Earticle

Explainable artificial intelligence (XAI) has expanded from a niche concern into a defining requirement for the deployment of opaque machine-learning models, yet the field's rapid growth has produced a fragmented body of evidence in which the same handful of techniques are applied across very different domains with little agreement on how explanations should be evaluated. This study maps the structure of that evidence rather than pooling its effect sizes. Departing methodologically from variance-weighted metaanalysis, we conduct a systematic mapping with structured content-analytic coding and bibliometric network synthesis over a tri-lingual corpus spanning the English, Chinese, and Korean literatures from 2021 to 2025 (coded sample N = 240; macro indicators triangulated against an indexed corpus of 4,781 articles). Each study is coded for application domain, XAI method, explanation scope and locality, and evaluation practice, and the resulting landscape is analysed through method-prevalence profiles, a method×domain co-occurrence matrix, a method co-occurrence network, and a five-year compositional trend. Three findings stand out. First, the field is strongly SHAP-centric: post-hoc, model-agnostic attribution dominates, while intrinsically interpretable glassbox models remain marginal. Second, method choice is largely dictated by data modality rather than by explanatory goal, producing a stable but narrow method×domain coupling. Third, and most consequentially, a persistent evaluation gap remains: a plurality of applied studies present explanations illustratively, without faithfulness, robustness, or human-grounded validation. We argue that the central problem facing XAI is not a shortage of methods but a shortage of disciplined evaluation, and we outline an evaluation-first agenda and a minimal reporting standard for explainability research.

In the digital era, the “always-on” nature of social media generates distinct psychological reactions when individuals are offline. Some users experience a persistent urge to stay connected for fear of missing social updates or work-relevant information—the fear of missing out (FoMO)—whereas others gain comfort from disconnection, an experience termed the joy of missing out (JoMO). Although such missing-out experiences are widely studied among adolescents and general employees, they remain underexplored among university faculty, a highly knowledge-intensive group whose work requires continuous learning and academic exchange. This study examines how four facets of the missing-out experience—personal-information missing experience, social missing experience, missing state, and JoMO—affect faculty job performance, and whether work thriving mediates these effects. Drawing on self-determination theory, social comparison theory, and the socially embedded model of thriving at work, an integrated “missing experience–work thriving–job performance” model is constructed. Survey data were collected from faculty at SZ University via an online questionnaire, yielding 106 valid responses, and analyzed using partial least squares structural equation modeling (PLS-SEM). The results show that personal-information missing experience (β = 0.433), missing state (β = 0.283), and JoMO (β = 0.221) significantly and positively predict work thriving, whereas social missing experience exerts a significant negative effect (β = −0.341); work thriving in turn strongly predicts job performance (β = 0.744). The findings indicate that moderate missing-out experiences can stimulate reflective growth and resource-seeking behaviors, while excessive social FoMO depletes psychological resources. The study extends missing-experience research to higher education and offers universities evidence-based guidance for performance management and faculty support.

It is necessary to review the appropriability which is a method of securing the return on R & D investment, and to examine how to organize the mechanism of equity by industry and apply it systematically. Although there have been various attempts to identify the reasons for the necessary mechanism of self-sufficiency in companies, there is still no consensus on their effectiveness and appropriateness. This can be attributed to the fact that it did not reflect the causal relationship and the dynamic structure between the complex market mechanism of the market (industry, culture, etc.) and the various variables included in it. In this paper, we analyze the dynamic structure of R & D investment through the model of causal model of system dynamics.

As the new energy vehicle (NEV) industry accelerates its transformation from electrification toward intelligence, intelligent driving functions have become one of the key factors influencing user satisfaction and brand competitiveness. Traditional satisfaction surveys rely heavily on structured questionnaires and struggle to uncover the complex latent topics and emotions embedded in massive volumes of unstructured online reviews. This study therefore applies the BERTopic topic-modeling method to systematically mine the core concerns, sentiment tendencies, and topic-evolution patterns of NEV intelligent-driving users, and constructs an integrated “topic–sentiment–satisfaction” analytical framework that provides data-driven decision support for automakers. Review texts were collected from four platforms—Dongchedi, Autohome, Weibo, and Zhihu—yielding 5,071 valid comments after cleaning, word segmentation, and SnowNLP-based sentiment annotation. BERTopic was then used for topic modeling, combined with sentiment analysis, brand comparison, and topic-evolution analysis. The results show that negative reviews account for 55.0% while positive reviews account for only 1.8%, indicating a strong overall tendency toward negative expression; 62 topics were identified, of which “overall vehicle experience” was dominant at 47.9%. Intelligent-driving topics such as “Tesla FSD,” “AITO M7 pricing,” and “NIO ET5T & VLA” were highly active, with negativesentiment proportions generally exceeding 60%. Brand comparison revealed differentiated discussion foci around each brand’s signature technologies, and topic-evolution analysis found explosive growth of intelligent-driving topics from 2024 onward, closely tracking technology iterations and product launches. Theoretically, the study extends BERTopic to the vertical NEV intelligent-driving domain and verifies its effectiveness on short, term-dense texts; practically, it reveals widespread user dissatisfaction with the maturity of intelligent-driving functions and offers automakers actionable recommendations.

Process Virtualization Theory (PVT) has become a central framework for explaining which processes can be moved from physical to virtual settings, yet after a decade of large-scale virtualization in remote work, online education, and e-government, the empirical evidence on its core propositions remains fragmented and, in places, contradictory. This study integrates that evidence using a mixed approach that combines a systematic literature review with a random-effects meta-analysis. Following PRISMA guidance, 27 deduplicated records were screened and 11 studies were retained for qualitative synthesis, of which six reported extractable standardized effect sizes for quantitative synthesis (k = 3–4 independent samples per relationship; cumulative N ≈ 1,047–1,390). Standardized path coefficients and zero-order correlations were harmonized as effect sizes, transformed via Fisher’s z, and pooled with the DerSimonian–Laird estimator; effect-direction vote-counting was used as a robustness check. Sensory requirements exhibited a robust and consistent negative association with process virtualizability (r = −0.38, 95% CI [−0.50, −0.24], p < 0.001). Relationship requirements showed a weaker but significant negative association (r = −0.21, p = 0.015). In contrast, the pooled effects for synchronism requirements (r = −0.22, p = 0.16) and for identification-and-control requirements (r = −0.12, p = 0.43) were non-significant and highly heterogeneous (I² > 95%), with sign reversals across contexts. IT representation and IT reach were consistently supported as positive enablers, whereas monitoring capability was not. The evidence indicates that PVT’s requirement propositions are partially supported and strongly context-bounded. We derive three research priorities: harmonizing dependent variables, reconceptualizing context-sensitive constructs, and integrating PVT with adjacent theories of presence and crisis-driven transformation.

The digital industry encompasses the ICT sector and the broader transformation of value creation through data, artificial intelligence, cloud computing, and software. This study develops a transferable policy framework for fostering digital industries without assuming any specific city or province. Six strategic pillars are proposed: place-based prioritization; staged digital transformation support for SMEs; growth and market creation for digital firms; trusted data, computing, and testbed infrastructure; demand-linked talent development; and dedicated governance with performance management. A five-year roadmap, digital maturity model, project-selection criteria, public procurement mechanisms, data governance, cybersecurity controls, and outcome-oriented KPIs are specified. Basic local governments should act as ecosystem orchestrators connecting local demand, technology, talent, finance, and markets.