Designing Extended Intelligence: A Taxonomy of Psychobiological Effects of XR–AI Systems for Human Capability Augmentation

Abstract

Extended Reality (XR) and Artificial Intelligence (AI) are increasingly converging within cyber–physical infrastructures, including digital twins, the Spatial Web, and smart-city systems. These environments require new frameworks for understanding how human performance emerges through sustained interaction with immersive interfaces and adaptive computational agents. This paper introduces the TAXI–XI-CAP framework, a two-layer model that links psychobiological mechanisms of XR–AI interaction to higher-level, experimentally testable capability constructs. The TAXI layer defines 42 mechanisms spanning perception, cognition, physiology, sensorimotor control, and social coordination, while XI-CAP organizes these into capability patterns such as remote dexterity, distributed cognition, and adaptive workload regulation. Derived through a theory-guided synthesis across XR, neuroscience, and human–automation interaction, the framework models performance as emerging from interacting mechanisms under real-world constraints. A validation-oriented research agenda is proposed, emphasizing mechanism-level measurement, capability-level evaluation, and longitudinal testing. The TAXI–XI-CAP framework provides a structured basis for hypothesis generation, comparative analysis, and empirical validation of XR–AI systems, supporting the development of reliable, scalable, and human-centered Extended Intelligence infrastructures.

1. Introduction

The idea that digital networks can function as environments for human augmentation has evolved across paradigms from hypertext and cyberspace to the metaverse and Spatial Web [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17]. More recently, this trajectory has converged toward persistent spatial computing infrastructures in which eXtended Reality (XR), artificial intelligence (AI), digital twins, and real-time systems operate as a continuous cyber–physical layer rather than as isolated applications [18,19,20,21,22,23,24]. These developments suggest that digital environments are becoming operational platforms for industrial control, remote collaboration, immersive training, and distributed decision-making, making their human factors and psychobiological effects a critical area of study.

Early work on hypertext and networked computing framed digital systems as tools for extending human cognition and collaboration [1,2,3,4,5]. Subsequent concepts such as cyberspace introduced inhabitable digital environments in which identity, presence, and agency could be reconfigured [10,11,12,13], while parallel policy and infrastructure perspectives emphasized global connectivity and economic transformation, often without systematically addressing long-term human consequences [11,12,15,25].

These developments have led to persistent digital environments in which virtual and physical systems are tightly integrated. Concepts such as mirror worlds, metaverse architectures, and immersive web infrastructures anticipated large-scale spatial computing platforms supporting real-time simulation, collaboration, and remote operation [12,15,16,17,20,22]. More recent advances in XR, digital twins, real-time rendering, and low-latency cloud–edge infrastructures further support this transition toward continuous cyber–physical systems [18,19,22,23,26,27,28]. At the same time, AI increasingly functions as an adaptive agent within these environments, capable of planning, filtering information, and acting on behalf of users [29,30,31,32,33,34].

Within this trajectory, Extended Intelligence (XI) conceptualizes intelligence as a distributed property of coordinated human–AI–tool systems rather than an autonomous machine capability, building on earlier conceptualizations of intelligence as a system-level property emerging from human–technology integration [35,36]. Research on human–AI synergy further supports this perspective by emphasizing how coordinated interaction between humans and intelligent systems can enhance decision-making and performance in complex environments [37]. This perspective is further reinforced by work on ethical frameworks for AI, as well as studies on immersive and augmented environments that shape cognition, learning, and collective intelligence in distributed contexts [38,39,40,41,42,43]. XR interfaces are increasingly proposed as the primary interaction medium through which such distributed intelligence becomes operational, enabling embodied interaction with simulations, data, and remote systems [16,44,45,46,47].

Despite the rapid convergence of XR, AI, and cyber–physical infrastructures, current research lacks a unified account of how these systems operate in real-time interaction. Existing approaches often treat XR as an interface layer and AI as an autonomous decision-making entity, without addressing their dynamic coupling with human perceptual, cognitive, and affective processes. As a result, interaction remains limited in its capacity for adaptive coordination, shared situational awareness, and context-aware autonomy, constraining performance in complex scenarios such as human–robot collaboration and remote operation. Addressing this gap requires models that treat perception, cognition, and action as continuously co-adaptive processes within integrated human–AI–environment systems.

To clarify this conceptual gap and situate the present contribution,

Table 1. Conceptual positioning of Extended Intelligence (XI) relative to Extended Reality (XR), Artificial Intelligence (AI), augmented cognition, and cyber-physical systems.

Concept	Definition	Primary System Focus	Role of XR	Role of AI	Implications for Human Capability	Typical Application Contexts
Extended Reality (XR)	Umbrella term covering virtual, augmented, and mixed reality technologies that modify or extend perception through immersive interfaces.	Interface technology	Primary component	Optional	Enhances perception and interaction	Simulation, visualization, training, immersive environments
Artificial Intelligence (AI)	Computational systems capable of performing tasks that normally require human intelligence, including perception, prediction, and decision support.	Algorithmic intelligence	Optional	Primary component	Supports cognition and decision making	Automation, decision support, adaptive systems, data analysis
Augmented Cognition	Human–technology systems designed to enhance cognitive performance through adaptive interfaces, physiological sensing, and feedback.	Human performance optimization	Optional	Optional	Improves attention, workload regulation, and learning	Neuroergonomics, adaptive interfaces, training systems
Cyber-Physical Systems (CPS)	Integrated systems combining computation, sensing, and physical processes operating in real time and distributed environments.	System-level integration	Optional	Optional	Coordinates human, machine, and environment	Smart cities, robotics, industrial control, infrastructure monitoring
Digital Twins	Virtual representations of physical systems used for monitoring, simulation, and decision support in real time.	Operational modeling	Often used	Often used	Supports understanding and control of complex systems	Industry, infrastructure, healthcare, smart cities
Extended Intelligence (XI)	Closed-loop integration of immersive XR interfaces and adaptive AI within cyber-physical environments, enabling psychobiological mechanisms (TAXI) that converge into capability-level outcomes (XI-CAP).	Human–XR–AI capability systems	Core interface layer	Core adaptive layer	May extend or constrain human capability depending on design	Teleoperation, digital twins, immersive collaboration, training, smart-city systems

positions Extended Intelligence (XI) in relation to adjacent paradigms, including XR, AI, augmented cognition, cyber–physical systems, and digital twin infrastructures. Such differentiation is necessary because these domains are frequently conflated, leading to conceptual ambiguity regarding their respective roles, levels of integration, and operational scope.

Experience with earlier digital technologies shows that large-scale deployment often precedes systematic study of long-term human effects. Platforms are frequently introduced with narratives of empowerment and efficiency, while their behavioral and cognitive consequences are examined only after widespread adoption. Social media illustrates this pattern, where persuasive interface design, algorithmic engagement optimization, and sustained attentional capture were not fully anticipated, and policy responses have largely focused on mitigating harms after they emerge, particularly among younger populations [48,49,50,51,52,53,54,55,56,57,58].

This pattern suggests that immersive and AI-supported infrastructures should be evaluated proactively, with attention to the psychobiological pathways through which interaction influences attention, motivation, decision-making, and social behavior.

Human factors and ergonomics research shows that even mature computing paradigms produce persistent strain effects. Long-term use of conventional desktop interfaces is associated with musculoskeletal load, repetitive strain injuries, and visual fatigue, despite extensive optimization [59,60]. XR systems introduce additional constraints, including cybersickness, vestibular conflict, visual discomfort, and increased cognitive workload during spatial interaction [61,62,63,64,65,66]. As immersive environments become part of continuous professional workflows, these effects must be understood at the level of underlying psychobiological mechanisms rather than through usability evaluation alone.

At the same time, immersive XR–AI environments may enable forms of capability augmentation not achievable with conventional interfaces. Embodied interaction, multisensory integration, adaptive assistance, and shared spatial environments can support improved learning, remote manipulation, distributed cognition, and sustained performance under demanding conditions [45,67,68,69,70]. However, the mechanisms through which these systems influence perception, cognition, physiology, affect regulation, motor performance, and social coordination remain so far fragmented across research domains, including neuroscience, human–computer interaction (HCI), medical XR (MXR), neuroergonomics, human–robot interaction (HRI), and human–AI interaction.

Understanding how these effects should be studied and translated into system design remains a methodological challenge. In XR research, constructs such as presence, immersion, and cybersickness became practically useful only after being operationalized through standardized measures, experimental paradigms, and empirically derived design guidelines [66,71,72,73,74]. This progression enabled immersive systems to be evaluated systematically rather than intuitively.

A comparable progression has not yet been established for XR–AI environments, where immersive interfaces, adaptive AI, and cyber–physical systems form continuous interaction loops. As a result, capability-level claims—such as improved learning, collaboration, or decision-making—are often made without a clear mapping between underlying psychobiological mechanisms and observable performance outcomes.

This work addresses this gap by introducing a mechanism-based framework for XR–AI systems. First, a taxonomy of Extended Intelligence (TAXI) is proposed to organize the psychobiological mechanisms activated during immersive interaction. Building on this, XI Capability Augmentation Patterns (XI-CAP) are defined as higher-level constructs that group these mechanisms into functionally relevant human capability enhancements for real-world deployment. Together, these layers provide a structured basis for translating mechanism-level evidence into experimentally testable capability outcomes. An example of operationalization is presented, followed by a research agenda outlining key directions for future work.

Figure 1 complements Table 1 by visually situating XI relative to the overarching spatial web (represented by the digital twins/smart cities with XR-AI Infrastructures), with XR, AI, augmented cognition, and cyber–physical systems, while also illustrating how these domains connect to the TAXI and XI-CAP layers within the proposed framework.

In the clinical domain, MXR has begun to identify neuropsychological and biological mechanisms through which XR interventions influence outcomes [47]. This work extends the MXR mechanisms into XR–AI contexts by incorporating additional mechanisms relevant to perception, cognition, physiology, affect regulation, motor performance, and social coordination in phygital environments, from adjacent domains and disciplines.

TAXI and XI-CAP provide a structured basis for formulating and testing hypotheses about how XR–AI interaction influences human performance, well-being, and decision-making. By organizing psychobiological mechanisms into human capability domains, the framework enables systematic empirical evaluation using controlled experimental methods.

XI Capability Augmentation Patterns (XI-CAP) group combinations of TAXI mechanisms into higher-level capability outcomes relevant to real-world XR–AI applications, such as digital twins, immersive collaboration, and AI-assisted decision systems. For example, mechanisms related to attentional focus, cognitive load regulation, and multisensory integration can be combined into a capability pattern that supports enhanced situational awareness in remote operations.

As domain-dependent structuring tools, taxonomies are often used to provide an overview of emerging fields. While existing XR taxonomies focus on interface or application domains, the present framework organizes psychobiological mechanisms underlying capability-level constructs. This enables the systematic assessment of how XR–AI systems influence human performance and supports the design of systems that maximize beneficial effects while mitigating potential risks.

Table 2. Structure of the TAXI–XI-CAP framework linking psychobiological mechanisms to capability patterns, empirical validation, and deployment considerations.

Framework Level	Description	Role in the XI Framework	Typical Methods	Example Constructs/Outputs
Conceptual scope of Extended Intelligence (XI)	Definition of XR–AI systems as closed-loop human–technology environments operating in phygital and cyber-physical contexts.	Establishes theoretical scope and distinguishes XI from XR, AI, and augmented cognition.	Conceptual analysis, literature synthesis, framework definition	XR–AI environments, digital twins, immersive collaboration, cyber-physical systems
TAXI mechanisms (mechanism level)	Taxonomy of psychobiological efficacy mechanisms through which XR–AI interaction may influence perception, cognition, physiology, affect, and action.	Provides empirically grounded building blocks for capability analysis.	Experimental XR studies, neuroscience, human factors, neuroergonomics	Presence, embodiment, multisensory integration, workload regulation, trust calibration, fatigue dynamics
Mechanism grouping	Organization of mechanisms into functional clusters reflecting perceptual, physiological, sensorimotor, social, and cognitive processes.	Improves interpretability and supports compositional analysis of mechanisms.	Taxonomy synthesis, cross-domain comparison	Experiential, regulatory, sensorimotor, affective, cognitive, and distributed-interaction mechanisms
XI-CAP capability patterns	Higher-level performance outcomes emerging from combinations of TAXI mechanisms under realistic interaction conditions.	Translates mechanism-level evidence into capability-level hypotheses.	Conceptual synthesis, neuroergonomics, motor learning, human–AI interaction	Remote super-dexterity, distributed cognition, adaptive workload regulation, collective coordination
Empirical validation layer	Experimental testing of whether proposed capability patterns can be produced, stabilized, and reproduced under realistic conditions.	Defines research agenda and validation requirements.	Controlled experiments, longitudinal studies, multi-user studies, physiological measurement	Mechanism stability, fatigue effects, trust calibration, long-duration usability
Design and deployment guidelines	Translation of validated capability patterns into system design principles and operational constraints.	Supports safe and reliable XI system development.	Human factors engineering, usability testing, field trials	Interface design, AI adaptation rules, workload limits, safety constraints
Operational XI systems	Real-world XR–AI environments in which validated mechanisms and capabilities are used under realistic conditions.	Target domain of the framework.	Deployment studies, cyber-physical evaluation, large-scale testing	Digital twins, teleoperation, smart cities, medical collaboration, immersive training

outlines the progression from mechanisms to capability patterns, empirical validation, and deployment considerations for XI systems.

The remainder of this article is structured as follows. Section 2 describes the methodological approach used to identify and synthesize a taxonomy of psychobiological mechanisms underlying human capability augmentation in XR–AI interaction. Section 3 presents the resulting taxonomy of these mechanisms (TAXI) and outlines their relationship to XI Capability Augmentation Patterns (XI-CAP). Section 4 demonstrates how TAXI and XI-CAP can be applied for empirical validation. Section 5 outlines a research agenda for advancing empirical validation toward real-world deployment of human capability augmentation in XR–AI systems. Section 6 concludes with implications for XR–AI system design and directions for future research, including the extension of TAXI and XI-CAP.

2. Method and Materials

2.1. Study Design

This study adopts a theory-guided synthesis approach to develop a psychobiological mechanism-based framework for human capability enhancement through interaction with cyber-physical environments via XR–AI systems leading to XI. The objective is to integrate empirically supported psychobiological mechanisms into a structured taxonomy and show how they enable capability-oriented constructs relevant to XI.

The primary aim is to establish a comprehensive yet conservative taxonomy of psychobiological mechanisms underlying XR–AI interaction. The identification of mechanisms was conducted as exhaustively as possible based on available evidence; however, the citation strategy is selective rather than exhaustive. References were chosen to include seminal and recent empirical studies that substantiate each mechanism and provide a representative entry point into the literature.

Statistical aggregation is not the focus at this stage, as the objective is to organize and define the underlying mechanisms that structure the domain. Such taxonomy-building provides a necessary foundation for consistent definitions, comparability across studies, and the formulation of testable hypotheses. This approach aligns with framework-building reviews in XR, neuroergonomics, HCI, and human–AI systems, where the goal is to establish conceptual structure and translational relevance prior to quantitative synthesis [37,47,69,70,75].

The methodological workflow comprised three distinct activities:

• construction of a psychobiological mechanism taxonomy (TAXI) layer;
• derivation of the connection to an augmented capability (XI-CAP) layer;
• formulation of a research agenda to validate the XI-CAP constructs.

This structure enables analysis of XR–AI effects at the level of underlying processes and emergent capabilities, and formulates the required empirical validation.

The synthesis process followed a structured pipeline consisting of (i) identification of core mechanisms from the MXR domain, (ii) targeted expansion through adjacent domains, (iii) screening using inclusion and distinctness criteria, and (iv) consolidation into a unified taxonomy with consistent definitions and operational scope.

2.2. Conceptual Scope

The review focuses on psychobiological mechanisms through which XR–AI interaction influences human performance and behavior in sustained operational environments.

A mechanism was included if it satisfied three conditions:

• has empirical support in peer-reviewed literature;
• represents a psychobiological or cognitive-behavioral process rather than an application domain;
• is relevant to capability augmentation in immersive or phygital XR–AI environments.

Because XR–AI systems combine immersive interfaces, adaptive automation, and distributed collaboration, relevant mechanisms draw from historically separate research domains, including neuroscience, ergonomics, motor learning, distributed cognition, and human–machine interaction [46,66,69,70,75].

To avoid conceptual conflation, the taxonomy distinguishes psychobiological mechanisms from system-level properties and design features. Mechanisms are defined as processes operating within or through human perceptual, cognitive, physiological, or behavioral systems, even when shaped by interaction with AI or external representations. System architectures, interface features, and application contexts are treated as conditions under which mechanisms are activated rather than as mechanisms themselves.

2.3. Construction of the TAXI Mechanism Layer

The initial mechanism inventory was grounded in the MXR taxonomy developed by Spiegel et al. [47], which identifies psychobiological mechanisms underlying XR effects, including presence, embodiment, multisensory integration, attentional modulation, and physiological regulation.

These mechanisms were adopted as a validated foundation for TAXI, as they represent empirically supported psychobiological effects of interacting with XR, at the level of underlying human processes rather than application domains.

Accordingly, the MXR-derived mechanisms were extended to include processes relevant to human capability-enhancing effects of interacting with XR–AI operational environments.

2.4. Extension Beyond XR to XR-AI

Additional human capability-enhancing XR-AI mechanisms were identified through targeted synthesis across adjacent domains, including neuroergonomics, distributed cognition, human–automation interaction, motor learning, and social coordination research [66,69,70,75,76,77,78].

These additions incorporate established psychobiological processes that have not yet been systematically linked to XR–AI interaction. This extension is necessary when XR is treated as a persistent interface for teleoperation, collaborative environments, digital twins, when adding AI-mediated functionality to XR systems. The extended mechanism set reflects deployment conditions characterized by XR-AI-mediated:

• sustained or repeated exposure;
• high-precision cognitive and motor demands;
• multi-user coordination;
• perception and decision support;
• supervision of remote or simulated systems;
• long-duration workload and fatigue effects.

These conditions define the operational context of XI in XR–AI systems and justify extending the MXR mechanism inventory.

2.5. Mechanism Inclusion and Distinctness Criteria

Because related phenomena are described differently across research traditions, explicit criteria were applied to determine whether a process should be treated as a distinct mechanism.

A mechanism was retained if it met at least one of the following criteria:

• operates at a distinct psychobiological level;
• involves a different causal pathway;
• produces different measurable outcomes;
• cannot be reduced to another mechanism without loss of explanatory clarity.

These criteria limit conceptual inflation while preserving processes that are functionally important in XR–AI environments.

The taxonomy does not claim the discovery of new biological effects, but reorganizes mechanisms described across different domains (XR, neuroscience, psychology, ergonomics, and human–machine interaction, etc.), into a unified framework for the empirical assessment of XR–AI interaction effects.

2.6. Derivation of XI-CAP

Following the construction of the TAXI mechanism layer, a second layer was developed to group mechanisms into higher-level functional outcomes: Extended Intelligence Capability Augmentation Patterns (XI-CAP), representing augmented human capabilities enabled through XR–AI-mediated interaction with cyber–physical systems.

XI-CAP constructs were derived by identifying clusters of mechanisms that converge on shared performance outcomes, such as dexterity, distributed cognition, learning, and workload regulation. This approach is consistent with human factors and distributed cognition frameworks, in which performance emerges from coordinated human–technology interaction rather than isolated system components [37,69,70,79,80].

Mechanisms were grouped into capability patterns when they met one or more of the following criteria:

• convergence toward a shared functional outcome;
• relevance to high-precision or long-duration tasks;
• plausibility of modulation through immersive system design;
• relevance to real-world deployment constraints.

XI-CAP constructs are therefore defined as emergent capability patterns arising from the interaction of multiple psychobiological mechanisms within XR–AI environments. As such, XI-CAP serves as a translational layer linking mechanism-level evidence to system-level capability design and provides a structured basis for subsequent validation.

2.7. Literature Identification

Supporting literature for the TAXI mechanisms and XI-CAP constructs was identified through targeted searches in PubMed, Scopus, IEEE Xplore, and Google Scholar using terms related to XR, cognition, neuroergonomics, human–AI interaction, and performance. Publications were included if they provided empirical support for identified mechanisms, informed capability patterns, or addressed safety and deployment constraints. Selection was purposive rather than systematic, consistent with the framework-development focus of the study.

2.8. Evidence Provenance

The mechanisms included in the taxonomy originate from multiple research domains, including XR, neuroscience, psychology, ergonomics, and human–machine interaction. While not all mechanisms have been studied directly in XR contexts, their inclusion reflects processes relevant to XR–AI systems. The taxonomy is therefore presented as a synthesis of current knowledge rather than a definitive or exhaustive account.

2.9. Derivation of Research Agenda and Research Priorities

A research agenda with priorities was derived through expert synthesis by a multidisciplinary team spanning expertise in psychology, neuropsychology, human factors, ergonomics, XR, AI, and robotics.

Research priorities were selected when they addressed one or more of the following:

• mechanisms likely to become safety-critical under long-duration use;
• interactions between mechanisms that remain insufficiently validated;
• areas where XR evidence, adjacent disciplines, and XR–AI deployment conditions intersect but remain untested;
• barriers to real-world implementation, including workload stability, trust calibration, coordination, interpretability, and long-term adaptation.

These priorities define a validation research agenda for XI and XR–AI systems, identifying the mechanisms and interactions that require empirical testing before reliable deployment.

The research agenda is intentionally selective, focusing on factors most likely to determine whether XR–AI systems can transition from conceptual promise to reliable deployment.

The resulting framework should be understood as an XI–oriented taxonomy grounded in the medical field and extended through adjacent disciplines to support translational evaluation of interaction with XR–AI systems.

3. Results

The TAXI framework identified 42 psychobiological mechanisms, describing how XR–AI interaction influences human perception, cognition, physiology, and behavior. These mechanisms are grouped into seven functional domains:

A.

Experiential and Cognitive–Perceptual Mechanisms (M1–M6)

B.

Regulatory and Neurophysiological Mechanisms (M7–M13)

C.

Social and Bodily-State Mechanisms (M14–M18)

D.

Sensorimotor and Action Mechanisms (M19–M22)

E.

Affective and Identity Mechanisms (M23–26)

F.

Cognitive and Neuroergonomic Mechanisms (M27–M34)

G.

Distributed and Adaptive System Mechanisms (M35–M42)

The mechanisms (M1–M42) are presented in Section 3.2.

TAXI is not intended to claim the discovery of new biological effects. Instead, it organizes empirically supported mechanisms of action into a structure usable for (i) capability-augmentation-oriented analyses, (ii) XI usability and usefulness evaluations, and (iii) systematic research prioritization of XI-supporting implementations in large-scale Spatial Web and cyber–physical deployments, including for persistent digital twins and metaverse-scale environments [18,21,23,24,79].

Similar methodological progression has occurred in immersive systems research, where constructs such as presence, embodiment, and cybersickness became practically manageable or useful only after they were operationalized through standardized measures, experimental paradigms, and human factors design guidelines [66,71,73].

XR roadmapping and Delphi consensus studies identify immersive technologies as key enabling infrastructures for industrial, educational, and societal transformation, while emphasizing that human factors and psychophysiological constraints remain major barriers to adoption [80]. This perspective aligns with broader developments in smart city and digital twin research, where large-scale data-driven infrastructures are used to model and manage complex urban systems in real time [81]. Related work further highlights the importance of mechanism-level understanding in domains such as neurorehabilitation, adaptive training, and narrative-based perspective modulation, where behavioral and physiological effects depend strongly on interaction design [82]. Building on these developments, there is increasing recognition of the need for integrated system-level approaches that link interaction processes with computational and cyber–physical infrastructures in XR systems [83,84].

3.1. TAXI: Source Grounding and Expansion Beyond XR Mechanisms

The TAXI framework is grounded in the taxonomy of the breadth and depth of the Medical eXtended Reality (MXR) field developed by Spiegel et al. [47], which provides a clinically grounded classification of many aspects relevant to MXR, including neuropsychological and biological mechanisms of action observed when users interact with XR applications. They identified empirically supported pathways through which XR systems influence psychological, physiological, and neurobiological processes [47].

The extension of the mechanism set beyond the MXR mechanisms follows a consistent selection logic aligned with the object of classification in TAXI. Specifically, mechanisms were included when they are (i) supported in the literature as contributing to human perception, cognition, physiology, or behavior, and (ii) directly relevant to XR–AI interaction conditions, including persistent use, adaptive system coupling, and task-oriented deployment. This ensures that all included mechanisms are grounded in established research, while extending the taxonomy to cover processes that become salient only when XR is considered within broader extended intelligence systems.

Mechanisms M1–M13, M16, and M17 in the TAXI framework correspond directly to the neuropsychological and biological mechanisms of action in the MXR taxonomy and are the core psychobiological efficacy mechanisms underlying XR interaction.

Table 3. Conceptual mapping between XR mechanism categories reported by Spiegel et al. [47] and the TAXI mechanism set proposed in the present XI framework, representing an original synthesis and reinterpretation by the authors, in which reported medical constructs are reorganized, generalized, and extended, to show how they are included in the broader XR–AI operational contexts.

XR Mechanism (as reported by Spiegel et al. [47]	XR Domain Classification (based on Spiegel et al. [47]	TAXI Mechanism (ID and Label)	Mapping Interpretation within TAXI Framework
Presence/immersion	XR neuropsychological and biological mechanisms of action	M1 Presence and Immersion	Direct adoption
Virtual embodiment	XR neuropsychological and biological mechanisms of action	M2 Virtual Embodiment	Direct adoption
Proteus effect	XR neuropsychological and biological mechanisms of action	M3 Proteus Effect	Direct adoption
Multisensory input in XR	XR neuropsychological and biological mechanisms of action	M4 Multisensory Integration	Direct adoption (generalized label)
Distraction/spotlight of attention	XR neuropsychological and biological mechanisms of action	M5 Distraction and Attentional Modulation	Direct adoption (expanded label)
Alterations of time perception	XR neuropsychological and biological mechanisms of action	M6 Altered Time Perception	Direct adoption
Biofeedback in XR	XR neuropsychological and biological mechanisms of action	M7 Biofeedback/Neurofeedback Integration	Direct adoption (closed-loop framing)
Physiological effects of XR	XR neuropsychological and biological mechanisms of action	M8 Physiological Regulation	Direct adoption (refined scope)
Neurobiological effects of XR	XR neuropsychological and biological mechanisms of action	M9 Neurobiological Modulation	Direct adoption (refined scope)
Placebo/sham effects	XR neuropsychological and biological mechanisms of action	M10 Placebo and Expectancy Effects	Direct adoption
Biophilia/nature environments	XR neuropsychological and biological mechanisms of action	M11 Biophilia and Nature-Based Effects	Direct adoption
Meditation effects of XR	XR neuropsychological and biological mechanisms of action	M12 Meditation and Mindfulness Effects	Direct adoption
Immunological effects of XR	XR neuropsychological and biological mechanisms of action	M13 Immunological Effects	Direct adoption (evidence maturity noted)
Vestibular/motion cue effects	Ethics, safety, privacy, and adverse effects of MXR	M16 Vestibular Engagement and Motion Cue Integration	Cross-category adoption
Cybersickness/simulator sickness	Ethics, safety, privacy, and adverse effects of MXR	M17 Sensory Conflict and Cybersickness Dynamics	Cross-category adoption

provides a mapping between MXR mechanism items and the corresponding TAXI mechanisms, including minor relabeling for definitional clarity.

The table draws on mechanism categories reported by Spiegel et al. [47], but does not reproduce their original presentation. Instead, it provides an analytical mapping in which these constructs are reorganized, selectively relabeled, and extended to reflect XR–AI operational contexts beyond the original medical XR focus. The MXR taxonomy was developed to provide an overview of XR as an evolving field, rather than to define a mechanism-level framework for XR–AI or Extended Intelligence (XI) systems. The selected mechanisms from Spiegel et al. [47] therefore offer a validated foundation for some of the items included in the mechanism-level analysis presented in the TAXI framework, but do not fully account for XR–AI conditions involving persistent interaction, adaptive AI coupling, and operational deployment in cyber–physical environments.

To address this, the TAXI framework extends the MXR mechanism inventory through targeted synthesis across adjacent domains, including AI, neuroergonomics, distributed cognition, human–automation interaction, motor learning, social coordination, embodiment, and workload research [42,43,65,69,70,75,85,86]. This integration incorporates mechanisms relevant to XR–AI operational contexts, such as shared situation awareness, interpersonal synchrony, trust calibration, fatigue dynamics, tool-mediated action, predictive control, and AI-guided attentional regulation [46,73,79,80,87,88]. These mechanisms are not introduced as novel biological effects, but as relevant recontextualizations of established processes within XR–AI coupling scenarios.

In addition, processes previously treated as adverse effects in XR research are formalized as mechanisms when they represent measurable constraints on performance. For example, cybersickness reflects sensory-conflict physiology and constitutes a primary limiting factor for sustained XR–AI use, and is therefore treated as a distinct psychobiological constraint mechanism rather than solely as a safety outcome [61,64,65,66,89].

3.2. TAXI Mechanism Inventory and Presentation Format

The complete TAXI mechanism inventory is presented below (see subsections A–G), followed by the respective Tables which list the mechanisms, operational definitions, example XI use contexts, and supporting literature. The inventory and the tables follow the functional grouping introduced in Section 3.1 and serve as a reference layer for identifying which mechanisms are active in a given XR–AI scenario and how they may be measured or experimentally manipulated.

A.

Experiential and Cognitive–Perceptual Mechanisms: Perception and subjective experience in XR environments, includes M1 Presence and Immersion; M2 Virtual Embodiment; M3 Proteus Effect; M4 Multisensory Integration; M5 Attentional Modulation; M6 Altered Time Perception.

B.

Regulatory and Neurophysiological Mechanisms: Physiological and neural state regulation, includes M7 Biofeedback/Neurofeedback Integration; M8 Physiological Regulation; M9 Neurobiological Modulation; M10 Placebo and Expectancy Effects; M11 Biophilia and Nature-Based Effects; M12 Meditation and Mindfulness Effects; M13 Immunological Effects.

C.

Social and Bodily-State Mechanisms: Social coordination and bodily stability, includes M14 Action Observation and Imitation Learning; M15 Social Synchrony and Interpersonal Coordination; M16 Vestibular Engagement and Motion Cue Integration; M17 Sensory Conflict and Cybersickness Dynamics; M18 XR-Induced Analgesia and Pain Modulation).

D.

Sensorimotor and Action Mechanisms: Motor control and action capability extension, includes M19 Motor Skill Acquisition and Procedural Automation; M20 Tool Incorporation and Agency Extension; M21 Visuospatial Rescaling and Scale Transfer; M22 Predictive Perception and Anticipatory Control.

E.

Affective and Identity Mechanisms: Emotion, identity, and social cognition, includes M23 Affective Priming; M24 Self-Concept Plasticity; M25 Perspective-Taking and Moral–Cognitive Modulation; M26 Social Presence and Affective Resonance.

F.

Cognitive and Neuroergonomic Mechanisms: Higher-order cognition and performance regulation, includes M27 Cognitive Offloading and Externalized Cognition; M28 Shared Situation Awareness and Distributed Cognition; M29 Metacognitive Calibration and Confidence Regulation; M30 Trust Calibration and Automation Reliance; M31 Fatigue Dynamics and Vigilance Sustainment; M32 Cognitive Workload Regulation; M33 Adaptive Error Shaping and Learning Acceleration; M34 Flow State Induction and Sustained Engagement.

G.

Distributed and Adaptive System Mechanisms: Distributed cognition, agency, and long-term adaptation, includes M35 Mental Time Travel and Temporal Simulation; M36 High-Dimensional Abstraction and Spatial Reasoning Support; M37 Parallel Presence and Distributed Attentional Allocation; M38 Delegated Embodiment and Proxy Agency; M39 Attentional Orchestration by Autonomous Agents; M40 Long-Term Neuroplastic Adaptation; M41 Future-Self Embodiment and Motivational Commitment; M42 Replay-Based Self-Modeling and Reflective Learning.

To provide a structural overview of the TAXI mechanism inventory, Figure 2 presents the functional organization of the 42 psychobiological mechanisms grouped into the seven domains (A–G) described above. The figure serves as a visual reference for how mechanisms are distributed across perceptual, cognitive, physiological, sensorimotor, affective, and system-level interaction processes in XR–AI environments.

The grouping reflects recurring functional domains rather than mutually exclusive categories. Mechanisms are expected to co-occur and interact across domains, and their combined effects form the basis for the capability patterns introduced in Section 4. Figure 2 therefore complements the detailed tables by providing a high-level structural representation of the taxonomy.

Document S1 provides extended definitions of the TAXI mechanism groups and the 42 psychobiological mechanisms, complementing the concise descriptions presented above and in the tables below. It also outlines representative approaches for their operationalization in empirical and applied XR–AI contexts.

The detailed TAXI mechanism inventory is presented in the following Table 4, Table 5, Table 6, Table 7, Table 8, Table 9 and Table 10, organized according to the seven functional domains (A–G), defined above. Each table lists the mechanisms within a domain, along with their operational definitions, example XR–AI use contexts, and supporting literature, providing a structured reference layer for identifying, measuring, and experimentally validating mechanism-level processes in XR–AI systems.

• (A) Experiential and Cognitive–Perceptual Mechanisms

Table 4. Core XR experiential and cognitive–perceptual mechanisms in the TAXI framework, including presence, embodiment, multisensory integration, attentional modulation, and altered time perception during immersive interaction.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M1 Presence and Immersion	Subjective sense of “being there” in a virtual environment, eliciting realistic cognitive and physiological responses.	High-fidelity simulation training; digital twin walkthroughs; remote operations requiring perceptual realism.	[46,65,67,71,90,91,92,93,94,95,96]
M2 Virtual Embodiment	Sense of ownership and control over a virtual body or body part.	Teleoperation with embodied control; rehabilitation simulation; avatar-based collaboration.	[46,97,98,99,100,101,102]
M3 Proteus Effect	Behavioral and cognitive changes driven by avatar characteristics and identity cues.	Role rehearsal; leadership training; bias modulation; identity-based behavioral shaping.	[65,103,104,105]
M4 Multisensory Integration	Coherent perception arising from synchronized multimodal sensory input.	Surgical and technical simulation with haptics; multimodal alert systems; accessibility via redundant sensory cues.	[73,106,107,108,109,110,111,112]
M5 Attentional Modulation (Distraction and Focus Control)	Allocation and control of attention through salience, sensory load, and task-relevant cues.	Focus guidance in complex interfaces; attentional steering in control systems; distraction-based modulation (e.g., pain reduction).	[73,86,113,114,115,116,117,118]
M6 Altered Time Perception	Subjective compression or expansion of perceived time during immersive interaction.	Extended training sessions; sustained monitoring tasks; endurance support in repetitive workflows.	[65,119,120,121]

Table 4. Core XR experiential and cognitive–perceptual mechanisms in the TAXI framework, including presence, embodiment, multisensory integration, attentional modulation, and altered time perception during immersive interaction.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M1 Presence and Immersion	Subjective sense of “being there” in a virtual environment, eliciting realistic cognitive and physiological responses.	High-fidelity simulation training; digital twin walkthroughs; remote operations requiring perceptual realism.	[46,65,67,71,90,91,92,93,94,95,96]
M2 Virtual Embodiment	Sense of ownership and control over a virtual body or body part.	Teleoperation with embodied control; rehabilitation simulation; avatar-based collaboration.	[46,97,98,99,100,101,102]
M3 Proteus Effect	Behavioral and cognitive changes driven by avatar characteristics and identity cues.	Role rehearsal; leadership training; bias modulation; identity-based behavioral shaping.	[65,103,104,105]
M4 Multisensory Integration	Coherent perception arising from synchronized multimodal sensory input.	Surgical and technical simulation with haptics; multimodal alert systems; accessibility via redundant sensory cues.	[73,106,107,108,109,110,111,112]
M5 Attentional Modulation (Distraction and Focus Control)	Allocation and control of attention through salience, sensory load, and task-relevant cues.	Focus guidance in complex interfaces; attentional steering in control systems; distraction-based modulation (e.g., pain reduction).	[73,86,113,114,115,116,117,118]
M6 Altered Time Perception	Subjective compression or expansion of perceived time during immersive interaction.	Extended training sessions; sustained monitoring tasks; endurance support in repetitive workflows.	[65,119,120,121]

Table 4 presents core XR experiential and cognitive–perceptual mechanisms, including presence, embodiment, multisensory integration, attentional modulation, and altered time perception. These mechanisms describe how immersive systems influence perception and cognition.

• (B) Regulatory and Neurophysiological Mechanisms

Table 5. Regulation, neurophysiological, and expectancy-related mechanisms in the TAXI framework, including biofeedback, physiological regulation, neurobiological modulation, expectancy effects, nature-based restoration, mindfulness-related regulation, and immunological pathways.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M7 Biofeedback/Neurofeedback Integration	Closed-loop physiological sensing and feedback enabling self-regulation.	HRV/EEG-based stress training; adaptive XR environments responding to user state; performance regulation.	[122,123,124,125,126]
M8 Physiological Regulation	Modulation of autonomic and physiological state (e.g., arousal, stress responses) during XR interaction.	Recovery design for high-stress work; resilience training; sustained operational readiness.	[67,86,116,126]
M9 Neurobiological Modulation	Changes in neural activation and plasticity associated with learning and cognitive adaptation.	Neuroergonomic optimization; accelerated skill acquisition; cognitive performance tuning.	[67,127,128,129,130]
M10 Placebo and Expectancy Effects	Belief-driven modulation of perception, physiology, and performance.	Framing-driven motivation; expectancy-enhanced training; analgesia modulation.	[131,132,133,134]
M11 Biophilia and Nature-Based Effects	Stress reduction and cognitive restoration through exposure to natural environments.	Micro-break recovery environments; burnout reduction; wellbeing design in XR workspaces.	[135,136,137,138]
M12 Meditation and Mindfulness Effects	Attentional and emotional regulation supported by meditation-based XR interventions.	Readiness priming; stress resilience; cognitive reset before high-demand tasks.	[139,140,141]
M13 Immunological Effects	Indirect modulation of immune function through stress and autonomic regulation pathways.	Long-term wellbeing interventions; occupational stress–health mitigation.	[142,143]

Table 5. Regulation, neurophysiological, and expectancy-related mechanisms in the TAXI framework, including biofeedback, physiological regulation, neurobiological modulation, expectancy effects, nature-based restoration, mindfulness-related regulation, and immunological pathways.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M7 Biofeedback/Neurofeedback Integration	Closed-loop physiological sensing and feedback enabling self-regulation.	HRV/EEG-based stress training; adaptive XR environments responding to user state; performance regulation.	[122,123,124,125,126]
M8 Physiological Regulation	Modulation of autonomic and physiological state (e.g., arousal, stress responses) during XR interaction.	Recovery design for high-stress work; resilience training; sustained operational readiness.	[67,86,116,126]
M9 Neurobiological Modulation	Changes in neural activation and plasticity associated with learning and cognitive adaptation.	Neuroergonomic optimization; accelerated skill acquisition; cognitive performance tuning.	[67,127,128,129,130]
M10 Placebo and Expectancy Effects	Belief-driven modulation of perception, physiology, and performance.	Framing-driven motivation; expectancy-enhanced training; analgesia modulation.	[131,132,133,134]
M11 Biophilia and Nature-Based Effects	Stress reduction and cognitive restoration through exposure to natural environments.	Micro-break recovery environments; burnout reduction; wellbeing design in XR workspaces.	[135,136,137,138]
M12 Meditation and Mindfulness Effects	Attentional and emotional regulation supported by meditation-based XR interventions.	Readiness priming; stress resilience; cognitive reset before high-demand tasks.	[139,140,141]
M13 Immunological Effects	Indirect modulation of immune function through stress and autonomic regulation pathways.	Long-term wellbeing interventions; occupational stress–health mitigation.	[142,143]

Table 5 presents mechanisms related to physiological regulation, neural adaptation, and context-dependent modulation, including biofeedback integration, autonomic regulation, mindfulness-related effects, expectancy effects, and nature-based restoration. These mechanisms influence stress, arousal, learning, and recovery, and are particularly relevant in sustained or high-demand XR–AI environments.

These mechanisms directly affect physiological state, cognitive readiness, and resilience, and can be measured through physiological signals, behavioral performance, and subjective reports. They are therefore well suited for empirical evaluation in XR–AI systems.

In operational contexts, such as digital-twin supervision and remote robotics, these mechanisms contribute to sustained attention, stress regulation, learning, and recovery, and should be considered when designing experiments and evaluating system performance.

• (C) Social and Bodily-State Mechanisms

Table 6. Social, motion-related, and bodily-state mechanisms in the TAXI framework, including action observation, interpersonal synchrony, vestibular engagement, sensory conflict and cybersickness dynamics, and XR-induced analgesia.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M14 Action Observation and Imitation Learning	Learning through observing and replicating embodied actions.	Remote apprenticeship training; expert-guided skill transfer; AI-generated embodied demonstrations.	[144,145,146]
M15 Social Synchrony and Interpersonal Coordination	Temporal and behavioral alignment between individuals supporting coordinated action.	Multi-user assembly; crisis response coordination; time-critical team workflows.	[41,147,148,149,150]
M16 Vestibular Engagement and Motion Cue Integration	Integration of vestibular and sensory cues supporting orientation and motion perception.	Flight and vehicle simulation; navigation tasks; telepresence locomotion; orientation training.	[61,63,73,151,152]
M17 Sensory Conflict and Cybersickness Dynamics	Adverse physiological responses arising from mismatch between sensory inputs and motion cues.	Long-duration XR use; safety-critical applications; cybersickness mitigation design.	[61,63,66,87,89,152]
M18 XR-Induced Analgesia and Pain Modulation	Reduction in pain perception through distraction, embodiment, and affective modulation.	Rehabilitation; procedural tolerance; endurance support in physically demanding tasks.	[67,153,154,155,156]

Table 6. Social, motion-related, and bodily-state mechanisms in the TAXI framework, including action observation, interpersonal synchrony, vestibular engagement, sensory conflict and cybersickness dynamics, and XR-induced analgesia.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M14 Action Observation and Imitation Learning	Learning through observing and replicating embodied actions.	Remote apprenticeship training; expert-guided skill transfer; AI-generated embodied demonstrations.	[144,145,146]
M15 Social Synchrony and Interpersonal Coordination	Temporal and behavioral alignment between individuals supporting coordinated action.	Multi-user assembly; crisis response coordination; time-critical team workflows.	[41,147,148,149,150]
M16 Vestibular Engagement and Motion Cue Integration	Integration of vestibular and sensory cues supporting orientation and motion perception.	Flight and vehicle simulation; navigation tasks; telepresence locomotion; orientation training.	[61,63,73,151,152]
M17 Sensory Conflict and Cybersickness Dynamics	Adverse physiological responses arising from mismatch between sensory inputs and motion cues.	Long-duration XR use; safety-critical applications; cybersickness mitigation design.	[61,63,66,87,89,152]
M18 XR-Induced Analgesia and Pain Modulation	Reduction in pain perception through distraction, embodiment, and affective modulation.	Rehabilitation; procedural tolerance; endurance support in physically demanding tasks.	[67,153,154,155,156]

Table 6 presents mechanisms related to social interaction, bodily state, and motion-dependent perception, including action observation, interpersonal synchrony, vestibular engagement, cybersickness dynamics, and XR-induced analgesia. These mechanisms are particularly relevant in collaborative XR environments, teleoperation, and long-duration use, where social coordination and physiological stability directly affect performance and safety.

These mechanisms influence action synchronization, spatial orientation, motion tolerance, and sustained interaction, and can be measured through behavioral coordination, physiological responses, and performance outcomes.

In operational contexts, such as multi-user collaboration and remote robotics, they play a central role in maintaining coordination, stability, and safe system use.

• (D) Sensorimotor and Action Mechanisms

Table 7. Sensorimotor and action-extension mechanisms in the TAXI framework, including motor skill acquisition, tool incorporation, visuospatial rescaling, and predictive control supporting skilled action and remote manipulation.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M19 Motor Skill Acquisition and Procedural Automation	Skill learning and procedural memory consolidation through immersive practice.	Surgical simulation; industrial assembly; high-dexterity training; sports skill acquisition.	[67,129,130,157,158]
M20 Tool Incorporation and Agency Extension	Integration of external tools into the body schema and control system.	Robot teleoperation; AR-assisted manipulation; exoskeleton control; remote maintenance.	[69,70,78,159,160,161]
M21 Visuospatial Rescaling and Scale Transfer	Perception and action across altered spatial scales (micro to macro).	City-scale digital twins; infrastructure inspection; microscale medical visualization; large-scale logistics control.	[18,23,24,162,163,164,165,166]
M22 Predictive Perception and Anticipatory Control	Feedforward prediction enabling proactive action and error minimization.	Hazard anticipation; predictive maintenance; time-critical remote operations.	[75,85,163,167,168,169,170]

Table 7. Sensorimotor and action-extension mechanisms in the TAXI framework, including motor skill acquisition, tool incorporation, visuospatial rescaling, and predictive control supporting skilled action and remote manipulation.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M19 Motor Skill Acquisition and Procedural Automation	Skill learning and procedural memory consolidation through immersive practice.	Surgical simulation; industrial assembly; high-dexterity training; sports skill acquisition.	[67,129,130,157,158]
M20 Tool Incorporation and Agency Extension	Integration of external tools into the body schema and control system.	Robot teleoperation; AR-assisted manipulation; exoskeleton control; remote maintenance.	[69,70,78,159,160,161]
M21 Visuospatial Rescaling and Scale Transfer	Perception and action across altered spatial scales (micro to macro).	City-scale digital twins; infrastructure inspection; microscale medical visualization; large-scale logistics control.	[18,23,24,162,163,164,165,166]
M22 Predictive Perception and Anticipatory Control	Feedforward prediction enabling proactive action and error minimization.	Hazard anticipation; predictive maintenance; time-critical remote operations.	[75,85,163,167,168,169,170]

Table 7 presents mechanisms related to motor learning, body–tool coupling, spatial scaling, and predictive control. These mechanisms describe how XR–AI systems extend human action capabilities and support skilled performance in virtual, remote, and hybrid environments, particularly in teleoperation, robotics, and high-precision training.

They directly influence task performance, accuracy, reaction time, and procedural learning, and can be evaluated using objective metrics such as task completion time, error rate, and learning speed.

• (E) Affective and Identity Mechanisms

Table 8. Affective, identity-related, and moral–cognitive mechanisms in the TAXI framework, including affective priming, identity plasticity, perspective-taking, and social–affective coupling influencing motivation, trust, empathy, and decision-making.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M23 Affective Priming	Modulation of emotional state through immersive cues and environmental design.	De-escalation training; mood regulation for performance; emotional context shaping in collaboration.	[46,65,171,172,173]
M24 Self-Concept Plasticity	Changes in self-perception through role adoption and embodied identity.	Confidence building; leadership rehearsal; identity-based training scenarios.	[65,98,103,104,105]
M25 Perspective-Taking and Moral–Cognitive Modulation	Modulation of empathy, perspective-taking, and moral reasoning through simulated viewpoint shifts.	Ethics training; policy evaluation; intergroup understanding; conflict mediation.	[37,65,174,175,176]
M26 Social Presence and Affective Resonance	Perceived co-presence and emotional attunement between individuals in XR environments.	Remote teamwork; telemedicine; distributed leadership; collaborative negotiation.	[46,147,148,149,177,178]

Table 8. Affective, identity-related, and moral–cognitive mechanisms in the TAXI framework, including affective priming, identity plasticity, perspective-taking, and social–affective coupling influencing motivation, trust, empathy, and decision-making.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M23 Affective Priming	Modulation of emotional state through immersive cues and environmental design.	De-escalation training; mood regulation for performance; emotional context shaping in collaboration.	[46,65,171,172,173]
M24 Self-Concept Plasticity	Changes in self-perception through role adoption and embodied identity.	Confidence building; leadership rehearsal; identity-based training scenarios.	[65,98,103,104,105]
M25 Perspective-Taking and Moral–Cognitive Modulation	Modulation of empathy, perspective-taking, and moral reasoning through simulated viewpoint shifts.	Ethics training; policy evaluation; intergroup understanding; conflict mediation.	[37,65,174,175,176]
M26 Social Presence and Affective Resonance	Perceived co-presence and emotional attunement between individuals in XR environments.	Remote teamwork; telemedicine; distributed leadership; collaborative negotiation.	[46,147,148,149,177,178]

Table 8 presents mechanisms related to affective modulation, identity plasticity, perspective-taking, and social–emotional coupling in XR–AI environments. These mechanisms influence emotional state, self-representation, interpersonal perception, and decision-making, and are particularly relevant in collaborative, role-based, and ethically sensitive contexts.

They operate at the level of emotional appraisal, self-concept, and social cognition, shaping motivation, trust, cooperation, and judgment, and can be evaluated through behavioral, physiological, and subjective measures.

• (F) Cognitive and Neuroergonomic Mechanisms

Table 9. Cognitive augmentation and neuroergonomic mechanisms in the TAXI framework, including cognitive offloading, distributed cognition, metacognitive calibration, trust regulation, workload control, fatigue dynamics, adaptive learning, and sustained engagement.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M27 Cognitive Offloading and Externalized Cognition	Use of external representations or AI to support memory, reasoning, and task execution.	Guidance overlays; procedural checklists; annotated digital twins.	[69,75,85,159,179]
M28 Shared Situation Awareness and Distributed Cognition	Alignment of mental models across individuals enabling coordinated understanding and decision-making.	Multi-team coordination; digital twin operations centers; distributed mission planning; real-time collaborative decision-making in complex environments.	[41,69,75,81,85,180]
M29 Metacognitive Calibration and Confidence Regulation	Alignment between confidence and actual performance through feedback and self-monitoring.	Competency validation; reducing overconfidence; operator self-check workflows.	[143,163,170,181,182,183]
M30 Trust Calibration and Automation Reliance	Adjustment of reliance on AI systems to avoid overuse or underuse.	AI copilots; explainable decision support; supervisory control and override training.	[37,49,75,85,184,185]
M31 Fatigue Dynamics and Vigilance Sustainment	Changes in alertness and vigilance over time during prolonged task engagement.	Long-duration monitoring; control rooms; transport safety operations.	[86,114,186,187]
M32 Cognitive Workload Regulation (Neuroergonomic Load Shaping)	Adjustment of task demand and information load to maintain performance.	Adaptive interfaces; complexity management; multitask coordination.	[42,75,86,114,116]
M33 Adaptive Error Shaping and Learning Acceleration	Use of adaptive feedback and task difficulty to optimize learning and performance improvement.	Intelligent tutoring; adaptive simulators; accelerated skill acquisition.	[129,130,158,188,189]
M34 Flow State Induction and Sustained Engagement	Facilitation of deep engagement and sustained high-performance states.	Long-duration complex tasks; training adherence; creative problem solving.	[65,189,190,191]

Table 9. Cognitive augmentation and neuroergonomic mechanisms in the TAXI framework, including cognitive offloading, distributed cognition, metacognitive calibration, trust regulation, workload control, fatigue dynamics, adaptive learning, and sustained engagement.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M27 Cognitive Offloading and Externalized Cognition	Use of external representations or AI to support memory, reasoning, and task execution.	Guidance overlays; procedural checklists; annotated digital twins.	[69,75,85,159,179]
M28 Shared Situation Awareness and Distributed Cognition	Alignment of mental models across individuals enabling coordinated understanding and decision-making.	Multi-team coordination; digital twin operations centers; distributed mission planning; real-time collaborative decision-making in complex environments.	[41,69,75,81,85,180]
M29 Metacognitive Calibration and Confidence Regulation	Alignment between confidence and actual performance through feedback and self-monitoring.	Competency validation; reducing overconfidence; operator self-check workflows.	[143,163,170,181,182,183]
M30 Trust Calibration and Automation Reliance	Adjustment of reliance on AI systems to avoid overuse or underuse.	AI copilots; explainable decision support; supervisory control and override training.	[37,49,75,85,184,185]
M31 Fatigue Dynamics and Vigilance Sustainment	Changes in alertness and vigilance over time during prolonged task engagement.	Long-duration monitoring; control rooms; transport safety operations.	[86,114,186,187]
M32 Cognitive Workload Regulation (Neuroergonomic Load Shaping)	Adjustment of task demand and information load to maintain performance.	Adaptive interfaces; complexity management; multitask coordination.	[42,75,86,114,116]
M33 Adaptive Error Shaping and Learning Acceleration	Use of adaptive feedback and task difficulty to optimize learning and performance improvement.	Intelligent tutoring; adaptive simulators; accelerated skill acquisition.	[129,130,158,188,189]
M34 Flow State Induction and Sustained Engagement	Facilitation of deep engagement and sustained high-performance states.	Long-duration complex tasks; training adherence; creative problem solving.	[65,189,190,191]

Table 9 presents mechanisms related to cognitive augmentation, workload regulation, human–AI interaction, and sustained performance under complex conditions. These mechanisms support distributed cognition, adaptive assistance, trust calibration, fatigue management, and performance stabilization in long-duration XR–AI use.

They determine how effectively users manage complexity, maintain situation awareness, and interact with AI-supported systems, and can be evaluated through behavioral, physiological, and performance-based measures.

• (G) Distributed and Adaptive System Mechanisms

Table 10. Distributed presence, agent-mediated, and long-term adaptation mechanisms in the TAXI framework, including temporal simulation, spatial reasoning support, distributed attention, proxy agency, agent-guided attention, neuroplastic adaptation, future-self embodiment, and reflective learning.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M35 Mental Time Travel and Temporal Simulation	Simulation of past experiences and possible future scenarios in an embodied form.	Scenario planning; disaster rehearsal; strategic foresight.	[65,192,193]
M36 High-Dimensional Abstraction and Spatial Reasoning Support	Understanding complex systems through spatialized and high-dimensional representations.	AI model interpretability; complex systems education; industrial and scientific visualization.	[20,92,162,163,164]
M37 Parallel Presence and Distributed Attentional Allocation	Allocation of attention across multiple environments or representations simultaneously.	Multi-site monitoring; distributed supervision; simultaneous remote inspection.	[81,96,114,170,194]
M38 Delegated Embodiment and Proxy Agency	Delegation of actions to agents or avatars operating on user intent.	Agent-based monitoring; smart-city delegation; escalation management.	[33,70,195,196,197]
M39 Attentional Orchestration by Autonomous Agents	AI-driven prioritization and direction of user attention toward relevant events.	Safety monitoring; predictive alerting; operations center triage.	[33,75,85,170,185]
M40 Long-Term Neuroplastic Adaptation	Persistent neural changes resulting from repeated XR interaction.	Long-term training; workforce upskilling; sustained XR use.	[67,127,128,130]
M41 Future-Self Embodiment and Motivational Commitment	Changes in motivation driven by embodied simulation of future selves.	Health adherence; sustainability behavior; long-term goal reinforcement.	[65,189,198]
M42 Replay-Based Self-Modeling and Reflective Learning	Learning through replay, embodied review, and self-model refinement.	Training debrief; expert coaching; post-incident analysis.	[129,130,188,199]

Table 10. Distributed presence, agent-mediated, and long-term adaptation mechanisms in the TAXI framework, including temporal simulation, spatial reasoning support, distributed attention, proxy agency, agent-guided attention, neuroplastic adaptation, future-self embodiment, and reflective learning.

TAXI Mechanism (No. + Name)	Short Definition	Example XI Use Contexts	Supporting Literature
M35 Mental Time Travel and Temporal Simulation	Simulation of past experiences and possible future scenarios in an embodied form.	Scenario planning; disaster rehearsal; strategic foresight.	[65,192,193]
M36 High-Dimensional Abstraction and Spatial Reasoning Support	Understanding complex systems through spatialized and high-dimensional representations.	AI model interpretability; complex systems education; industrial and scientific visualization.	[20,92,162,163,164]
M37 Parallel Presence and Distributed Attentional Allocation	Allocation of attention across multiple environments or representations simultaneously.	Multi-site monitoring; distributed supervision; simultaneous remote inspection.	[81,96,114,170,194]
M38 Delegated Embodiment and Proxy Agency	Delegation of actions to agents or avatars operating on user intent.	Agent-based monitoring; smart-city delegation; escalation management.	[33,70,195,196,197]
M39 Attentional Orchestration by Autonomous Agents	AI-driven prioritization and direction of user attention toward relevant events.	Safety monitoring; predictive alerting; operations center triage.	[33,75,85,170,185]
M40 Long-Term Neuroplastic Adaptation	Persistent neural changes resulting from repeated XR interaction.	Long-term training; workforce upskilling; sustained XR use.	[67,127,128,130]
M41 Future-Self Embodiment and Motivational Commitment	Changes in motivation driven by embodied simulation of future selves.	Health adherence; sustainability behavior; long-term goal reinforcement.	[65,189,198]
M42 Replay-Based Self-Modeling and Reflective Learning	Learning through replay, embodied review, and self-model refinement.	Training debrief; expert coaching; post-incident analysis.	[129,130,188,199]

Table 10 presents mechanisms relevant to persistent XR–AI systems, including distributed attention, proxy agency, agent-guided interaction, long-term adaptation, and reflective learning. These mechanisms characterize interaction in environments where immersive interfaces are coupled with adaptive agents and users engage over extended periods.

They support distributed attention, delegation to automated agents, adaptation to evolving system states, and sustained performance over time, and are particularly relevant for large-scale, continuous XR–AI deployments.

Because these mechanisms unfold over longer time scales, they are important for evaluating long-term learning, adaptation, and potential side effects of XR–AI use.

3.3. Interpretation, Scope, and Evidentiary Status of the TAXI Inventory

Interpretation of the mechanism inventory:

The TAXI inventory should be interpreted as a compositional set rather than a list of independent effects. In XR–AI environments, multiple mechanisms typically co-occur, and their interaction determines performance, safety, and usability. The taxonomy supports identification of relevant mechanisms within a scenario and analysis of how their combinations influence higher-level outcomes.

Scope of included mechanisms:

The TAXI inventory represents a synthesis of psychobiological mechanisms that are supported across multiple research domains. This integrative approach is consistent with established scoping review methodologies, which are designed to structure heterogeneous bodies of evidence, map key concepts, and identify relationships across interdisciplinary research areas [200,201]. These include mechanisms extensively studied in XR contexts (e.g., presence, embodiment, multisensory integration), as well as mechanisms established in adjacent fields such as neuroscience, neuroergonomics, and human–automation interaction (e.g., distributed cognition, trust calibration, workload regulation).

All included mechanisms are grounded in empirical or theoretically well-established research on human perception, cognition, physiology, or behavior. Differences between mechanisms reflect the contexts in which they have been studied, rather than differences in their conceptual validity. Their integration within TAXI reflects the objective of the taxonomy: to capture mechanisms that are relevant to XR–AI interaction and extended intelligence systems, regardless of the disciplinary origin of their supporting evidence.

Role of the taxonomy in the present work:

TAXI provides a structured mechanism-level foundation for analyzing XR–AI interaction. Rather than defining a closed or exhaustive set of effects, it supports systematic identification of relevant processes, comparison across scenarios, and formulation of experimentally testable hypotheses.

Transition to capability patterns:

While TAXI describes how XR–AI interaction affects the human user at the level of underlying mechanisms, real-world system performance is evaluated in terms of functional outcomes. Such outcomes do not arise from individual mechanisms, but from their coordinated interaction under task and system constraints.

This motivates the introduction of a second abstraction layer, in which TAXI mechanisms are organized into Extended Intelligence Capability Augmentation Patterns (XI-CAP), representing higher-level performance constructs derived from mechanism convergence. The relationship between these two layers forms the basis of the TAXI–XI-CAP framework introduced in the following section. In this sense, TAXI functions as a mechanism-level integration layer, enabling consistent analysis across XR–AI systems that would otherwise be described using fragmented domain-specific constructs.

4. Extended Intelligence Capability Augmentation Patterns (XI-CAP)

4.1. From Psychobiological Mechanisms to Capability Constructs

The TAXI framework provides a mechanism-level description of how XR–AI interaction influences human perception, cognition, physiology, affect, sensorimotor control, and social coordination. However, in applied XR–AI contexts, system performance is evaluated in terms of functional outcomes rather than individual mechanisms. These outcomes—such as precision, learning speed, coordination, or sustained performance—arise from the coordinated interaction of multiple mechanisms under task and system constraints [75,85,87,170].

To bridge this gap, a capability-level abstraction is introduced in which TAXI mechanisms are organized into Extended Intelligence Capability Augmentation Patterns (XI-CAP). Each XI-CAP construct represents a performance-level pattern emerging from the convergence of multiple psychobiological mechanisms within XR–AI environments.

The XI-CAP constructs were derived through structured conceptual synthesis, based on the identification of mechanism clusters that converge on shared functional outcomes. This derivation builds on the methodological approach outlined in Section 2.6 and reflects recurring patterns of co-occurrence and functional interdependence across perceptual, cognitive, sensorimotor, regulatory, and social domains [69,75,85,159].

Candidate capability patterns were retained when they satisfied the following criteria:

• convergence of multiple validated TAXI mechanisms toward a shared functional outcome;
• relevance to observable performance outcomes in XR–AI or human-factors contexts;
• applicability to realistic deployment scenarios, including high-precision, long-duration, or distributed tasks;
• non-reducibility to a single mechanism without loss of explanatory clarity;
• potential for empirical evaluation using behavioral, physiological, or task-performance measures.

This approach is consistent with established perspectives in neuroergonomics, distributed cognition, and human–automation interaction, where performance outcomes are understood as emergent properties of interacting subsystems rather than isolated variables [75,85,114,159,163].

Accordingly, XI-CAP constructs should be interpreted as structured, hypothesis-driven capability patterns that define the level at which XR–AI performance effects can be formulated, compared, and experimentally tested. They do not introduce new mechanisms, but organize existing, empirically supported processes into capability-oriented constructs relevant to system design and evaluation [87,91].

4.2. XI Capability Augmentation Patterns (XI-CAP)

Based on the derivation process described above, twelve Extended Intelligence Capability Augmentation Patterns (XI-CAP) were identified. Each construct represents a capability-level pattern that emerges when specific combinations of TAXI mechanisms are present and sufficiently coordinated within XR–AI environments.

The XI-CAP constructs capture functional performance outcomes relevant to applied contexts such as teleoperation, immersive training, digital twin supervision, and distributed human–AI collaboration. They are defined at a level that enables comparison across systems and supports experimental evaluation using behavioral, physiological, and task-based measures.

The XI Capability Augmentation Patterns (XI-CAP) and supporting mechanisms are:

C1 Remote Super-Dexterity: Capacity to perform fine, high-precision actions remotely with near co-located control fidelity—maps onto M1 Presence and Immersion; M2 Virtual Embodiment; M4 Multisensory Integration; M19 Motor Skill Acquisition and Procedural Automation; M20 Tool Incorporation and Agency Extension; M22 Predictive Perception and Anticipatory Control.

C2 Scale Transcendence Action (Micro–Macro): Capacity to perceive and act effectively across non-human spatial scales—maps onto M1 Presence and Immersion; M2 Virtual Embodiment; M4 Multisensory Integration; M21 Visuospatial Rescaling and Scale Transfer; M36 High-Dimensional Abstraction and Spatial Reasoning Support.

C3 Accelerated Skill Acquisition and Transfer: Accelerated acquisition and transfer of skills under immersive conditions—maps onto M6 Altered Time Perception; M9 Neurobiological Modulation; M14 Action Observation and Imitation Learning; M19 Motor Skill Acquisition and Procedural Automation; M33 Adaptive Error Shaping and Learning Acceleration; M40 Long-Term Neuroplastic Adaptation; M42 Replay-Based Self-Modeling and Reflective Learning.

C4 Continuous Cognitive Load Regulation: Dynamic regulation of attention, arousal, and workload during complex tasks—maps onto M5 Attentional Modulation; M7 Biofeedback/Neurofeedback Integration; M8 Physiological Regulation; M12 Meditation and Mindfulness Effects; M31 Fatigue Dynamics and Vigilance Sustainment; M32 Cognitive Workload Regulation; M34 Flow State Induction and Sustained Engagement; M39 Attentional Orchestration by Autonomous Agents.

C5 Resilient Long-Duration Performance: Maintenance of performance quality under prolonged use—maps onto M6 Altered Time Perception; M8 Physiological Regulation; M10 Placebo and Expectancy Effects; M11 Biophilia and Nature-Based Effects; M12 Meditation and Mindfulness Effects; M31 Fatigue Dynamics and Vigilance Sustainment; M32 Cognitive Workload Regulation; M40 Long-Term Neuroplastic Adaptation.

C6 Pain-Tolerant Task Execution: Maintenance of task performance despite discomfort or minor pain—maps onto M5 Attentional Modulation; M10 Placebo and Expectancy Effects; M18 XR-Induced Analgesia and Pain Modulation.

C7 Collective Distributed Cognition: Coordinated group-level cognition and action across distributed agents—maps onto M15 Social Synchrony and Interpersonal Coordination; M26 Social Presence and Affective Resonance; M28 Shared Situation Awareness and Distributed Cognition; M30 Trust Calibration and Automation Reliance.

C8 Identity-Based Performance Modulation: Modulation of behavior and performance through role-based embodiment—maps onto M2 Virtual Embodiment; M3 Proteus Effect; M23 Affective Priming; M24 Self-Concept Plasticity.

C9 Moral and Perspective Expansion: Enhanced ethical reasoning and empathic perspective-taking through immersive experience—maps onto M25 Perspective-Taking and Moral–Cognitive Modulation; M26 Social Presence and Affective Resonance.

C10 Human–AI Co-Regulated Action: Joint regulation of perception, decision-making, and action between humans and AI systems—maps onto M27 Cognitive Offloading and Externalized Cognition; M28 Shared Situation Awareness and Distributed Cognition; M29 Metacognitive Calibration and Confidence Regulation; M30 Trust Calibration and Automation Reliance; M32 Cognitive Workload Regulation; M39 Attentional Orchestration by Autonomous Agents; M20 Tool Incorporation and Agency Extension.

C11 Embodied Tool and Machine Assimilation: Integration of tools or machines into the user’s sense of agency and control—maps onto M2 Virtual Embodiment; M19 Motor Skill Acquisition and Procedural Automation; M20 Tool Incorporation and Agency Extension; M22 Predictive Perception and Anticipatory Control.

C12 Multi-Space Supervision and Agent-Orchestrated Attention: Distributed awareness across multiple environments supported by agent-driven prioritization and delegation—maps onto M37 Parallel Presence and Distributed Attentional Allocation; M38 Delegated Embodiment and Proxy Agency; M39 Attentional Orchestration by Autonomous Agents; M28 Shared Situation Awareness and Distributed Cognition; M30 Trust Calibration and Automation Reliance; M32 Cognitive Workload Regulation.

The XI-CAP constructs represent convergence patterns rather than independent variables. Each capability emerges from the interaction of multiple mechanisms, and no single mechanism is sufficient to produce these outcomes in isolation. Full definitions and interpretive details for each XI-CAP construct are provided in Document S2.

Table 11. Extended Intelligence Capability Augmentation Patterns (XI-CAP) derived from TAXI mechanisms in XR–AI environments, illustrating how multiple psychobiological mechanisms converge to enable higher-level performance outcomes relevant to remote operation, distributed cognition, adaptive control, and long-duration interaction.

No.	XI Capability	Short Definition	Supporting TAXI Mechanisms	Supporting Literature
C1	Remote Super-Dexterity	Capacity to perform fine, high-precision actions remotely with near co-located control fidelity	M1, M2, M4, M19, M20, M22	[72,97,106,130,160,167]
C2	Scale Transcendence Action (Micro–Macro)	Capacity to perceive and act effectively across non-human spatial scales	M1, M2, M4, M21, M36	[18,73,162,164]
C3	Accelerated Skill Acquisition and Transfer	Accelerated acquisition of skills under immersive conditions	M6, M9, M14, M19, M33, M40, M42	[120,128,130,144,158,188,199]
C4	Continuous Cognitive Load Regulation	Potential for dynamic regulation of attention, arousal, and workload during complex tasks	M5, M7, M8, M12, M31, M32, M34, M39	[113,114,122,141,185,186,190]
C5	Resilient Long-Duration Performance	Maintenance of performance quality under prolonged use	M6, M8, M10, M11, M12, M31, M32, M40	[120,128,133,136,137,187]
C6	Pain-Tolerant Task Execution	Potential to maintain task performance despite discomfort or minor pain	M5, M10, M18	[133,134,152]
C7	Collective Distributed Cognition	Hypothesized capacity for coordinated group-level cognition and action across distributed agents	M15, M26, M28, M30	[69,75,147,148]
C8	Identity-Based Performance Modulation	Potential for modulation of behavior and performance through role-based embodiment	M2, M3, M23, M24	[100,103,172]
C9	Moral and Perspective Expansion	Potential for enhanced ethical reasoning and empathic perspective-taking through immersive experience	M25, M26	[175,176]
C10	Human–AI Co-Regulated Action	Capacity for joint regulation of perception, decision-making, and action between humans and AI systems	M27, M28, M29, M30, M32, M39, M20	[75,85,159,182,185]
C11	Embodied Tool and Machine Assimilation	Potential for integration of tools or machines into the user’s sense of agency and control	M2, M19, M20, M22	[78,130,160,167]
C12	Multi-Space Supervision and Agent-Orchestrated Attention	Hypothesized capacity for distributed awareness across multiple environments supported by agent-driven prioritization and delegation	M37, M38, M39, M28, M30, M32	[75,114,185,194,196]

summarizes the XI-CAP constructs, including short definitions, dominant contributing TAXI mechanisms, and representative application contexts.

The XI-CAP constructs represent convergence patterns rather than independent variables. Each capability emerges from the interaction of multiple mechanisms, and no single mechanism is sufficient to produce these outcomes in isolation. Expanded definitions in Document S2 specify the contributing mechanisms, operational scope, measurement variables, and design implications for each capability, supporting their interpretation and empirical evaluation.

The XI-CAP layer serves as an intermediate abstraction linking mechanism-level evidence to experimentally testable capability constructs. Rather than focusing on individual mechanisms, it enables analysis of how combinations of mechanisms give rise to specific performance outcomes and how these outcomes can be evaluated under operational conditions [87,170].

This perspective is consistent with research in neuroscience, cognitive science, and human factors, where complex performance outcomes—such as sustained attention, fine motor control, and coordinated teamwork—emerge from interacting perceptual, cognitive, physiological, and socio-affective processes [113,114].

Accordingly, XI-CAP defines the level at which capability claims can be formulated and empirically tested. Each capability represents a convergent pattern rather than an isolated effect, providing a structured basis for experimental evaluation and XR–AI system design.

4.3. Mapping Mechanisms to Capability Patterns

The relationship between TAXI mechanisms and XI-CAP constructs is many-to-many and should be understood as a convergence pattern rather than a fixed mapping. Individual capabilities arise from the interaction of multiple psychobiological mechanisms spanning perceptual, cognitive, sensorimotor, regulatory, and social domains. Conversely, a single mechanism may contribute to multiple capabilities depending on task context, system design, and user characteristics.

Accordingly, XI-CAP constructs should be interpreted as conditional and context-dependent outcomes rather than deterministic effects. This convergence logic provides a synthesis layer linking mechanism-level evidence to performance-oriented constructs, supporting both comparative analysis and empirical validation across XR–AI systems. This many-to-many relationship forms the conceptual basis for the capability-oriented analysis and validation framework developed in the following sections.

4.4. Operationalizing XI-CAP Constructs

Figure 3 illustrates a mechanism-level operationalization of XI capability evaluation, showing how multiple TAXI mechanisms can be linked to a target XI-CAP construct and associated measurement variables.

The example demonstrates how capability claims—such as remote super-dexterity—can be evaluated by identifying contributing mechanisms and measuring their combined effect under controlled conditions. Rather than treating capabilities as abstract constructs, this approach enables their translation into experimentally testable configurations.

In practice, operationalization involves:

• selecting a target capability (e.g., remote super-dexterity);
• identifying the contributing TAXI mechanisms;
• defining measurable variables for each mechanism (e.g., task accuracy, physiological load, latency tolerance); and
• evaluating their combined effect under controlled or ecologically valid conditions.

TAXI is designed to support mechanism composition rather than mutually exclusive categorization. Individual mechanisms may contribute to multiple capability patterns, and multiple mechanisms may operate concurrently to produce performance effects, particularly in XR–AI environments involving adaptive systems, multisensory feedback, and distributed interaction.

Mechanisms operate across multiple levels of analysis, ranging from intrapersonal psychobiological processes (e.g., attentional modulation, physiological regulation) to system-level interaction processes such as distributed presence, shared situation awareness, and agent-mediated coordination. Accordingly, TAXI functions as a set of analytic building blocks for describing complex XR–AI interaction states and for guiding capability-oriented analysis and empirical investigation.

5. Research Agenda: Priorities for Validating and Scaling XI

5.1. Research Priorities

The TAXI–XI-CAP framework indicates that XR–AI systems depend on the coordinated and stable interaction of multiple psychobiological mechanisms operating under real-world conditions. TAXI defines the mechanism layer, while XI-CAP provides capability-level constructs that can be formulated as testable hypotheses. The central challenge is therefore not whether individual mechanisms exist, but whether they can be reliably combined into stable and reproducible performance outcomes [76,86].

As XR–AI systems evolve into persistent cyber–physical infrastructures—such as digital twins and distributed collaboration environments—the focus shifts from short-duration effects to long-duration stability, repeated use, adaptive AI mediation, and multi-user coordination [87,90,91]. These conditions introduce time-dependent dynamics, including fatigue, attentional drift, habituation, and shifts in automation reliance, which directly affect system performance and reliability [75,128,185,186,187].

These challenges are particularly critical in safety-sensitive domains such as remote surgery, industrial teleoperation, and infrastructure supervision, where human factors directly influence system reliability and risk [88,170,202]. This underscores the need for systematic and proactive validation prior to large-scale deployment.

Addressing these challenges requires validation infrastructures capable of evaluating XR–AI systems as integrated, closed-loop environments. This includes mechanism-sensitive measurement, ecologically valid experimental protocols, and evaluation frameworks that account for AI-driven adaptation, transparency, and user reliance [75,85,185]. In addition, practical constraints such as ergonomics, accessibility, and sustainability must be incorporated into evaluation design [49,51,87].

A structured research agenda is therefore required to guide the transition from mechanism-level evidence to deployment-ready XR–AI systems. See Figure 4 for a schematic representation of this process.

The priorities proposed here focus on mechanism validation, capability benchmarking, risk mitigation, and infrastructure readiness, defining the empirical pathways needed to establish reliable and scalable performance.

Figure 4 and

Table 12. Research priority mapping across XI-CAP capability patterns and TAXI mechanisms, organizing priority themes (R1–R15) as translational validation challenges. Each priority links mechanism-level processes to capability outcomes, deployment constraints, and evaluation contexts relevant to XR–AI systems, providing a structured and reproducible basis for empirical validation and XR-AI research planning.

Priority Theme	TAXI/XI-CAP Linkages	Evaluation Context/Benchmarks	Supporting Literature
R1 Mechanism validity under operational load	TAXI: M1–M12, M30–M32; XI-CAP: C1, C4, C5, C10, C11	Control rooms; industrial pilots; shift-work simulations; long-duration exposure (4–8 h); high-stakes tasks; time-on-task stability	[86,116,186]
R2 Multi-mechanism synergy and interference	TAXI: M1–M9, M14–M15, M23–M34, M37–M39; XI-CAP: C1–C6, C10, C12	Factorial XR experiments; multimodal cue combinations; controlled interaction studies; synergy/interference mapping under prolonged use	[76,86]
R3 Workload shaping via adaptive XR–AI	TAXI: M5, M27, M29–M30, M32, M39; XI-CAP: C4, C5, C10, C12	Adaptive interfaces; real-time workload sensing; adaptive pacing experiments; performance and error benchmarking under dynamic UI conditions	[75,86,185]
R4 Fatigue and vigilance sustainment	TAXI: M6, M8, M31–M32, M34; XI-CAP: C5	Extended monitoring; vigilance decrement curves; endurance simulations (8–12 h); cumulative fatigue tracking	[120,186,187]
R5 Trust calibration and safe AI reliance	TAXI: M28–M30, M39; XI-CAP: C10, C12	AI reliability manipulation experiments; override drills; human–AI teaming benchmarks; longitudinal trust drift monitoring	[75,85,185]
R6 Distributed cognition and shared situation awareness	TAXI: M15, M26–M28; XI-CAP: C7	Multi-team simulations; digital twin operations; collaborative mission rehearsal; coordination stability over time	[69,147,150]
R7 Proxy agents and distributed supervisory intelligence	TAXI: M37–M39, M30–M32; XI-CAP: C12 (and C10)	Multi-asset monitoring; anomaly response testing; escalation and triage evaluation; false-alarm fatigue	[75,185]
R8 Embodiment and agency extension boundaries	TAXI: M2, M16–M17, M19–M20, M22; XI-CAP: C1, C11	Teleoperation latency trials; cybersickness testing; agency attribution; long-duration headset use; multi-day tolerance	[78,154]
R9 Training transfer and reflective replay learning	TAXI: M14, M19, M33, M40, M42; XI-CAP: C3	Certification benchmarks; simulator-to-real transfer; replay-based debriefing; retention tests; stress-condition transfer	[129,130,144]
R10 Affective priming and ethical performance optimization	TAXI: M23, M10, M8 (opt. M24); XI-CAP: C8, C9	Nudging experiments; affect modulation safety testing; stress–performance trade-offs; longitudinal affect drift	[132,172]
R11 Social presence and empathy stability	TAXI: M25–M26 (opt. M23–M24); XI-CAP: C9	Longitudinal empathy studies; behavioral follow-up; adverse outcome monitoring; emotional fatigue	[175,176]
R12 Temporal simulation and counterfactual planning	TAXI: M35, M42, M10; XI-CAP: cross-cutting (C10, C12)	Disaster rehearsal; foresight experiments; counterfactual reasoning evaluation; planning accuracy stability	[192,193]
R13 High-dimensional abstraction and systems insight	TAXI: M36, M27, M32 (opt. M39); XI-CAP: C2	Visualization benchmarks; interpretability tasks; systems debugging; delayed insight recall	[164]
R14 Longitudinal adaptation and unintended consequences	TAXI: M9, M27, M30–M32, M40; XI-CAP: C3, C5, C10	Longitudinal pilots; cognitive drift tracking; wellbeing baselines; multi-week XR–AI exposure; dependency indicators	[127,128,186]
R15 Governance, ethics, safety, and interoperability	Cross-cutting (all mechanisms; C1–C12)	Standards mapping; regulatory readiness; accessibility validation; auditability; sustainability assessment	[49,51,53,202]

summarize these priorities by mapping research themes to TAXI mechanisms and XI-CAP capability patterns. Together, they highlight the need for multi-level evaluation across psychobiological processes, capability outcomes, and operational constraints.

Table 12 organizes the research agenda into priority themes spanning multiple TAXI mechanisms and XI-CAP capability patterns. Each priority represents a translational challenge linking mechanism-level evidence to capability outcomes, deployment constraints, and validation requirements for XR–AI systems [86,87,91,185].

The inclusion of a priority should not be interpreted as evidence of deployment readiness. Rather, these priorities identify areas where validation remains incomplete, long-duration effects are insufficiently characterized, or technical and human factors constraints have yet to be reconciled [154,170,186].

The priorities are structured from foundational challenges—such as mechanism stability and measurement infrastructure—to system-level considerations, including governance, interoperability, and long-term societal impact [49,53,87,202]. Expanded descriptions are provided in Document S3.

Collectively, these priorities indicate that XI cannot be achieved through isolated advances in XR hardware, artificial intelligence, or interface design. Progress instead requires integrated research infrastructures capable of evaluating XR–AI systems under long-duration, high-variability, and safety-critical conditions [75,86,185]. Establishing such infrastructures is a prerequisite for reliable deployment across domains such as industrial digital twins, remote healthcare, immersive training, and large-scale cyber–physical coordination, enabling the transition from capability hypotheses to validated and deployable XR–AI systems.

5.2. Toward Integrative, Interdisciplinary Research Infrastructures

Across the identified priorities, a consistent requirement is the development of research infrastructures capable of evaluating XR–AI systems as integrated human–XR–AI environments. Many of the effects described in the TAXI–XI-CAP framework emerge only through sustained interaction, adaptive system behavior, and multi-mechanism coupling, and cannot be captured through short-duration or single-variable studies [86,91,116,170,186].

Progress will therefore depend on interdisciplinary testbeds that combine immersive XR platforms, multimodal sensing, real-time AI, and ecologically valid task environments. These systems must support longitudinal monitoring of workload, fatigue, trust, adaptation, and adverse effects under realistic operational constraints, including latency, device limitations, and multi-user coordination [75,87,154,185,187].

To support reproducibility and cross-domain comparison, future research will require shared datasets, standardized protocols, and interoperable benchmarking frameworks [202]. Coordinated funding and cross-sector collaboration will be necessary to sustain long-duration and deployment-scale validation efforts [49,51,53,87].

Without such infrastructures, XI risks developing as fragmented prototypes rather than a scalable and scientifically grounded capability layer. Establishing shared validation environments is therefore a prerequisite for reliable, reproducible, and responsible deployment of XR–AI systems at scale.

5.3. Stakeholder Alignment and Translational Relevance

The proposed research agenda is oriented toward multiple stakeholder domains, including industry, medicine, education, and governance. As XR–AI systems evolve into persistent infrastructures, each domain introduces distinct performance requirements, risk constraints, and validation standards.

Industrial stakeholders prioritize precision, reliability, and safety in applications such as teleoperation and digital twin supervision, requiring validation of mechanisms related to embodiment, predictive control, workload regulation, and trust calibration [75,86,87,187].

Medical applications emphasize reproducibility, ethical safeguards, and clinically validated outcomes, particularly in training, rehabilitation, and decision-support contexts [129,185].

Educational use cases focus on scalability, accessibility, and learning transfer, requiring evidence that immersive environments improve retention and skill acquisition without excessive cognitive load or adverse effects [129,154].

Governance and smart-city applications introduce requirements related to interoperability, transparency, and system-level trust, particularly in large-scale cyber–physical coordination environments [90,202].

Mapping research priorities to TAXI mechanisms and XI-CAP capability patterns clarifies how mechanism-level evidence translates into domain-specific system requirements. This alignment supports the development of shared evaluation standards and enables comparison across applications, contributing to the establishment of XI as a reliable, interoperable, and scalable infrastructure layer rather than a collection of isolated XR applications.

6. Conclusions

This review examined the convergence of immersive XR and adaptive AI within emerging cyber–physical infrastructures, including the Spatial Web, digital twins, and smart-city systems [90,202]. Within this context, XI is framed as a human-centered paradigm in which performance emerges from closed-loop interaction between humans, immersive interfaces, and adaptive computational agents.

The TAXI framework provides a taxonomy of psychobiological mechanisms through which XR–AI interaction influences perception, cognition, physiology, affect regulation, sensorimotor control, and social coordination. XI Capability Augmentation Patterns (XI-CAP) extend this structure by organizing these mechanisms into capability-oriented constructs relevant to applied domains such as training, teleoperation, and distributed collaboration.

Together, TAXI and XI-CAP establish a structured pathway from mechanism-level evidence to experimentally testable capability constructs and system-level evaluation. This framing emphasizes that XR–AI performance must be operationalized and validated under realistic conditions rather than inferred from isolated effects.

The present work has several limitations. The TAXI mechanism inventory represents a theory-guided synthesis rather than an exhaustive catalog, and the evidentiary maturity of included mechanisms varies across domains. While the framework integrates findings from XR, neuroscience, human factors, and human–AI interaction, not all mechanisms have been empirically validated under long-duration, adaptive XR–AI conditions. Similarly, the XI-CAP constructs are proposed as convergent capability patterns that require systematic experimental validation across contexts, tasks, and user populations. The framework should therefore be understood as a structured basis for hypothesis generation, comparative analysis, and empirical validation of XR–AI systems, rather than a complete or definitive account of validated effects.

A central implication is that XR–AI effects must be understood as interacting systems rather than independent variables. Performance emerges from coordinated perceptual, cognitive, physiological, and social processes, consistent with research in neuroergonomics and human–automation interaction [69,75,86].

Capability amplification should not be assumed to be inherently beneficial. The same mechanisms that support improved learning or coordination may also introduce workload, fatigue, sensory conflict, or automation-related risks if poorly calibrated. As XR–AI systems move toward persistent use, these constraints must be treated as core design requirements alongside technical parameters [87,154,187].

From an infrastructure perspective, XR–AI systems are increasingly embedded within large-scale cyber–physical ecosystems, including industrial digital twins, remote collaboration platforms, and smart-city coordination environments [90,202]. As in other safety-critical domains, performance depends on stable coupling between human intention, system behavior, and environmental feedback [78,86].

At the governance level, the development of XI systems must align with human-centered principles emphasizing safety, transparency, accessibility, and societal benefit. While XR and AI offer opportunities for scalable training, healthcare access, and decision support, their combined deployment requires validated evidence, standardized evaluation, and clearly defined human factors constraints [87].

The TAXI–XI-CAP framework provides a foundation for analyzing how XR–AI systems may extend human capability while introducing new constraints and risks. Advancing XI will require shared validation infrastructures, longitudinal studies, and interdisciplinary research capable of evaluating XR–AI interaction as a continuous human–machine system.

XI should therefore be evaluated not by immersion alone, but by its capacity to sustain stable, safe, and meaningful human performance over time under real-world conditions.