Abstract: Extended Reality (XR) is increasingly used for training in military and emergency response contexts, yet its pedagogical application to high-risk and cognitively demanding environments remains underdeveloped. This paper argues that XR’s primary value lies in its ability to regulate cognitive load through staged, elemental training rather than holistic simulation.
Problem statement: How can Extended Reality technologies be structured to train military personnel and first responders effectively, while avoiding the pitfall of overwhelming trainees with complex scenarios?
So what?: Effective XR training for high-stakes domains requires decomposing elemental skills and employing staged progression rather than immediate full-scenario immersion. Training systems must regulate cognitive demands while maintaining high contextual coherence to ensure valid skill transfer. This has direct implications for XR system design, procurement decisions, and cross-organisational standardisation efforts.
A Transformative Force
Extended Reality (XR) has become a transformative force in serious games and training applications.[1] Through immersive simulations, XR provides platforms for practising skills and enhancing decision-making. However, simply placing trainees in high-fidelity simulations often leads to cognitive overload.[2] In a typical XR training session, a trainee wearing a head-mounted display is immersed in a virtual scenario—such as a tactical engagement or a mass-casualty incident—and must perceive, decide, and act in real time while the system records performance data for subsequent review. The fidelity of these simulations can range from screen-based desktop trainers to fully immersive mixed-reality environments incorporating physical props and haptic feedback.
This pedagogical challenge is particularly acute in operational contexts characterised by volatile, uncertain, complex, and ambiguous conditions, where learners must simultaneously manage high-stakes decision-making and physiological stress.[3] Yet precisely because these scenarios are too dangerous or complex to replicate physically, these immersive technologies are particularly beneficial to ChE. Challenging Environments is, in line with Skarbez, Smith, and Whitton, an umbrella term for high-stakes military and emergency response domains,[4] such as combat operations, CBRN (Chemical, Biological, Radioactive, Nuclear) incidents, firefighting, and law enforcement interventions, that are situational, difficult to anticipate, and potentially harmful XR technology bridges this gap by offering safe yet realistic training grounds. This work defines ChE, maps training approaches to the XR framework, and identifies the main benefits and challenges of using XR for specialised training.
The Pedagogical Imperative of Challenging Environments
The Challenge-Threat Mechanism
To understand why traditional holistic simulation often fails for trainees, it is necessary to first define the cognitive mechanics of the operational environment. The Biopsychosocial Model of Challenge and Threat (BPSM), integrates biological, psychological, and social factors to explain how individuals appraise and respond to motivated performance situations.[5] This model serves to define ChE not merely by their physical danger, but by their impact on trainee cognitive resources.[6]
The BPSM posits that when an individual confronts a high-stakes task, they instantaneously evaluate situational Demands against their personal Resources.[7] In this context, Demands refer to the situational and task-specific requirements imposed on the individual, such as time pressure, operational complexity, or threat exposure, while Resources encompass the individual’s available cognitive capacities, prior training, coping strategies, and domain expertise. This evaluation triggers one of two distinct physiological states, arising from intertwined Biological, Psychological, and Social Factors.[8]
When Resources outweigh Demands (Resources > Demands), the individual enters a Challenging State, characterised by improved cardiac efficiency and heightened cognitive focus; when Demands exceed Resources (Demands > Resources), a Threat State ensues — a stress response that actively inhibits decision-making and cognitive function.[9]
When Demands exceed Resources, a Threat State ensues — a stress response that actively inhibits decision-making and cognitive function.
This paper defines Challenging Environments (ChE) as operational contexts specific to the trainee, specifically within military and first-responder domains, characterised by seven attributes that inherently impose maximum Demands on the user:
- Situated: Challenges deeply rooted in specific, real-world circumstances;
- Stressful: Inducing physiological or psychological tension;
- Complex: Featuring multiple, unpredictable variables;
- High Risk: Mistakes result in severe consequences;
- High Stakes: Profoundly consequential outcomes;
- Interactive/Changing: Dynamic conditions requiring real-time adaptability; and
- Social/Communicative: Involving multiple agents demanding effective collaboration.
In these environments, biological stress responses and social pressures (e.g. command hierarchy) interact to rapidly deplete a trainee’s cognitive resources.[10]
The Training Gap: Why Elemental Approaches are Necessary
The specific training goals in these domains–ranging from rigid Standard Operating Procedures (SOPs) and tactical communication to complex decision-making–require the internalisation of “automated behaviour”.[11] Approach in XR has increasingly converged toward complex systems–comprehensive simulations designed to capture the full spectrum of operational requirements of the domain (Knowledge of processes and rules,[12] cognitive and emotional skills relating to environmental threats,[13] technical skills,[14] physical skills,[15] and communication[16]).
However, while high-fidelity simulation is required for contextual realism,[17] placing a trainee directly into a complex ChE scenario often causes Demands to overwhelm Resources. This triggers a Threat State, where learning shuts down in favour of survival mechanisms. Therefore, effective XR training cannot simply replicate the environment; it must structurally manipulate the Demands to align with the trainee’s cognitive Resources.
This necessitates a shift from purely holistic simulation to Elemental Training–a part-task approach where complex competencies (Technical Skills, Communication, SA) are isolated and mastered to automaticity.[18] By building these Resources in isolation, it is ensured that when the trainee is eventually exposed to the full Challenging Environment, they remain in a “Challenge State” rather than succumbing to threat.
The XR Framework: Mechanisms for Cognitive Control
The Dials of Demand: IM, EWK, and Coherence
Having established that ChE impose a high intrinsic cognitive load–the mental complexity arising from the operational task itself– the role of the training apparatus shifts. The XR system is not merely a simulation tool but a “Cognitive Load Management System”, conceptually aligning with Cognitive Load Theory’s distinction between intrinsic, extraneous, and germane load.[19] Cognitive Load Theory posits that working memory has limited capacity, and that instructional design must manage three types of load: intrinsic load imposed by the task itself, extraneous load generated by suboptimal instruction, and germane load dedicated to schema construction and skill automatization.[20] This mapping represents a conceptual framework warranting further empirical investigation. It acts as a dynamic valve, allowing instructional designers to suppress situational demands (Extraneous Load) while trainees build automatized resources (Germane Load).[21]
To operationalise the stated control and assess the achievable training goals for XR, a clear definition and classification are needed. The present paper builds on the framework proposed by Uhl et al., which adapts the XR dimensions[22] originally introduced by Skarbez, Smith, and Whitton. Skarbez et al. propose three fundamental XR dimensions, which were subsequently refined and contextualised by Uhl et al. for immersive training scenarios.[23]
This framework provides three distinct dials that align directly with the segmenting principle of multimedia learning, which states that complex material should be broken into learner-paced segments to manage essential processing, thus allowing complex operational tasks to be decomposed and mastered elementally.[24] The first dial, Immersion (IM), refers to the technical system’s fidelity in providing sensory input as it would in reality, relating to Slater’s Place Illusion, to be understood as the subjective sensation of physically being present in the virtual environment.[25] By manipulating IM, instructional designers conceptually control aspects of Exogenous Load – removing visual distractions (low IM) to focus on procedure[26] or adding “fog of war” (high IM) to test resilience. Crucially, this relationship is non-linear; while high IM can support Germane Load (processing that contributes to skill acquisition) by providing valid context, for novices, this sensory density often manifests as excessive Extraneous Load (processing unrelated to the learning objective) that must be managed.[27] Building on this, the second dial, Extent of World Knowledge (EWK), describes the system’s awareness of its surroundings—whether the simulation has environmental or object models. This dimension controls Germane Load related to psychomotor skills; low EWK forces abstract thinking, while high EWK supports embodied action and the development of muscle memory in realistic contexts.[28] The third and perhaps most critical dial is Coherence, which refers to the simulation’s internal logic. For the “Load Manager”, the XR system itself, which regulates training intensity through its dimensional properties rather than a human instructor, to function, the simulation must behave consistently; in cognitive load terms, violations of coherence introduce unnecessary extraneous load by forcing learners to process glitches and contradictions instead of task‑relevant information. This is especially central for occupational training: virtual environments and events must be plausible in respective contexts. This is considered an inherent necessity for virtual training in ChE, as it closely relates to the need for context-specific realism. For medical first responders, virtual training requires medical realism like bleeding wounds leading to unconsciousness if untreated, procedures requiring specific time, and wrong treatment worsening the patient’s state. This fidelity of context-specific realism varies by occupation but remains essential for plausible, coherent scenarios. Incoherence of a system furthermore creates Extraneous Load, forcing the brain to process glitches rather than learning objectives.
In cognitive load terms, violations of coherence introduce unnecessary extraneous load by forcing learners to process glitches and contradictions instead of task‑relevant information.
Topology of Task Decomposition
Task decomposition, in this context, refers to the systematic process of breaking down complex operational competencies–such as a full tactical engagement–into discrete, individually trainable skill elements that can be mastered in isolation before recombination. Following ‘strong concepts’ principles,[29] theory relates to practice by mapping main training goals onto the XR framework. In the focused Elemental Training context, these are not competing technologies, but staged pedagogical zones that allow the instructional designer to ramp up Demands only as Resources increase
- Quadrant 1 (Low IM / Low EWK)– The Procedural Zone: Screen-based training for logistics, organisation, and communication. Trainees build cognitive Resources (learning the steps of a 9-Liner or SOP) without the Demands of physical stress. The probability of entering a “Threat State” is minimised here, thereby maximising procedural retention.[30]
- Quadrant 2 (Low IM / High EWK) – The Contextual Zone: AR training with motion-tracking enables real skills training with environmental augmentation, lowering Demand on working memory to allow the trainee to focus on physical manipulation.[31]
- Quadrant 3 (High IM / Low EWK) – The Psychological Zone: Conventional VR for decision-making in simulated environments. By removing physical consequences, the Demands of visual chaos can be safely spiked to train decision-making before physical risks are introduced.
- Quadrant 4 (High IM / High EWK) – The Integrated Zone: MR with haptic feedback for technical skills in immersive environments[32] and decision-making with crucial contextual cues. This is the final stage where Demands (Chaos + Physicality) mimic the ChE. Trainees should enter this zone only after the component skills have been automated in Q1–Q3, ensuring they remain in a “Challenge State” rather than a “Threat State”.[33]
The Necessity of High Coherence
Across this topology, task decomposition strategies serve as the unifying thread. Whether simulating a radio call in a desktop trainer (Q1) or a kinetic breach in mixed reality (Q4), the underlying “elemental” skills—such as communication protocols—remain consistent. This allows trainees to master individual schema in lower-complexity quadrants before progressing to integrated, high-fidelity scenarios in Quadrant 4.[35]
However, a critical distinction must be made regarding the manipulation of these variables. While instructional designers can and should reduce Immersion or World Knowledge to manage cognitive load during early training phases, Coherence (contextual plausibility) must remain high across all quadrants.[36]
Even in a simplified Q1 desktop trainer, the tactical doctrine and physics logic must be flawless. If the elemental training teaches a procedure that conflicts with real-world ballistics or Standard Operating Procedures (low coherence), it creates negative transfer.[37] In this scenario, the trainee internalises a falsehood that must be “unlearned” cognitively under stress, actively endangering them in real-world environments.[38] Thus, while a simulation need not look or feel exactly like the real world, it must behave exactly like it.
From Isolation to Integration
The Elemental Training Approach
To address the pedagogical challenge of cognitive overload under volatile, uncertain, complex, and ambiguous conditions, this paper proposes integrating Elemental Training into the XR framework. Operational success in ChE relies on the automatisation of fundamental skills to free up cognitive resources for high-stakes decision-making.[39] Elemental Training in XR is operationally defined as the systematic isolation and progressive mastery of discrete skill components before integration into complex scenarios. This approach shares foundations with micro-learning and part-task training but extends these concepts to XR-mediated instruction, where dimensional properties (IM, EWK, Coherence) enable precise manipulation of training complexity.[40]
By decomposing complex procedures into constituent elements, XR allows trainees to master individual psychomotor and cognitive schemas in controlled, lower-complexity settings (Quadrants 1–3 of the XR Framework) before integrating into full operational scenarios in Quadrant 4. This specifically addresses the Biopsychosocial Model dynamics: by increasing the trainee’s Resources through automatization, the likelihood of a threat state when Demands increase is reduced.[41] This approach is particularly effective in military and first-responder domains, where complex, multi-component competencies must be developed systematically to prevent trainees from becoming overwhelmed.[42]
From Decomposition to Concrete Application Examples
To illustrate the practical application of elemental training, four critical areas of officer education are examined where XR offers distinct advantages over traditional methods. These examples illustrate how specific skills can be isolated and trained to automaticity:
The first area concerns mechanical automaticity, as applied to weapon handling and mechanics. Before a recruit proceeds to a live-fire range, Mixed Reality (Quadrant 4) with haptic feedback enables rigorous training in elements such as posture accuracy and handling mechanics, including stoppage drills, magazine changes, and sight alignment.[43] Closely related is the automaticity of procedural communication skills (e.g. the “9-Liner” as radio/reporting protocol). Complex reports, such as Call for Fire or MEDEVAC, require strict adherence to syntax under pressure. XR systems isolate this element by tasking trainees to recite reports under induced cognitive load, with AI-controlled Virtual Instructors verifying syntax in real-time until the trainee’s performance is flawless, thereby minimising cognitive resources required before full scenario integration. Thirdly, XR significantly enhances training for Room Clearing Procedures (e.g. CQB/MOUT). Entering and clearing rooms requires precise sectorization and team movement. Unlike physical training, XR enables visualisation of movement paths and lines of sight via Eye-Tracking. In the After Action Review (AAR), commanders can see precisely which angles remained unsecured, providing objective data for what was previously a subjective assessment.
Complex reports, such as Call for Fire or MEDEVAC, require strict adherence to syntax under pressure
Finally, XR provides a platform for tactical decision-making where trainees issue orders while managing multiple information channels (C2, radio, battle-chat). XR’s distinct advantage is its ability to dynamically modulate complexity, keeping training within the zone of proximal development.
Resulting Benefits
When Elemental Training is implemented within high-fidelity XR, several distinct advantages emerge. These benefits are analysed specifically through the lens of the previously established XR Framework.
Safe Decoupling of Risk and Realism: First-responder and military deployments often occur in non-permissive, hostile, or otherwise hazardous environments (e.g., mines,[44] construction sites,[45] CBRN conditions,[46] fires[47]). Traditional training faces a fundamental tension: authentic skill development requires exposure to realistic stressors, yet physical danger precludes meaningful failure without serious consequences. XR decouples psychological realism from physical risk. In Quadrants 3 and 4, the Plausibility Illusion triggers valid stress responses without actual harm. Safe failure opportunities allow trainees to build Resources (procedural confidence, stress inoculation) necessary to maintain Challenge States during actual deployment. Furthermore, immersive virtual training enables realistic scenario participation, fostering a sense of presence and the transfer of acquired skills to real-world scenarios.[48]
Traditional training faces a fundamental tension: authentic skill development requires exposure to realistic stressors, yet physical danger precludes meaningful failure without serious consequences.
Repeatability and Experimental Control: XR enables precise repetition without material wear—a key advantage for automation development.[49] This is an advantage for training effectiveness, as repetition is at the root of learning and fosters automaticity.[50]
Digital simulation (Quadrants 3 and 4) ensures the extent of World Knowledge (EWK) remains constant, guaranteeing that every trainee faces the exact same set of variables, which is essential for standardised competency measurement. By enabling high-frequency, low-cost repetition of elemental skills in lower quadrants before integration, XR directly supports the transfer of cognitive Resources from controlled to automatic processing—the prerequisite for maintaining performance under the elevated Demands of operational environments.
Training Personalisation through Adaptive Virtual Environments: Within XR simulation, optimal scenarios balance challenge with engagement. Monitoring physiological signals (ECG, EDA, eye tracking) enables real-time trainee state analysis.[51]
Adaptive Virtual Environments (AVEs) alter content, stimuli, or tasks based on trainee performance and objectives,[52] e.g. detecting when a trainee is approaching a Threat State and reducing Demands (e.g., removing distractors, slowing scenario pace). This adaptability manifests through flexibility to adjust complexity, pacing, and content of training scenarios in real-time. Advanced systems, such as those incorporating physiological monitoring and scenario engines that are fully controllable by trainers or intelligent AI models, can adapt the sequencing and granularity of skill elements, determining when trainees are ready to progress from isolated component practice to integrated scenarios.[53] This ensures learners remain within the optimal BPSM zone—a capability not possible in unmediated reality.
Interoperability: Interoperability enables a single virtual environment to train diverse roles effectively. The military Synthetic Training Environment (STE) seeks to replace non-interoperable programs with a unified platform.[54] Many institutions, despite having fundamentally similar objectives, pursue standalone developments. Establishing central platforms where organisations collaboratively exchange digital assets could significantly reduce development costs and improve training quality.
Key Challenges
Several practical and technical challenges must be addressed for XR training to fulfil its potential in operational contexts. At the organisational level, resistance to new technology requires identifying organisational ‘champions’ and securing ongoing support from technology partners. Beyond organisational inertia, achieving effective remote collaboration presents a distinct technical hurdle: while XR can, in principle, enable joint training across distributed locations, multi-user synchronisation across networks remains technically demanding, particularly for latency-sensitive tactical coordination exercises.[55]
While XR can, in principle, enable joint training across distributed locations, multi-user synchronisation across networks remains technically demanding, particularly for latency-sensitive tactical coordination exercises.
A further challenge concerns tangibility – the integration of physical props into virtual worlds enables authentic motor-skills training, yet current tracker requirements constrain the range and granularity of props that can be reliably supported. Communication with Non-Player Characters (NPCs) driven by Large Language Models raises complementary concerns: although these agents are becoming increasingly capable of processing voice commands, detecting emotional cues, and adapting behaviour to trainee actions,[57] achieving consistent, domain-appropriate responses in high-stress training scenarios remains an active research challenge.[58]
Finally, achieving the high contextual realism demanded by ChE training requires digitising actual operational environments. The manual effort to model real-world locations remains substantial. Emerging techniques such as Gaussian Splatting, a 3D scene reconstruction method that represents environments as collections of volumetric primitives to enable rapid capture and real-time rendering from sparse image data, offer promising approaches to rapid environment capture, but these technologies are still in their infancy and not yet mature enough to reliably produce training-ready assets at scale.[59]
The Future of XR for Challenging Environments
AI/ML Shaping the Future of XR Training
Artificial Intelligence (AI) and Machine Learning (ML) integration has the potential to revolutionise training personalisation.[60] Organisations harness AI and ML to create training programs catering to each trainee’s unique needs, offering personalised experiences that lead to improved performance.[61]
ML algorithms, especially those rooted in pattern recognition and predictive modelling, show promise in decoding intricate links between physiological signals and stress levels.[62] In AVEs, ML operates through two phases: learning stress patterns via supervised learning with labelled physiological data, then classifying real-time stress levels to dynamically adjust training intensity—maintaining optimal challenge-engagement balance while generating performance insights for after-action review.
ML algorithms, especially those rooted in pattern recognition and predictive modelling, show promise in decoding intricate links between physiological signals and stress levels.
Optimising Task Decomposition and Progressive Skill Integration
An emerging research direction focuses on optimising task decomposition strategies for XR training. While the benefits of the elemental training approach are well-established, determining the appropriate granularity of skill decomposition across domains remains an open question. Future research should investigate domain-specific principles for balancing fine-grained element practice with whole-task training, particularly in military and first-responder contexts, where operational complexities vary significantly.
AI-driven systems may enable automated optimisation of learning sequences, identifying prerequisite skill elements and determining optimal timing for integration into complete scenarios based on individual performance patterns.[63] The military domain, with its emphasis on distributed team training and multi-domain operations, presents particular opportunities for investigating how elemental competencies can be trained individually before coordination in joint exercises.
Limitations
Several limitations constrain the present analysis and warrant acknowledgement. First, this position paper presents a conceptual framework rather than empirical validation; the proposed mapping between Cognitive Load Theory constructs and XR dimensions remains theoretically motivated but requires systematic experimental verification across different operational domains and trainee populations. Second, the four-quadrant topology necessarily simplifies a complex multidimensional design space—real training systems may occupy intermediate positions or require dynamic traversal across quadrants within single sessions. Third, our treatment of “Elemental Training” presupposes that complex operational competencies can be meaningfully decomposed into trainable subcomponents, an assumption that may not hold equally across all skill types, particularly for highly integrated perceptual-motor tasks or tacit-knowledge domains.
Additionally, generalisation across professional domains assumes sufficient commonality in cognitive demands; domain-specific adaptations will likely be necessary. Finally, the rapid evolution of XR hardware and AI capabilities may render specific technological assumptions obsolete; the framework’s value lies in its pedagogical principles rather than platform-specific implementations.
The rapid evolution of XR hardware and AI capabilities may render specific technological assumptions obsolete; the framework’s value lies in its pedagogical principles rather than platform-specific implementations.
Conclusion
Extended Reality’s (XR) transformative potential in training for challenging occupations and environments is undeniable. This paper’s definition of ChE is multifaceted, encompassing scenarios demanding specialised training and preparation. Mapping training goals in these environments to the XR space provides a comprehensive framework for understanding that XR is not merely a simulation tool but a strategic asset for operational readiness.
The integration of Artificial Intelligence and Machine Learning into XR training platforms promises deeply personalised training experiences. Systematic task decomposition and progressive skill integration, supported by adaptive AI, enable organisations to build procedural automaticity efficiently, which is critical for military training, where complex competencies must be developed under time constraints. XR offers a safe proving ground for high-stakes decision-making; by simulating operational complexity, future leaders build resilience and cognitive adaptability before deployment. XR should be considered not an auxiliary tool but rather a fundamental pillar for achieving mission readiness.
[1] Alexandra D. Kaplan et al., “The Effects of Virtual Reality, Augmented Reality, and Mixed Reality as Training Enhancement Methods: A Meta-Analysis,” Human Factors 63, no. 4 (June 2021): 706–26, https://doi.org/10.1177/0018720820904229.
[2] Kristin L. Fraser, Paul Ayres, and John Sweller, “Cognitive Load Theory for the Design of Medical Simulations,” Simulation in Healthcare 10, no. 5 (October 2015): 295–307, https://doi.org/10.1097/SIH.0000000000000097.
[3] Louise Giaume et al., “Psychological, Cognitive, and Physiological Impact of Hazards Casualties’ Trainings on First Responders,” Frontiers in Psychology 15 (2024): 1336701, https://doi.org/10.3389/fpsyg.2024.1336701.
[4] Richard Skarbez, Missie Smith, and Mary C. Whitton, “Revisiting Milgram and Kishino’s Reality-Virtuality Continuum,” Frontiers in Virtual Reality 2 (2021): 647997, https://doi.org/10.3389/frvir.2021.647997.
[5] Jim Blascovich and Joe Tomaka, “The Biopsychosocial Model of Arousal Regulation,” in Advances in Experimental Social Psychology, vol. 28 (New York: Elsevier, 1996), 1–51, https://doi.org/10.1016/S0065-2601(08)60235-X.
[6] Blascovich and Tomaka, “Biopsychosocial Model of Arousal Regulation.”
[7] Jim Blascovich and Wendy B. Mendes, “Challenge and Threat Appraisals: The Role of Affective Cues,” in Feeling and Thinking: The Role of Affect in Social Cognition, ed. Joseph P. Forgas (New York: Cambridge University Press, 2000), 59–82.
[8] Blascovich and Mendes, “Challenge and Threat Appraisals.”
[9] Jim Blascovich et al., “Social ‘Facilitation’ as Challenge and Threat,” Journal of Personality and Social Psychology 77, no. 1 (1999): 68–77, https://doi.org/10.1037/0022-3514.77.1.68.
[10] John Sweller, Paul Ayres, and Slava Kalyuga, Cognitive Load Theory (New York: Springer, 2011), https://doi.org/10.1007/978-1-4419-8126-4.
[11] Richard M. Shiffrin and Walter Schneider, “Controlled and Automatic Human Information Processing II,” Psychological Review 84, no. 2 (1977): 127–90, https://doi.org/10.1037/0033-295X.84.2.127.
[12] Anke S. Baetzner et al., “Preparing Medical First Responders for Crises,” Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine 30, no. 1 (2022): 41, https://doi.org/10.1186/s13049-022-01056-8.
[13] Jennifer Wild et al., “Pre-Incident Training to Build Resilience in First Responders,” Psychiatry 83, no. 2 (2020): 128–42, https://doi.org/10.1080/00332747.2020.1750215.
[14] J. Kevin Ford and Aaron M. Schmidt, “Emergency Response Training,” Journal of Hazardous Materials 75, nos. 2–3 (2000): 195–215, https://doi.org/10.1016/S0304-3894(00)00180-1.
[15] Patrick Sweeney, Michael D. Matthews, and Paul D. Lester, Leadership in Dangerous Situations, 2nd ed. (Annapolis, MD: Naval Institute Press, 2022).
[16] Balakrishan S. Manoj and Alexandra Hubenko Baker, “Communication Challenges in Emergency Response,” Communications of the ACM 50, no. 3 (2007): 51–53, https://doi.org/10.1145/1226736.1226765.
[17] Skarbez, Smith, and Whitton, “Revisiting Milgram and Kishino’s Reality-Virtuality Continuum.”
[18] Christopher D. Wickens et al., “Effectiveness of Part-Task Training,” Human Factors 55, no. 2 (2013): 461–70, https://doi.org/10.1177/0018720812451994.
[19] Jeroen J. G. Van Merriënboer and John Sweller, “Cognitive Load Theory and Complex Learning,” Educational Psychology Review 17, no. 2 (2005): 147–77, https://doi.org/10.1007/s10648-005-3951-0.
[20] Van Merriënboer and Sweller, “Cognitive Load Theory and Complex Learning.”
[21] Van Merriënboer and Sweller.
[22] Jakob C. Uhl et al., “XR for First Responders,” in Virtual Reality and Mixed Reality, vol. 14410, ed. Gabriel Zachmann et al. (Cham: Springer, 2023), 192–200, https://doi.org/10.1007/978-3-031-48495-7_13.
[23] Skarbez, Smith, and Whitton.
[24] Richard E. Mayer, ed., The Cambridge Handbook of Multimedia Learning, 2nd ed. (Cambridge: Cambridge University Press, 2014), https://doi.org/10.1017/CBO9781139547369; Richard E. Mayer and Celeste Pilegard, “Principles for Managing Essential Processing,” in The Cambridge Handbook of Multimedia Learning, 316–44.
[25] Mel Slater, “Place Illusion and Plausibility,” Philosophical Transactions of the Royal Society B 364, no. 1535 (2009): 3549–57, https://doi.org/10.1098/rstb.2009.0138.
[26] Van Merriënboer and Sweller, “Cognitive Load Theory and Complex Learning.”
[27] Fraser, Ayres, and Sweller, “Cognitive Load Theory for the Design of Medical Simulations.”
[28] Van Merriënboer and Sweller, “Cognitive Load Theory and Complex Learning.”
[29] Kristina Höök and Jonas Löwgren, “Strong Concepts,” ACM Transactions on Computer-Human Interaction 19, no. 3 (2012): 1–18, https://doi.org/10.1145/2362364.2362371.
[30] Sweller, Ayres, and Kalyuga, Cognitive Load Theory.
[31] Van Merriënboer and Sweller, “Cognitive Load Theory and Complex Learning.”
[32] Van Merriënboer and Sweller.
[33] Shiffrin and Schneider, “Controlled and Automatic Human Information Processing II”; Mark D. Seery, “Challenge or Threat?” Neuroscience & Biobehavioral Reviews 35, no. 7 (2011): 1603–10, https://doi.org/10.1016/j.neubiorev.2011.03.003.
[34] Uhl et al., “XR for First Responders.”
[35] Van Merriënboer and Sweller, “Cognitive Load Theory and Complex Learning.”
[36] Skarbez, Smith, and Whitton.
[37] John D. Lee, “Fifty Years of Driving Safety Research,” Human Factors 50, no. 3 (2008): 521–28, https://doi.org/10.1518/001872008X288376.
[38] Seery, “Challenge or Threat?”; Blascovich and Mendes, “Challenge and Threat Appraisals.”
[39] Shiffrin and Schneider.
[40] Wickens et al., “Effectiveness of Part-Task Training.”
[41] Blascovich and Tomaka, “Biopsychosocial Model of Arousal Regulation”; Blascovich and Mendes, “Challenge and Threat Appraisals.”
[42] Eduardo Salas et al., “Using Simulation-Based Training,” Joint Commission Journal on Quality and Patient Safety 31, no. 7 (2005): 363–71, https://doi.org/10.1016/S1553-7250(05)31049-X.
[43] Daniel Fleming, Firearms Training Technologies for Accuracy and Qualification, HDIAC Report AD1210099 (Belcamp, MD: HDIAC, 2023), https://hdiac.dtic.mil/wp-content/uploads/2023/06/TI-Response-Report_HDIAC_Firearms-Training-Technologies-for-Accuracy-and-Qualification_AD1210099.pdf.
[44] Kurt Andersen, Simon Jose Gaab, and Frederick C. Harris, “METS VR,” in Proceedings of the 17th International Conference on Information Technology—New Generations (2020): 325–32, https://doi.org/10.1007/978-3-030-43020-7_43.
[45] Rafael Sacks, Amotz Perlman, and Ronen Barak, “Construction Safety Training Using Immersive Virtual Reality,” Construction Management and Economics 31, no. 9 (2013): 1005–17, https://doi.org/10.1080/01446193.2013.828844.
[46] Annette Mossel et al., “Immersive Virtual Reality Training System for CBRN Preparedness,” in International Multidisciplinary Modeling & Simulation Multiconference (2015).
[47] Andrzej Grabowski, “Practical Skills Training in Enclosure Fires,” Fire Safety Journal 125 (2021): 103440, https://doi.org/10.1016/j.firesaf.2021.103440.
[48] Fabrizia Mantovani et al., “Virtual Reality Training for Health-Care Professionals,” CyberPsychology & Behavior 6, no. 4 (2003): 389–95, https://doi.org/10.1089/109493103322278772.
[49] Markus Murtinger et al., “CBRNe Training in Virtual Environments,” International Journal of Safety and Security Engineering 11, no. 4 (2021): 295–303, https://doi.org/10.18280/ijsse.110402.
[50] Shiffrin and Schneider.
[51] Paulo Blikstein and Marcelo Worsley, “Multimodal Learning Analytics,” Journal of Learning Analytics 3, no. 2 (2016): 220–38, https://doi.org/10.18608/jla.2016.32.11.
[52] Charles R. Kelley, “What Is Adaptive Training?” Human Factors 11, no. 6 (1969): 547–56, https://doi.org/10.1177/001872086901100602.
[53] Maryam Zahabi and Ashiq Mohammed Abdul Razak, “Adaptive Virtual Reality-Based Training,” Virtual Reality 24 (2020): 725–52, https://doi.org/10.1007/s10055-020-00434-w.
[54] Jeremiah Rozman, The Synthetic Training Environment, AUSA Spotlight 20–6 (2020).
[55] Sam Van Damme et al., “Impact of Latency on QoE,” Applied Sciences 14, no. 6 (2024): 2290, https://doi.org/10.3390/app14062290.
[56] Jakob C. Uhl et al., “Tangible Immersive Trauma Simulation,” CHI Conference Proceedings (2023), https://doi.org/10.1145/3544548.3581292; Davide Calandra et al., “Immersive Virtual Reality and Passive Haptics,” Virtual Reality 26 (2022): 985–1012, https://doi.org/10.1007/s10055-021-00620-8.
[57] Joon Sung Park et al., “Generative Agents,” in UIST ’23 Proceedings (2023), https://doi.org/10.1145/3586183.3606763.
[58] Süeda Özkaya, Santiago Berrezueta-Guzman, and Stefan Wagner, “How LLMs Are Shaping VR,” IEEE Access 13 (2025): 193335–55, https://doi.org/10.1109/ACCESS.2025.3631594.
[59] Guikun Chen and Wenguan Wang, “A Survey on 3D Gaussian Splatting,” arXiv (2025), https://doi.org/10.48550/arXiv.2401.03890.
[60] Souvik Maity, “Artificial Intelligence in Training and Development,” Journal of Management Development 38, no. 8 (2019): 651–63, https://doi.org/10.1108/JMD-03-2019-0069.
[61] Russell Li and Zhandong Liu, “Stress Detection Using Deep Neural Networks,” BMC Medical Informatics and Decision Making 20 (2020): 285, https://doi.org/10.1186/s12911-020-01299-4.
[62] Aditi Bhutoria, “Personalized Education and AI,” Computers and Education: Artificial Intelligence 3 (2022): 100068, https://doi.org/10.1016/j.caeai.2022.100068.
[63] Joakim Laine et al., “Intelligent Tutoring Systems in VR,” International Journal of Technology in Education and Science 6, no. 2 (2022): 178–203, https://doi.org/10.46328/ijtes.334.
