Factors influencing the use of VR technology for safety training among electric industry workers in China: an extended TAM

Abstract

Despite the potential of virtual reality (VR)-based safety training to reduce fatalities in the electric industry, its success depends on worker acceptance. This study validated an extended technology acceptance model incorporating self-efficacy (SE), facilitating conditions (FC), and technology anxiety (TA) using data from 341 electric industry workers. Results confirmed that while SE and FC are significant positive predictors of perceived usefulness (PU) and perceived ease of use (PEOU) (p < 0.05), TA exerts a significant negative influence on PEOU and PU (path coefficients of − 0.194 and − 0.217, respectively). Structural equation modeling indicated that the research model explained a substantial proportion of variance: 38.1% for PEOU, 61.5% for PU, 65.6% for attitude, and 44.7% for intention to use. These findings provided an evidence-based framework for stakeholders to bolster user confidence and mitigate technological anxiety, thereby accelerating the strategic adoption of VR safety training in high-risk electrical environments.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-49473-4.

Introduction

The global energy landscape is undergoing a profound transformation, driven by the rapid modernization and expansion of the electric grid to integrate renewable energy sources^1,2. This evolution from centralized models to decentralized and digitally-controlled networks creates increasingly sophisticated and intricate infrastructure, which in turn introduces and exacerbates significant occupational hazards for the electrical workforce³. Personnel routinely face life-threatening risks such as electric shock, arc flash, and falls⁴. Existing safety protocols are struggling to keep pace with the escalating operational demands of next-generation grid development⁵.

As a regulatory mandate and a fundamental pillar of workforce safety, traditional training aims to equip electrical workers with the critical knowledge, practical skills, risk perception, and situational awareness necessary to mitigate operational hazards^6–9. The conventional approaches for safety training in the electric industry typically rely on classroom lectures, printed manuals, and instructor-led demonstrations¹⁰. The passive nature of the conventional approaches often fosters disengagement, provides insufficient hands-on experience, and fails to adequately simulate the high-risk and high-pressure scenarios where critical decision-making is paramount¹¹. Consequently, the conventional approaches struggle with low knowledge retention rates because electric industry workers lack opportunities for repeated practice in realistic settings¹². When compounded by the significant logistical and financial burdens of organizing in-person sessions, these deficiencies underscore the urgent need for innovative solutions, positioning virtual reality (VR) as a promising technology to address the shortcomings of traditional safety programs.

VR is a technology that generates a fully immersive, computer-generated, and three-dimensional environment¹³, replacing the user’s real-world surroundings to instill a compelling sense of presence and interaction in a digital world¹³. This unique capability allows for the creation of interactive training platforms where workers can safely engage with simulated hazards and practice complex procedures in real time¹⁰. VR technology facilitates adaptive, repeatable, and scalable training experiences that enhance knowledge retention and prepare personnel for rare but critical events without exposure to actual danger¹⁴.

In the relevant literature, there are limited studies on VR-based safety training acceptance among workers in other high-risk industries. For instance, Guo, et al.¹⁵ applied an extended unified theory of acceptance and use of technology to identify the drivers of construction workers’ willingness to adopt metaverse-based safety training. They found that performance and effort expectancy, social influence, and facilitating conditions (FC) directly boost the acceptance, while perceived trust moderates the transition from willingness to actual participation. Similarly, Liu, et al.¹⁶ integrated the technology acceptance model (TAM) and the theory of planned behavior supplemented by risk perception, safety climate, and self-efficacy (SE) to analyze Chinese construction workers’ acceptance of VR-based safety training. However, no studies focused on the VR-based safety training acceptance among the workforce in the electric industry, which remains a significant and unexamined research gap. While recent studies have examined VR acceptance in high-risk sectors such as construction^15,16, this study extended those patterns by identifying the unique psychological drivers specific to the electric power industry, where the margin for error is significantly narrower. Different from construction-based models that frequently highlight social influence and general performance expectancy, this study demonstrated that for electric industry workers, SE and technology anxiety (TA) are the dominant predictors of their adoption of VR-based safety training. The novelty of this study lied in its application of the extended TAM to the high-consequence and zero-tolerance environment of the electric industry¹⁷, where the drivers of technology acceptance differ fundamentally from general corporate or office settings. In this specific professional context, PU is not merely a measure of task efficiency but a critical evaluation of life safety and hazard mitigation. This study thus provided a specialized framework for understanding workers’ acceptance in the electric industry.

This study aimed to provide an in-depth understanding of the acceptance of VR-based safety training among electric industry workers. To achieve this, this study introduced and validated an expanded TAM that integrated the critical constructs of self-efficacy (SE), FC, and TA. This research constituted a pioneering effort to examine the specific influence of these factors on the adoption of VR technology within this high-risk occupational group. The practical insights generated help VR developers, policymakers, and practitioners engineer superior, tailored VR-based safety training programs that effectively meet the needs of the electrical workforce. While general technology acceptance is well documented, the high-stakes and high-risk nature of the electric industry creates unique psychological barriers, such as AT about safety-critical errors, that standard models do not fully capture. By integrating SE and TA into a domain-specific framework for Chinese electric workers, this study provides a nuanced understanding of safety-critical acceptance, which differs fundamentally from general office-based technology adoption. The novelty of this research lies in developing a domain-specific acceptance framework that accounts for the psychological interplay between risk-related anxiety and SE in high-voltage operational environments. A comprehensive list of all abbreviations is summarized in Appendix 1.

Literature review and research hypotheses

Technology acceptance model

TAM, introduced by Davis, et al.¹⁸, provides a foundational framework for evaluating user acceptance of technology. TAM posits that an individual’s intention to use technology (ITUT) is primarily determined by their perceived usefulness (PU) and attitude toward using technology (ATUT). In turn, ATUT is shaped by PU and perceived ease of use (PEOU), while PU is influenced by PEOU and external factors, with PEOU itself being affected by external variables. Within this study’s context, PEOU refers to the degree to which electric industry workers find VR-based safety training effortless. PU represents their belief that VR-based safety training enhances personal safety performance. ATUT captures their positive or negative impressions of VR-based safety training. ITUT denotes their willingness to adopt VR-based safety training. Because TAM is a preeminent theory widely validated for explaining worker acceptance of various new technologies, including cobots¹⁹ and digital procedures²⁰, it offers a robust theoretical basis for this investigation.

Self-efficacy

SE, defined as an individual’s conviction in their own capacity to execute the behaviors necessary to produce specific performance outcomes²¹, is recognized as a pivotal external variable within the TAM framework²². The influence of SE on technology acceptance has been extensively validated across diverse research domains, establishing it as a primary determinant in users’ willingness to embrace innovations such as digital business²³, generative artificial intelligence tools²⁴, and e-learning systems²⁵. Specifically within the context of immersive technologies, empirical evidence confirms that higher levels of SE positively influence PEOU and PU of VR technology^26,27 because users who feel more capable are more likely to perceive the system as easy to use and beneficial.

Facilitating condition

FC is defined as the degree to which individuals believe the requisite resources exist to support their use of a new technology²⁸. The positive influence of FC on user acceptance is well-established in the literature. Empirical studies consistently demonstrate that it significantly enhances PU and PEOU for diverse technologies, including highly automated vehicles²⁹, online mapping services for older populations³⁰, and smart wearable devices³¹. In the context of this study, FC could be crucial in helping electric industry workers manage the technical demands and fully leverage the learning capabilities of VR for safety training.

Technology anxiety

TA, defined as an individual’s concern about their ability to correctly use new technological devices³², is a significant psychological barrier to adoption. TA frequently stems from a lack of familiarity or the perceived complexity of navigating innovative systems³³, often manifesting as avoidance behaviors and fostering resistance to the technology³². In the context of VR-based safety training, workers who fear making errors are less likely to engage with and employ the technology, undermining its training potential³⁴. Therefore, TA among electric industry workers is expected to negatively impact PEOU by making the system seem more intimidating, and PU because the anxiety itself may overshadow the training’s benefits³⁰. Consequently, the following hypotheses were put forth:

Research model

Synthesizing the preceding literature review, this study proposed an integrated research model to investigate the determinants of VR-based safety training acceptance among electric industry workers. This conceptual framework was theoretically grounded in the TAM but was expanded to incorporate three critical external variables identified as highly relevant to this specific context: SE, FC, and TA. The complete proposed model, which schematically articulates the 11 hypotheses developed from this synthesis, is presented in Fig. 1. The overarching objective of this framework is to comprehensively elucidate the factors influencing this specialized workforce’s willingness to adopt VR-based safety technology.

Fig. 1.

VR-based safety training acceptance model for electric industry workers.

H1: PEOU positively influences ATUT.

H2: PEOU positively influences PU.

H3: PU positively influences ATUT.

H4: PU positively influences ITUT.

H5: ATUT positively influences ITUT.

H6: SE positively influences PEOU.

H7: SE positively influences PU.

H8: FC positively influences PEOU.

H9: FC positively influences PU.

H10: TA negatively influences PEOU.

H11: TA negatively influences PU.

Methods

To quantitatively evaluate the hypotheses, this study employed a face-to-face questionnaire-based methodology. A critical component of the procedure involved a pre-survey familiarization phase. Before completing the questionnaire, all participants were briefed on and subsequently underwent a VR-based safety training experience that detailed specific safety procedures. The VR-based safety training consisted of a 25-minute immersive session (plus a 5-minute calibration and tutorial) utilizing a discovery-based “High-Voltage Switchgear Operations and Emergency Response” scenario. This high-risk/low-frequency scenario was selected to safely replicate sensory-heavy arc flash and electrocution hazards that are impossible to simulate in traditional classrooms. Regarding the level of interaction and immersion, the program employed a six-degree-of-freedom system within a 3 m x 3 m tracked space, allowing participants to physically navigate a virtual substation and use hand controllers for tasks such as operating disconnect switches and using multimeters. The interaction design featured a transition from routine maintenance to an active fault sequence, requiring a time-pressured Lockout/Tagout procedure. To ensure pedagogical depth, the program employed active feedback loops, where incorrect tool selection or a breach of safety boundaries triggered visual warnings and a simulated near-miss restart, providing immediate reinforcement of safety protocols. The primary learning objectives of the VR-based safety training were to enhance electric industry workers’ hazard recognition skills related to arc flash and electrocution risks and to achieve mastery of standardized safety protocols within a risk-free environment. This hands-on immersion was designed to ensure their responses were based on actual interaction, thereby enhancing the validity of the study results. The VR experience was administered using the Oculus Quest 2, a wireless, standalone headset selected for its low latency, which mitigates motion sickness and has contributed to its frequent use in worker safety training studies³⁵. The subsequent sections detail the specific questionnaire measurements, participant demographics, and the data analysis procedures employed.

Measurements

To gather empirical data, a self-administered questionnaire was developed, comprising three distinct sections. The first section was designed to gauge participants’ perceptions of the external variables: SE, FC, and TA. The second section evaluated the core TAM constructs, including PU, PEOU, ATUT, and ITUT. The final section collected relevant demographic information, such as age, gender, and educational level. All measurement items for the theoretical constructs were derived and adapted from established, validated scales within prior TAM-related literature, with specific modifications to ensure their relevance to the unique context of VR-based safety training. The 22 items, which were detailed with their corresponding citations in Appendix 2, were assessed using a five-point Likert-type scale ranging from ‘strongly disagree’ (1) to ‘strongly agree’ (5). To ensure high questionnaire comprehension among participants with diverse educational backgrounds, the instrument underwent a rigorous two-stage validation process prior to full-scale distribution. First, a pilot study was conducted with a small focus group of field technicians to identify and simplify overly academic or abstract terminology, translating TAM constructs into industry-specific language. Second, data collection was conducted via facilitated face-to-face sessions, where trained researchers were present to provide real-time clarifications and ensure consistent interpretation of the Likert-scale items without leading the respondents. Furthermore, reverse-coded items and visual anchors were utilized to verify internal consistency and minimize response bias.

Participant

This study, which received ethical approval from the research ethics committee of the South China University of Technology, recruited 341 electric industry workers in China using convenience sampling. Regarding the sampling strategy, this study employed a convenience sampling method primarily due to the specialized and restricted nature of the electric industry workforce, which presented notable barriers to random probability sampling given the stringent site safety and security protocols. The participants were recruited from established regional training centers where a diverse cross-section of field technicians and engineers were already gathered for professional development. The high-voltage operational environment and strict grid security regulations in China preclude the use of broad random sampling techniques because access to workers is limited to authorized training or maintenance windows. Given the high-security nature of the electric industry and the logistical challenges of accessing frontline workers during active duty, sampling via centralized professional training centers provided the most reliable way to reach a concentrated, relevant, and diverse population of electric industry workers. In addition, regional training centers serve as a mandatory nexus for the workforce, providing a participant pool that includes various age groups, experience levels, and technical specialties, thereby mitigating the typical selection bias associated with convenience sampling. This approach ensured that the sample remained highly representative of the target demographic despite the non-random selection. All methods were performed in accordance with the ethical standards of the Declaration of Helsinki. Participants were not required to have prior VR experience. However, to ensure safety and data validity, electric industry workers with poor physical health or cognitive/physical impairments were excluded. Before commencing the survey, all individuals provided informed consent, were assured of their right to withdraw at any time and were assured of the confidentiality of their data. The questionnaire was administered in Chinese, and to ensure data completeness and mitigate response bias, researchers with survey expertise reviewed each instrument upon completion, prompting participants to fill in any missing items. As detailed in Table 1, the sample had a mean age of 44.0, was predominantly male (90.6%), and largely possessed an informal or primary school education (74.5%).

Table 1.

Demographic information of participants (n = 341).

Factors	Description	Frequency	Percentage (%)
Age (years)	≤ 30	32	9.4
	31 − 40	47	13.8
	41 − 50	198	58.0
	51 − 60	64	18.8
Gender	Male	309	90.6
	Female	32	9.4
Education level	Informal education	175	51.3
	Primary school	79	23.2
	Secondary school	60	17.6
	Undergraduate	27	7.9

Data analysis

This study employed a two-stage analytical approach, beginning with confirmatory factor analysis (CFA) to rigorously examine the psychometric qualities of the measurement model. The adequacy of the model fit was evaluated against several key indices³⁶: the ratio of the Chi-square value to the degrees of freedom (χ²/df), the comparative fit index (CFI), the standardized root mean square residual (SRMR), and the non-normed fit index (NNFI). In accordance with established psychometric standards³⁷, an excellent fit is characterized by a χ²/df value below 5.0, CFI and NNFI values above 0.90, and RMSEA values below 0.08. Following the CFA, the model’s psychometric properties were further scrutinized by assessing internal consistency, convergent validity, and discriminant validity³⁸. Internal consistency was evaluated using Cronbach’s alpha, with values greater than 0.70 considered acceptable³⁹. Convergent validity, which measures the agreement among items within the same construct, was affirmed based on the criteria of Fornell and Larcker⁴⁰. All item factor.

loadings and composite reliability (CR) values were required to surpass the 0.70 threshold, and the average variance extracted (AVE) for each construct was required to be above 0.50. Discriminant validity, confirming that the constructs are empirically distinct, was established by verifying that the square root of each construct’s AVE was higher than its correlations with all other constructs. Upon establishing acceptable model fit, reliability, and validity for the measurement model, the proposed research hypotheses were tested by specifying a structural equation model (SEM). The adequacy of this structural model was evaluated using the same stringent goodness-of-fit criteria as the CFA (namely, χ²/df < 3.0, CFI > 0.90, SRMR < 0.08, and RMSEA < 0.08). All statistical analyses for both the CFA and SEM were conducted using the Analysis of Moment Structures (version 26.0)⁴¹. Harman’s single-factor test was conducted to detect common method bias by loading all measurement items into an exploratory factor analysis.

Result

Measurement model assessment

The CFA results indicated an excellent fit for the measurement model. All goodness-of-fit indices met the recommended benchmarks: χ²/df was 2.651 (< 5), CFI was 0.952 (> 0.90), NNFI was 0.941 (> 0.90), and RMSEA was 0.050 (< 0.08). These values collectively demonstrated that the measurement model aligns well with the collected empirical data. Furthermore, the model exhibited strong convergent validity. As detailed in Table 2, all item factor loadings and construct CR values exceeded the 0.70 threshold. Additionally, all AVE values ranged between 0.669 and 0.816, comfortably surpassing the 0.50 standard, confirming that the items accurately represent their intended latent constructs. Item FC1 (“It is important to have someone assist me when using VR technology for safety training.”) captures the perceived necessity of support, which differs slightly from the actual availability of support typically measured in office-based TAM studies. However, in the high-risk context of the electric industry, workers often perceive technology not as a standalone tool but as part of an institutional training system. This item was designed based on the framework by Venkatesh, et al.⁴², where the “importance of assistance” reflects workers’ recognition of the organizational support infrastructure required for safe operation. This item was not reverse-coded because it was intended to measure the positive valuation of available help. A factor loading check was performed and found that FC1 loaded significantly (> 0.70) onto the FC construct, suggesting it shares sufficient variance with the other items. While item FC1 deviates from standard resource-availability metrics, its inclusion is justified by the specialized nature of electrical safety training. For frontline workers, the facilitating condition is often the presence of a qualified instructor or safety officer. Therefore, the perceived importance of this assistance serves as a proxy for workers’ reliance on the organizational support system. While FC1 may partially reflect user confidence, a trait closely linked to SE, the high factor loadings and internal consistency (Cronbach’s alpha > 0.8) of the FC construct in this study confirmed that these items collectively represent the necessary external support environment for VR adoption. Discriminant validity was also firmly established, as shown in Table 3, where the square root of each construct’s AVE (in bold) was higher than any of its inter-construct bivariate correlations. This result confirmed that all constructs in the model are empirically distinct from one another. Finally, the internal consistency of the measures was high, with Cronbach’s alpha values for all constructs exceeding the 0.70 benchmark. The results of Harman’s single-factor test revealed that the first factor accounted for only 27.8% of the total variance (well below the 50% threshold), suggesting that no single factor explains the majority of the variance⁴³. Having demonstrated adequate reliability and validity, the measurement model was deemed robust and suitable for the subsequent SEM analysis.

Table 2.

CFA results.

Construct	Item	Mean	SD	Factor Loading	CR	AVE	Cronbach’s alpha
PEOU	PEOU1	3.84	0.961	0.858	0.901	0.752	0.900
	PEOU2	3.80	0.914	0.910
	PEOU3	3.77	0.936	0.833
PU	PU1	3.69	0.984	0.842	0.894	0.738	0.893
	PU2	3.71	0.965	0.865
	PU3	3.89	0.925	0.870
ATUT	ATUT1	3.75	0.858	0.892	0.878	0.707	0.869
	ATUT2	3.65	0.829	0.856
	ATUT3	3.69	0.932	0.770
ITUT	ITUT1	3.87	0.945	0.851	0.905	0.762	0.904
	ITUT2	3.67	0.938	0.888
	ITUT3	3.50	1.031	0.879
SE	SE1	3.54	0.909	0.811	0.941	0.800	0.941
	SE2	3.54	0.928	0.863
	SE3	3.55	0.980	0.949
	SE4	3.50	0.916	0.948
FC	FC1	3.75	0.911	0.762	0.858	0.669	0.853
	FC2	3.80	0.926	0.883
	FC3	3.69	0.886	0.804
TA	TA1	2.32	1.068	0.882	0.930	0.816	0.930
	TA2	2.29	1.107	0.911
	TA3	2.33	1.124	0.917

Table 3.

Correlations among constructs.

	PEOU	PU	ATUT	ITUT	SE	FC	TA
PEOU	0.867
PU	0.650***	0.859
ATUT	0.707***	0.714***	0.859
ITUT	0.546***	0.586***	0.629***	0.873
SE	0.564***	0.642***	0.775***	0.599***	0.894
FC	0.414***	0.499***	0.521***	0.420***	0.519***	0.818
TA	−0.482***	−0.581***	−0.622***	−0.577***	−0.604***	−0.476***	0.903

Note: Bold values are the square root of AVE values. ***: p < 0.001.

Structural model assessment.

The structural model’s adequacy was assessed using the same goodness-of-fit indices as the measurement model (χ²/df, CFI, NNFI, and RMSEA). The analysis yielded strong fit statistics (χ²/df = 3.109, CFI = 0.936, NNFI = 0.925, and RMSEA = 0.079), all of which met the recommended benchmarks. This outcome confirmed that the proposed conceptual model adequately represents the hypothesized relationships within the data. Subsequent hypothesis testing, detailed in Table 4, revealed that the data supported all 11 proposed hypotheses. Specifically, the core TAM relationships were validated. PEOU significantly and positively influences ATUT (H1) and PU (H2). PU, in turn, positively influences ATUT (H3) and ITUT (H4). ATUT is a significant positive determinant of ITUT (H5). Furthermore, the external variables behaved as anticipated. SE and FC exert positive influences on PEOU and PU (H6–H9), whereas TA negatively influences PEOU and PU (H10 and H11). The research model explained 38.1%, 61.5%, 65.6%, and 44.7% of the variance in PEOU, PU, ATT, and ITUT, respectively.

Table 4.

SEM results.

Hypotheses	Standardized Path coefficient	p-value
H1: PEOU positively influences ATUT.	0.403	< 0.001
H2: PEOU positively influences PU.	0.326	< 0.001
H3: PU positively influences ATUT.	0.485	< 0.001
H4: PU positively influences ITUT.	0.344	< 0.001
H5: ATUT positively influences ITUT.	0.370	< 0.001
H6: SE positively influences PEOU.	0.398	< 0.001
H7: SE positively influences PU.	0.293	< 0.001
H8: FC positively influences PEOU.	0.127	0.041
H9: FC positively influences PU.	0.130	0.015
H10: TA negatively influences PEOU.	−0.194	0.003
H11: TA negatively influences PU.	−0.217	< 0.001

Discussion

By integrating the critical external variables of SE, FC, and TA, this research proposed and validated an expanded TAM to comprehensively explain the adoption of VR-based safety training among electric industry workers. The model’s central proposition that acceptance within this specialized workforce is significantly dependent on psychosocial and situational factors was empirically substantiated using data from 341 participants. This study made theoretical contributions by validating the expanded TAM in a high-risk occupational setting and providing practical implications for enhancing electric industry workers’ adoption of VR-based safety training.

Theoretical implications

This study provided substantial empirical evidence validating the TAM as a robust framework for explaining the adoption of VR-based safety training among electric industry workers. In alignment with the model’s core theoretical propositions⁴⁴, the findings confirmed the critical causal relationships. Critically, this research extended beyond mere replication, significantly enhancing TAM’s explanatory power by successfully integrating the external constructs of SE, FC, and TA.

This study’s finding that SE exerts a significant positive influence on PU and PEOU was consistent with prior research on the adoption of new educational technologies^45,46. As a primary psychological driver, SE, an individual’s belief in their own capacity to execute a specific task²¹, profoundly shapes user perception. Electric industry workers with high SE approach VR technology with confidence rather than trepidation, trusting their innate ability to learn, adapt, and successfully operate the VR-based safety training system. This confidence directly mitigates perceived complexity and enhances PEOU⁴⁷. The positive relationship between SE and PEOU suggests a cognitive cushion effect. In the high-stress context of electrical safety, workers with high SE likely allocate fewer cognitive resources to the fear of failure, allowing for more efficient processing of the VR interface itself⁴⁸. Furthermore, this confidence in using VR technology translates directly into a belief in learning from it. Electric industry workers with high SE are likely to engage fully with the simulation, master the safety procedures, and believe they can successfully transfer those skills to the real world. This conviction in the successful acquisition of skills and improvement in performance enhances the system’s PU, framing it as an effective tool for professional development rather than an insurmountable challenge.

A significant finding of this investigation was the positive influence of FC on PU and PEOU. This result aligned with prior technology acceptance research, which similarly identified the pivotal role of FC in determining PU and PEOU in contexts such as e-learning^49,50. The impact of FC on PU indicated that institutional support is not merely a logistical requirement but a psychological validator. When a power company provides robust technical support, it signals to the worker that the technology is a legitimate and reliable safety tool, thereby reducing the trust gap often found in traditional industrial workforces transitioning to digital platforms. Within this high-risk industrial setting, FC may function as a foundational requirement rather than a primary motivator. The modest coefficients suggest that while a baseline level of support is required to reduce operational complexity and ensure the system functions smoothly, the presence of such support alone does not drastically shift a worker’s perception of the technology’s ease or usefulness. Support infrastructures may reduce the perceived operational complexity of the VR system by making electric industry workers feel adequately supported⁵¹, thereby boosting its PEOU. Moreover, these conditions ensure VR technology is reliable and seamlessly integrated into the safety training regimen. When the VR-based safety training functions smoothly and workers can readily access help, their cognitive focus can shift from troubleshooting the technology to mastering the safety procedures⁵². This focus, in turn, reinforces their belief that the VR-based safety training system is a valuable and effective tool for improving their occupational safety, consequently enhancing its PU.

This study’s finding that TA negatively impacts PU and PEOU is consistent with prior technology acceptance research in other populations, such as older adults⁵³. While standard TAM research often found PU to be the primary driver of intention, the results of this study highlighted that in safety-critical training, TA acts as a powerful gatekeeper. For an electric industry workers, a technical error in VR is not just a nuisance. It may be psychologically linked to real-world fatal mistakes, which explains why TA had such a significant negative path coefficient. This anxiety serves as a significant psychological barrier to VR-based safety training among electric industry workers. When electric industry workers feel apprehensive about using VR technology, their perception of the VR-based safety training system’s difficulty immediately increases. This difficulty manifests as a low intrinsic belief in their ability to operate the technology⁵⁴, causing them to perceive it as more complex and intimidating than it objectively is, thereby depressing PEOU. Furthermore, this cognitive apprehension diverts the user’s attentional focus from the pedagogical benefits of VR-based safety training to the procedural stress of the interaction itself. Instead of internalizing safety procedures, the anxious worker becomes preoccupied with fears of making errors or breaking the system, a distraction that prevents them from recognizing VR technology’s value as an effective learning tool and thus diminishes its PU⁵⁵. Therefore, these results strongly indicated that a great emphasis must be placed on mitigating TA to ensure the successful promotion and adoption of VR-based safety training within this workforce.

Practical implications

The findings can provide actionable guidance to developers, policymakers, and industry associations seeking to accelerate the adoption of VR-based safety training in the electric industry. To leverage the positive impact of PU and PEOU on electric industry worker attitudes, promotional activities must explicitly highlight the tangible benefits of VR-based safety training, such as improved hazard recognition and mastery of high-risk procedures. Moreover, VR-based safety training providers must offer structured, hands-on experiences to demystify the technology. Critically, implementation strategies must simultaneously address SE, FC, and TA. To bolster SE, organizations must cultivate worker confidence through low-stakes introductory sessions before advancing to complex simulations. To enhance FC, a robust support infrastructure is non-negotiable, requiring well-maintained equipment, clear guides, and immediate, on-demand technical support. Finally, to counteract the detrimental effects of TA, the psychological safety of electric industry workers must be prioritized by introducing VR in a sandbox environment for pressure-free exploration, explicitly normalizing initial apprehension, and reinforcing the simulation as a safe space to make mistakes. By strategically enhancing SE and FC while actively managing TA, organizations can significantly improve PEOU and PU, fostering acceptance of VR-based safety training among electric industry workers.

Limitations of this study

Despite the significant findings, several limitations must be acknowledged to provide a balanced interpretation. First, this study focused on electric industry workers’ initial intention to use VR technology following a single exposure. While intention is a strong precursor to behavior, it does not guarantee sustained and long-term adoption. Future research should employ longitudinal designs to track how SE and TA evolve as users transition from novices to regular users. Second, the reliance on self-reported data may introduce the potential for social desirability bias because participants may overstate their positive attitudes in a professional training environment. Future studies could integrate objective data, such as VR system logs (e.g., time spent in-headset) or supervisor evaluations, to triangulate self-reported perceptions. Third, the data were collected exclusively within the Chinese electric industry. Given that safety culture and technology acceptance are deeply influenced by organizational and national cultures, caution should be exercised when generalizing these results to other regions. Cross-cultural validation is recommended to test the model’s robustness in Western or other Asian industrial contexts. Lastly, because participants were recruited from professional training centers, they may possess higher baseline motivation than the general workforce. Future investigations should utilize stratified random sampling across diverse company sizes and geographic locations to capture a broader spectrum of digital literacy.

Conclusion

As VR technology becomes increasingly integrated into safety training within the electric industry, understanding workforce acceptance remains critical. This study addressed this gap by validating an expanded TAM among frontline workers in China. The key findings and implications are summarized as follows:

• The study successfully validated an extended TAM incorporating SE, FC, and TA, explaining a substantial proportion of variance in workers’ intentions to use VR training.

• Results confirmed that SE and FC are significant positive predictors of PU and PEOU.

• TA was identified as a significant barrier, exerting a detrimental negative influence on the core perceptions of the technology.

• This research enriches the existing literature by providing a specialized framework tailored to the unique high-risk environment of the electric industry.

• The findings offer an evidence-based roadmap for stakeholders and designers to prioritize interventions that bolster worker confidence, provide robust technical support, and proactively mitigate user anxiety to foster successful VR adoption.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

C.Z., conceptualization, formal analysis, funding acquisition, investigation, methodology, writ-ing—original draft; L.L., conceptualization, data curation, formal analysis, methodology, writ-ing—original draft; J.H., conceptualization, investigation, methodology, resources, writ-ing—original draft; F.H., conceptualization, project administration, resources, validation, writ-ing—original draft; C.L., conceptualization, supervision, writing—original draft; C.L., conceptu-alization, writing—review and editing; S.S.M., conceptualization, methodology, writing—original draft, writing—review and editing. All authors have read and agreed to the published version of the manuscript.

Funding

This study was supported by the 2025 Power Grid Construction Line and Equipment Training Resource Development Project (16064478).

Data availability

The data presented in this study are available on request from the corresponding author.

Declarations

Competing interests

The authors declare no competing interests.