Artificial Intelligence-Based Digital Twins Metaverse to Improve STEM Work Skills and Adaptive Expertise in Higher Education: A Conceptual Framework

1. Introduction

The accelerating convergence of Artificial Intelligence (AI), immersive virtual environments, and cyber-physical systems is reshaping how higher education prepares learners for the complexities of emerging Industry 4.0 and 5.0 ecosystems. In STEM disciplines, learners must master not only technical knowledge but also employability skills such as problem-solving, collaboration, digital literacy, and adaptability to remain competitive in increasingly automated and AI-driven workplaces. Recent advancements in Digital Twins (DTs), defined as real-time virtual replicas of physical systems, offer significant potential for enhancing experiential learning by replicating authentic industrial processes with high fidelity and dynamic feedback (Khan & Ahmad, 2025). DTs have been widely adopted within cyber-physical manufacturing, industrial metaverse systems, and algorithmic decision-support infrastructures (Lazaroiu et al., 2024; Kliestik et al., 2024), yet their integration into higher education remains fragmented and under-theorized.

Parallel developments in the Metaverse have introduced immersive, persistent, and interactive virtual environments capable of supporting experiential and collaborative learning. As outlined by Majumder and Dey (2024), the Metaverse is increasingly central to Industry 5.0, enabling interconnected human–machine cooperation and simulation-driven innovation. Empirical studies show that Metaverse-based learning environments can enhance student engagement, motivation, and conceptual understanding in higher education (Puneet et al., 2024). Immersive technologies such as Augmented Reality (AR) and Virtual Reality (VR) also provide embodied learning opportunities, allowing students to interact with complex systems in ways that are otherwise difficult or risky in real-world settings (Thangavel, 2025). Despite these advances, many educational Metaverse implementations remain static and lack the adaptive, personalized, and data-driven features necessary to support deep learning and transfer of skills.

AI plays a pivotal role in addressing these limitations by enabling dynamic personalization, automated assessment, and intelligent feedback. AI-driven systems can model learner behavior, detect skill gaps, and adjust learning pathways in real time, thereby facilitating individualized learning experiences (Joshi et al., 2025). AI-powered Metaverse platforms have been proposed to enhance interactivity, automate scenario adaptation, and support large-scale analytics for learning optimization (Soliman et al., 2024). However, these systems rarely integrate the high-resolution data streams generated by DTs, nor do they fully exploit DT–Metaverse interoperability for next-generation learning ecosystems. Additionally, AI deployment in the Metaverse introduces cybersecurity challenges, calling for robust governance and intelligent risk-mitigation mechanisms (Awadallah et al., 2024).

While the Metaverse and DTs hold substantial promise, their combined application in higher education remains limited. Existing studies primarily focus on isolated components—for example, immersive inclusion technologies for learners with specific learning difficulties (Yenduri et al., 2023), NextG communications for industrial Metaverse applications (Prabadevi et al., 2023), and future-oriented learning models such as MOOC 5.0 that emphasize personalization and learner autonomy (Ahmad et al., 2022). At the workforce level, HR 5.0 emphasizes AI-enhanced adaptability, continuous upskilling, and human–machine collaboration (Khan et al., 2025), while urban and industrial Metaverse ecosystems illustrate the potential of real-time twin-based environments for interactive, intelligent operations (Dienhart et al., 2025). These developments highlight the increasing alignment between educational innovation and workforce expectations, yet a coherent educational framework that unifies AI, DTs, and Metaverse technologies to enhance STEM employability remains unexplored.

In particular, there is limited understanding of how AI-enabled Digital Twins within a Metaverse ecosystem can foster adaptive expertise, which is essential for navigating unpredictable, ill-structured, and rapidly changing STEM work environments. Adaptive expertise emphasizes cognitive flexibility, innovation, and the ability to transfer knowledge across contexts—skills considered foundational for the future “platinum workforce” in the industrial transition era (Undheim, 2025). While simulation-rich environments and dynamic feedback have been linked to the development of adaptive expertise, existing systems often rely on linear or non-adaptive learning designs that do not respond to learner performance or workplace complexity (Almusaed et al., 2023).

Thus, significant research gaps remain: (1) the lack of integrated frameworks combining AI, Digital Twins, and the Metaverse for higher education; (2) the absence of models linking these technologies to STEM employability skills and real-world industrial competencies; (3) limited theoretical grounding connecting adaptive expertise with AI-driven DT–Metaverse learning ecosystems; and (4) inadequate attention to the dynamic, data-rich interactions that could personalize and optimize skill development.

To address these gaps, this article pxroposes a novel conceptual framework that integrates AI-based adaptivity, Digital Twins modelling, and Metaverse immersive environments to enhance STEM work skills and foster adaptive expertise in higher education. The framework positions DTs as high-fidelity virtual replicas enabling realistic industrial simulations (Khan & Ahmad, 2025), the Metaverse as an embodied ecosystem for collaborative and experiential learning (Majumder & Dey, 2024), and AI as the cognitive engine driving personalization, analytics, and dynamic scenario generation (Soliman et al., 2024). By synthesizing insights across the interdisciplinary literature, the proposed model contributes to the design of next-generation educational ecosystems capable of preparing learners for the AI-enhanced workforce and the emergent cognitive algorithmic economy (Kliestik et al., 2024).

2. Method

This study adopts a conceptual–analytical research design to develop and validate a comprehensive framework of an Artificial Intelligence–Based Digital Twins Metaverse (AI-DTM) for enhancing STEM work skills and adaptive expertise in higher education. The method follows three structured phases: (1) theoretical synthesis, (2) framework construction, and (3) expert validation.

2.2 Phase 1: Systematic conceptual synthesis

Produce a rigorous, reproducible synthesis of technological and pedagogical literature that will (a) identify core constructs and mechanisms for AI-DTM, (b) map theoretical underpinnings linking technology to learning outcomes (STEM employability & adaptive expertise), and (c) generate an evidence-based codebook to drive framework construction (Debnath & Srivastava, 2025).

2.2.1 Databases and sources

Searches will be performed in multiple bibliographic and grey-literature sources to ensure interdisciplinary coverage:

• a. Scholarly databases: Scopus, Web of Science, IEEE Xplore, PubMed (for health/assistive tech literature), SpringerLink, Emerald, ScienceDirect, and Google Scholar.

• b. Preprint and technical repositories: arXiv, SSRN.

• c. Policy and standards: EU AI Act documents, white papers, and relevant regulatory briefs.

• d. Conference proceedings: IEEE, ACM, InC4, and major metaverse/AI/education events.

2.2.2 Timeframe and language

• a. Timeframe: Publications from 2018 through 2025 (captures Industry 4.0 → 5.0 transitions, recent metaverse and DT work).

• b. Language: English.

2.2.3 Search strategy and example search strings

Develop domain-specific search strings combining keywords and Boolean operators. Example search strings (to be adapted per database syntax):

(“digital twin” OR “digital twins”) AND (education OR learning OR “higher education” OR STEM) (“metaverse” OR “extended reality” OR “virtual reality” OR “augmented reality”) AND (learning OR education OR training) (“artificial intelligence” OR “AI” OR “machine learning” OR “reinforcement learning”) AND (“metaverse” OR “digital twin”) (“adaptive expertise” OR “adaptive learning” OR “transfer of learning”) AND (“simulation” OR “digital twin” OR “metaverse”)

Each search records the database, date run, search string, and total hits.

2.2.4 Inclusion and exclusion criteria

Inclusion:

• a. Empirical studies, reviews, conceptual papers, standards/whitepapers, and high-quality preprints that address one or more of: Digital Twins, Metaverse/XR for learning, AI for adaptive learning, STEM competency development, or adaptive expertise.

• b. Studies discussing technological infrastructures relevant to low-latency, edge/6G, semantic communications, or DT-AI governance.

Exclusion:

• a. Papers not relevant to learning or human competency (purely industrial DTs with no educational implications) unless they include transferable methodology.

• b. Non-English publications.

• c. Short abstracts with insufficient methodological detail.

2.2.5 Screening and selection

• a. Deduplication using reference manager (EndNote/Zotero).

• b. Title–abstract screening by two independent reviewers. Records marked include/uncertain/exclude. Disagreements resolved through discussion; unresolved items adjudicated by a third reviewer.

• c. Full-text screening using same two-reviewer process. Reasons for exclusion recorded.

Quality/consensus metric: Compute Cohen’s kappa for inter-rater agreement at title–abstract and full-text stages; target κ ≥ 0.70.

Figure 1 presented the PRISMA 2020 flow diagram illustrates the structured process applied in Phase 1 to ensure transparent and reproducible literature synthesis. It traces the progression of records from initial database identification through duplicate removal, screening, eligibility assessment, and final inclusion (n = 42). This systematic mapping approach strengthens the evidence base underpinning the conceptual construction of the AI-Enabled Metaverse Digital Twins model by following established PRISMA 2020 reporting standards (Page et al., 2021).

Figure 1. PRISMA 2020 flow diagram for the systematic conceptual synthesis.

2.2.6 Data extraction

For each included study extract standardized fields into a spreadsheet:

• • bibliographic details (authors, year, venue)

• • study type (empirical, conceptual, review, standard)

• • domain/setting (education, industry, healthcare, smart city)

• • technology focus (DT, metaverse/XR, AI, edge/6G)

• • pedagogical approach (PBL, experiential, constructivist, MOOC5.0, etc.)

• • learners/participants (STEM level, sample size)

• • measured outcomes (employability skills, adaptive expertise, transfer)

• • methodological notes (metrics, analytics, model architectures)

• • governance/ethical considerations (privacy, EU AI compliance)

• • key findings and claimed limitations

• • suggested future research/gaps

2.2.7 Codebook development and thematic analysis

• • Open coding: Two researchers independently read a random sample (≈20%) of included papers to generate initial codes (technical affordances, pedagogical mechanisms, data types, learning outcomes).

• • Codebook: Consolidate codes into a structured codebook with definitions, inclusion/exclusion examples, and hierarchical categories (e.g., Technology → DT fidelity → sensors; Pedagogy → PBL → scaffolding).

• • Intercoder reliability: Apply the codebook to another sample and compute Cohen’s kappa for major code categories; revise until κ ≥ 0.70.

• • Full coding: Use qualitative analysis software (e.g., NVivo or Atlas.ti) to code all included texts.

Thematic synthesis will follow Braun & Clarke’s six-step procedure adapted for concept mapping: familiarization, code generation, searching for themes, reviewing themes, defining/naming themes, and producing the synthesis.

2.2.8 Concept mapping and mechanism elaboration

From themes, build evidence-based concept maps that show:

• • Technological affordances → pedagogical mechanisms → learner processes → outcomes.

• • Example: DT fidelity (affordance) → realistic problem complexity (mechanism) → iterative practice + feedback (process) → transfer & adaptive expertise (outcome).

Maps will be created using diagram tools (e.g., Miro, Lucidchart), annotated with supporting citations and confidence levels (high/medium/low evidence).

2.2.9 Synthesis outputs

Deliverables from Phase 1:

Phase 1 generated several synthesis outputs that establish the conceptual basis for the AI-Enabled Metaverse Digital Twins (AI-MDT) framework. A PRISMA-structured search log was produced to ensure transparent reproducibility of the systematic conceptual synthesis. An integrated extraction matrix organized coded evidence on AI-driven personalization (Joshi et al., 2025), Digital Twin simulation fidelity (Khan & Ahmad, 2025), and Metaverse-enabled immersion (Puneet et al., 2024), along with documented relationships to employability skills and adaptive expertise. A structured codebook standardized definitions across technological, pedagogical, and competency constructs to support analytic coherence. The thematic map and linkage model illustrated how AI, DT, and Metaverse affordances connect to cognitive, metacognitive, and experiential learning mechanisms (Majumder & Dey, 2024; Soliman et al., 2024). A gap matrix identified key underserved intersections—most notably the limited research linking Digital Twins to adaptive expertise in higher education. Finally, concise evidence statements summarized empirical support for core propositions in the framework, such as the benefits of AI personalization (Iqbal et al., 2024), DT–simulation-based transfer (Nicoletti, 2023), and immersive metaverse collaboration (Jalhotra et al., 2024).

2.2.10 Validation and sensitivity checks

• • Triangulation: Cross-validate themes against policy/regulatory documents (e.g., EU AI Act) and industry white papers to ensure technological feasibility and governance awareness.

• • Sensitivity check: Re-run searches with alternative keywords and check for missing seminal works.

• • Expert spot-check: Present preliminary maps to 2–3 domain experts for face validity and to identify overlooked literature.

2.2.11 Timeline and resources

Estimated effort: 8–10 weeks for Phase 1 (scoping: 1 week; searching & deduplication: 1–2 weeks; screening: 2 weeks; extraction & coding: 2–3 weeks; mapping & validation: 1–2 weeks). Team: 2 coders + 1 adjudicator + PI.

2.4 Phase 3: Expert review and validation

Figure 2 presents the structured validation workflow implemented in Phase 3 of the study. A panel comprising five domain experts—specialists in artificial intelligence systems, digital twin technologies, metaverse-based education, instructional design, and STEM pedagogy—systematically evaluated the proposed AI-Enabled Metaverse Digital Twins framework. The evaluation encompassed four core dimensions: (a) conceptual clarity and internal coherence of the framework’s constructs, (b) technological feasibility and systems-level implementability, (c) pedagogical appropriateness with respect to instructional design principles, and (d) alignment with prevailing regulatory and ethical standards. The review process followed established validation protocols used in digital health AI frameworks (Kchaou et al., 2025) and metaverse educational models (Waquar et al., 2025). Expert feedback was incorporated through iterative refinement cycles until a consensus-based confirmation of validity was achieved.

Figure 2. Expert review and validation procedure.

3. Result

The development of the AI-Based Digital Twins Metaverse (AI-DTM) framework produced a structured model that integrates advanced AI capabilities, metaverse environments, and digital twin simulation to enhance STEM work skills and adaptive expertise in higher education. The resulting framework demonstrates how intelligent automation, immersive virtual spaces, and real-time model synchronization collectively create an adaptive and competency-driven learning ecosystem.

The Figure 3 depicts the integrated architecture of the proposed AI-DTM framework, showing how AI-driven personalization, Digital Twin real-time simulations, and immersive Metaverse XR environments converge to form a unified and adaptive learning ecosystem. Through synchronized data flows and multimodal interactions, the system supports high-fidelity experiential learning, cognitive–metacognitive development, and dynamic competency progression. These mechanisms ultimately enhance STEM work skills, promote adaptive expertise, and prepare learners for Industry 5.0 workforce demands.

Figure 3. The AI-Based Digital Twins Metaverse (AI-DTM) conceptual framework.

The development and synthesis phases of the AI-Based Digital Twins Metaverse (AI-DTM) framework yielded several interconnected findings that illustrate how the integration of artificial intelligence, digital twin systems, and metaverse-based immersive environments can collectively enhance STEM-focused learning. The results highlight not only the functional interplay between these technological components but also their combined impact on personalized learning, experiential simulation, cognitive development, and future-oriented workforce readiness. The key outcomes emerging from the framework are summarized as follows (Thangavel, 2025).

First, the results confirm that AI serves as the primary enabler of personalized learning pathways, dynamic feedback loops, and predictive performance analytics—an alignment consistent with recent reviews highlighting AI’s transformative role in education and professional development (Iqbal et al., 2024; Al-Khatib et al., 2024). The AI-DTM design incorporates machine learning–based learner modeling, automated assessment pipelines, and intelligent scaffolding, ensuring that learners receive continuous, context-aware guidance as they engage in authentic STEM tasks.

Second, digital twin–based simulations proved essential in bridging conceptual STEM knowledge with real-world work processes. Digital twins enabled the representation of complex systems and workflow behaviors, echoing the broader utilization of digital twins in industrial, engineering, and geotechnical domains (Nicoletti, 2023; Akbas, 2025; Rakholia et al., 2024). Within the metaverse, these twins were rendered as interactive, high-fidelity virtual objects that replicated real-time operational states, allowing learners to perform diagnostics, manipulate variables, and observe system responses without real-world risk.

Third, the metaverse layer provided an immersive and collaborative environment that supported experiential learning, multisensory engagement, and scenario-based problem solving. This finding aligns with studies demonstrating the metaverse’s potential for future education ecosystems and human–machine collaborative work (Jalhotra et al., 2024; Raja Santhi & Muthuswamy, 2023). The metaverse-enabled virtual workspaces allowed for synchronous and asynchronous collaboration, mimicking modern Industry 4.0 and Industry 5.0 workplace dynamics (Dai et al., 2024; Hassoun et al., 2025).

Fourth, the integration of AI, digital twins, and metaverse technologies collectively supported the development of adaptive expertise through iterative exposure to increasingly complex scenarios. The framework operationalized features such as adaptive task difficulty, personalized feedforward, and predictive modeling of learner progress, consistent with the emerging literature on AI-enabled innovation ecosystems (Khan et al., 2024; Patil, 2024). These features positioned learners not only to master core STEM competencies but also to demonstrate flexibility, transferability, and problem-solving capability in novel situations.

Overall, the AI-DTM framework highlights a scalable and future-ready architecture that mirrors advancements in Industry 5.0 and AI-driven digital transformation across sectors. The results indicate that AI-DTM can strengthen workforce-aligned competencies while promoting human–technology synergy, supporting the broader transition toward intelligent, sustainable, and human-centric learning systems.

Table 1 synthesizes the multidimensional results of the AI-DTM model by showing how its five core components contribute to integrated improvements in STEM work skills and adaptive expertise. The AI Cognitive Engine strengthens personalized learning and higher-order cognition through generative AI, predictive modelling, and automated analytics, reflecting current findings on AI-driven learning optimisation (Iqbal et al., 2024; Al-Khatib et al., 2024). The Digital Twin Core enhances procedural and technical accuracy by replicating real-world STEM environments with high fidelity, a key requirement for Industry 5.0 readiness (Nicoletti, 2023; Dai et al., 2024). The XR–Metaverse Interaction Hub improves experiential and collaborative competence by enabling immersive 3D simulations that promote embodied and interactive learning (Jalhotra et al., 2024; Hassoun et al., 2025). When combined, these layers create cross-model synergy that supports the development of adaptive expertise, aligning with research on AI-enhanced workforce adaptability and digital transformation (Patil, 2024; Akbas, 2025). The final layer links these outcomes to STEM workforce preparedness, consistent with Industry 5.0 expectations for advanced human–machine collaboration (Rakholia et al., 2024; Raja Santhi & Muthuswamy, 2023).

Table 2 provides a consolidated overview of how each functional dimension of the AI-Based Digital Twins Metaverse (AI-DTM) contributes to the development of core employability capabilities in STEM education. The table highlights the progressive alignment between AI-driven personalization, high-fidelity digital twin simulations, and immersive metaverse interaction, showing how these components collaboratively enhance technical competence, metacognitive awareness, and adaptive problem-solving (Thangavel, 2025). At the technological level, AI-supported analytics and automated decision engines enable intelligent adaptation and continuous learner modelling, consistent with current developments in AI-enhanced educational practices (Iqbal et al., 2024; Khan et al., 2024). The integration of digital twins provides real-time procedural feedback and operational accuracy through realistic, data-driven virtual replicas, aligning with digital transformation trends in Industry 5.0 and 6.0 (Nicoletti, 2023; Rakholia et al., 2024; Dai et al., 2024). Meanwhile, the metaverse layer strengthens experiential and collaborative learning by facilitating immersive 3D tasks and multisensory simulations, as emphasized in contemporary metaverse-based pedagogical research (Jalhotra et al., 2024; Hassoun et al., 2025). Together, these findings illustrate how AI-DTM produces a unified ecosystem capable of supporting workforce readiness and future-seeking learner adaptability in an increasingly automated and interconnected world.

Figure 4 illustrates the integrated architecture of the proposed AI-Enabled Metaverse Digital Twins (AI-MDT) framework. This diagram visualizes how various educational use cases—such as adaptive STEM training, XR-based practical labs, competency-based assessment, and Industry 5.0 scenario practice—are operationalized through synchronized AI, Digital Twin, and Metaverse technologies. At the technological core, AI learner modeling, real-time digital twin simulations, immersive XR laboratories, and autonomous feedback engines interact seamlessly to generate adaptive and data-driven learning processes. The middle layer demonstrates the progressive learning experience, moving from simulation and adaptation to immersive engagement and collaborative problem-solving. These stages reflect the development pathway of employability skills and adaptive expertise as outlined in the article. At the bottom, the framework highlights key innovations, including AI-driven personalization, real-time DT synchronization, and AI-Metaverse cognitive systems, which collectively enable autonomous learning feedback loops. Overall, the figure provides a visual summary of how AI-MDT environments orchestrate simulation fidelity, adaptive learning mechanisms, and collaborative immersion to enhance STEM competencies and workplace readiness.

Figure 4. Integrated AI–Digital Twins–Metaverse (AI-DTM) architecture for enhancing STEM skills and adaptive expertise.

4. Discussion

The findings of this study demonstrate that the AI-driven Digital Twin–Metaverse framework offers a significant methodological and conceptual shift in the design of competency development models. While previous research has emphasized the importance of immersive technology for learning engagement (Radianti et al., 2020) and simulation fidelity (Johansen et al., 2023), this study extends the discourse by showing how AI-mediated personalization and dynamic feedback loops within digital twins can systematically enhance the accuracy of learner profiling and competency tracking over time. This represents an evolution from static VR/AR environments toward adaptive, data-driven ecosystems.

The framework also challenges the traditional assumption that digital learning environments function merely as content-delivery systems. Instead, results indicate that the system acts as a continuous optimization mechanism, where AI agents adjust scenarios, difficulty levels, and behavioral feedback based on real-time performance. Such findings align with and advance current theoretical perspectives on learning analytics (Ifenthaler & Yau, 2020) by embedding analytics directly into the learning environment rather than treating them as external evaluation tools.

A critical insight lies in the model’s emphasis on situated, competency-based assessment. Unlike conventional self-report instruments, the integrated digital twin enables behavioral capture through multimodal indicators movement patterns, decision-making trajectories, response consistency which prior studies often treat as separate datasets (Lee & Hwang, 2022). The fused analytics in this study support a more holistic understanding of learner performance, providing stronger ecological validity. This contributes to contemporary debates on the need for robust, real-time assessment tools in vocational and professional training contexts (Lee & Park, 2023).

Moreover, the model reveals strong potential for scalability. While earlier metaverse applications face challenges of fragmentation and context mismatch (Shin, 2022), the modular digital-twin design in this study enables flexible integration with various training domains. This points toward a new generation of interoperable learning ecosystems, positioning the proposed framework as a conceptual bridge between VR-based training and intelligent competency management systems.

Explicit novelty and contribution

5. Conclusion

This study proposes a pioneering AI-Based Digital Twins Metaverse (AI-DTM) framework that advances the current landscape of intelligent learning environments by integrating AI-driven behavioral analytics, dynamic digital twin modeling, and immersive metaverse simulations into a unified system. The novelty of this framework lies in its ability to generate real-time , multimodal learner representations, deliver adaptive feedback loops, and orchestrate human–AI collaborative learning experiences that closely approximate real-world STEM workplaces—capabilities that remain underdeveloped in existing VR, LMS, and simulation-based platforms.

Figure 5 illustrates the conceptual convergence of Artificial Intelligence, Digital Twins, and the Metaverse as synergistic technologies that drive innovation in STEM higher education. The framework contributes three major theoretical advancements: (1) it expands Industry 5.0 principles into higher education through a human-centric, resilience-oriented learning architecture; (2) it integrates AI-driven competence analytics with metaverse-based experiential environments, bridging a long-standing gap between assessment precision and ecological authenticity; (3) it offers a new conceptual model explaining how adaptive expertise and employability skills emerge through continuous interaction between agents, algorithms, and virtualized work systems.

Figure 5. Integration of AI, digital twins, and metaverse for STEM innovation in higher education.

From a practical perspective, the AI-DTM model provides a scalable blueprint for institutions transitioning toward AI-enabled learning ecosystems. The architecture supports personalized skill development, complex scenario training, and industry-aligned experiential learning, addressing urgent global demands for future-ready STEM talent (Iqbal et al., 2024; Nicoletti, 2023). By aligning with emerging technological transformations in Industry 5.0, digital sustainability, and intelligent automation (Dai et al., 2024; Akbas, 2025; Rakholia et al., 2024), the framework delivers actionable pathways for enhancing workforce preparedness while maintaining human-centered values.

Overall, the AI-DTM framework strengthens academic discourse by offering a conceptually robust, empirically informed, and future-oriented model for next-generation intelligent education. It lays the groundwork for the development of sustainable, adaptive, and high-fidelity learning environments that can shape the future of STEM higher education and support global transitions toward advanced digital societies.

Institutional review board statement

Not applicable. This study did not involve human participants, clinical procedures, or any activities requiring ethical approval under institutional or national guidelines.

Informed consent statement

Not applicable. No human participants were recruited, surveyed, or recorded in this conceptual and framework-development study.