. Notably, MR demonstrated a relatively balanced profile, achieving high engagement without a corresponding increase in workload, indicating its potential to support immersive interaction while maintaining manageable cognitive demands.

This pattern is consistent with recent findings in immersive learning research. For instance, Tang et al. (2026) show that immersion plays a mediating role in enhancing engagement and cognitive performance, while its effectiveness is influenced by factors such as training duration and cognitive load. At the same time, De Witte et al. (2025) demonstrate that immersive VR environments can impose higher cognitive load compared to conventional training, which may negatively affect learning outcomes under certain conditions. Similarly, Grecu et al. (2025) report that high levels of immersion can introduce additional cognitive costs, particularly for novice learners, highlighting the importance of cognitive load-aware instructional design. Taken together, these studies support the interpretation that immersive interfaces enhance engagement but also increase cognitive demands, reinforcing the need to balance experiential richness with cognitive efficiency. Consequently, the effectiveness of an interface modality should be understood as a balance between engagement benefits and cognitive demands, rather than being determined by immersion considered in isolation. From an XR interaction perspective, the higher engagement observed in immersive conditions may be related to embodied and situated interaction characteristics that support richer experiential interaction during task performance. However, these interaction forms may also increase cognitive load due to the additional sensorimotor and spatial processing required.

6.3 Task performance and usability efficiency

The results indicate clear differences in task performance and perceived usability across interface modalities. In addressing RQ3, immersive conditions, particularly VR, demonstrated lower error rates compared to the PC condition, indicating lower error frequency during procedural task execution. Completion scores were comparable across conditions, with no statistically significant differences observed. MR showed a similar performance profile to VR, while the PC condition exhibited higher error counts despite identical task structures. However, these performance differences were not reflected in usability ratings. The PC interface achieved the highest System Usability Scale (SUS) score, followed by MR, while VR received the lowest usability rating. These findings suggest that immersive environments may support procedural interaction through spatially aligned interaction and direct manipulation, but may also introduce interaction complexity that reduces perceived usability.

This interpretation is consistent with prior research on immersive procedural training and XR-based skill development. Jongbloed et al. (2024) report strong benefits of immersive training for procedural learning and transfer, while Azzam et al. (2025) show that MR provides a more intuitive user experience than VR with comparable performance. Similarly, Lang et al. (2026) demonstrate that MR-based training supports effective task execution in applied settings. Taken together, these studies suggest that interface modality should be evaluated in terms of both performance and usability. From an XR interaction perspective, improved task accuracy observed in VR and MR may be associated with embodied and situated interaction characteristics that enable more direct alignment between actions and spatial representations, potentially reducing the need to interpret abstract symbolic representations commonly found in PC-based interaction.

6.4 Implications for training design

Taken together, the findings across RQ1–RQ3 suggest that the selection of interface modality for esports event production training should be guided by instructional priorities rather than the assumption that greater immersion is inherently superior. In addressing RQ4, PC, VR, and MR each offer distinct advantages and constraints. PC is most suitable when usability, familiarity, and ease of interaction are prioritized, particularly for novice learners or early-stage conceptual instruction. VR may be advantageous for procedural rehearsal, spatially oriented tasks, and experiential engagement, although it introduces higher cognitive workload and lower perceived usability. MR demonstrates a more balanced profile, combining strong task performance and engagement with manageable workload and acceptable usability. These patterns indicate that interface modality should be treated as a strategic design variable, where PC supports foundational learning, VR supports procedural rehearsal, and MR supports context-rich training requiring both performance efficiency and usability.

This interpretation aligns with prior research emphasizing that XR training design should be objective-driven and context-sensitive. Bödding et al. (2025) highlight the effectiveness of MR while noting limited comparative evidence across immersive conditions, and Dincelli (2026) emphasizes that VR should be designed in relation to training goals and user characteristics. Similarly, Yudintseva (2024) found no significant difference in learning outcomes between immersion levels, despite higher engagement and cognitive load in more immersive conditions. From an XR interaction perspective, these findings suggest that effectiveness depends not only on immersion but also on the characteristics of the interaction paradigm. VR may support embodied interaction for procedural and spatial tasks, MR may support situated interaction for context-aware operation, and PC relies more heavily on symbolic interaction, which may be more suitable for analytical tasks and early-stage learning. Together, these results highlight that effective XR training design should align interaction affordances with learning objectives, task complexity, and user readiness rather than maximizing immersion alone.

6.5 Limitations and future work

Several limitations should be considered when interpreting the findings. First, the study involved a relatively homogeneous sample of undergraduate students who were novice learners in esports event production. While this supports internal validity, it may limit the generalizability of the results to professional or expert populations, where prior experience and domain knowledge could influence interaction efficiency and learning outcomes. Second, although the training tasks were designed to reflect real-world workflows, the study was conducted in a simulated environment, which may not fully capture the complexity, time pressure, and coordination demands of live esports production settings. Third, although inferential statistical analysis was conducted for key task performance metrics, the analysis was limited to aggregate measures (completion score and error count) and did not incorporate more fine-grained behavioral or temporal performance indicators. In addition, the study focused on short-term learning outcomes and did not examine knowledge retention, transfer of training, or long-term skill development. Finally, prior familiarity with different interaction modalities was not explicitly measured, which may have influenced usability perceptions and workload ratings. In addition, constructs such as situational awareness, attentional involvement, and psychological presence were discussed as possible explanatory mechanisms related to immersive interaction; however, these variables were not directly measured in the present study. Therefore, these interpretations should be considered inferential rather than conclusive.

Future research should extend this work in several directions to strengthen both theoretical understanding and practical application. Longitudinal studies are needed to examine retention and transfer effects, particularly to determine whether the performance advantages observed in immersive conditions translate into sustained operational competence in real-world environments. Future studies should also include more diverse participant groups, such as experienced users and professional practitioners, to better understand how expertise interacts with interface modality. Future studies should also incorporate validated measures of presence, attention, and situational awareness to better examine the mechanisms underlying XR training outcomes. In addition, integrating multimodal data sources—including behavioral logs, physiological measures (e.g., cognitive load indicators), and observational performance assessment—could provide a more comprehensive understanding of user interaction and learning processes. From a design perspective, future work should explore adaptive or hybrid training approaches that combine multiple interface modalities, leveraging the strengths of each (e.g., PC for foundational learning, VR for procedural rehearsal, and MR for contextualized application). Further investigation is also needed to optimize interaction design in immersive environments, particularly in reducing unnecessary cognitive load while preserving engagement and effective task performance.

7 Conclusion

This study compared PC, VR, and MR interfaces for esports event production training across multiple dimensions, including knowledge acquisition, engagement, cognitive workload, task performance, and usability. The results show that all modalities effectively support knowledge gain, with no significant differences observed across conditions. Immersive modalities (VR and MR) were associated with higher engagement and significantly lower error rates, although VR also introduced higher cognitive workload and lower perceived usability. In contrast, the PC interface provided higher usability and lower workload but was associated with lower engagement and higher error frequency during procedural task execution. These findings indicate that no single modality is universally optimal; rather, the effectiveness of a training system depends on aligning interface modality with learning objectives, task requirements, and user characteristics. By highlighting trade-offs among task performance, engagement, usability, and workload, this study provides practical guidance for the design of XR-based training systems that are both effective and context-appropriate.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frvir.2026.1856354/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Ramy Hammady, University of Southampton, United Kingdom

Reviewed by: Geovanny Francisco Ruiz Muñoz, University of Guayaquil, Ecuador

Yanfang Luo, The Hong Kong Polytechnic University, China