Leveraging artificial intelligence (AI) technologies to advance digital and intelligent transformation has become an important pathway for Chinese manufacturing enterprises to achieve high-quality development. At the same time, the rapid expansion of AI applications has sparked widespread societal concern regarding whether AI adoption inevitably substitutes for human labor. Accordingly, two interrelated questions have attracted growing attention from both academia and practice: which factors or combinations of factors shape AI adoption intensity in manufacturing enterprises, and how does AI adoption intensity affect the scale of labor demand? Existing research, however, tends to examine either the determinants of AI adoption behavior at the individual level or the effects of AI adoption intensity on the scale of enterprise labor demand. Systematic evidence on the enterprise-level antecedents of AI adoption intensity and their implications for labor demand remains limited. In particular, prior studies have rarely investigated the interactive effects of technological, organizational, and environmental (TOE) factors on AI adoption intensity from a configurational perspective, nor have they examined the mediating role of adoption intensity in linking TOE-based configurations to the scale of labor demand.
Against this backdrop, this study draws on the TOE framework to examine the configurational pathways through which TOE factor combinations drive AI adoption intensity in manufacturing enterprises and to assess how such adoption intensity subsequently influences the scale of labor demand. Methodologically, the study adopts a mixed-methods approach that combines fuzzy-set qualitative comparative analysis (fsQCA) and regression analysis, using data from 148 A-share listed manufacturing enterprises in 2023. Specifically, fsQCA is employed to identify multiple TOE configuration patterns associated with high AI adoption intensity, while regression analysis is used to test the mechanisms through which TOE configurations and AI adoption intensity jointly influence the scale of labor demand.
The fsQCA results reveal three distinct pathways to high AI adoption intensity. The first is an “Organizational Dual-Core Driven under Limited Technological and Environmental Constraints” pathway, in which enterprises compensate for limited technological and environmental conditions (e.g., low R&D investment, weak policy support, and limited innovation capacity) by strengthening internal organizational transformation, knowledge intensity, and risk-taking capacity. The second is a “Technology-Organization Collaborative Innovation with Policy Nudges under Limited R&D Investment” pathway, whereby enterprises with constrained R&D resources enhance AI adoption by building internal-external collaboration mechanisms and leveraging government policy support to bolster technological innovation, organizational knowledge intensity, and risk-taking capacity. The third is a pathway dominated by R&D investment and risk-taking, indicating that even when organizational capabilities and external support are weak, enterprises can still achieve high AI adoption intensity by substantially increasing R&D investment and organizational risk-taking.
Regression analyses further demonstrate the mediating role of AI adoption intensity between TOE configuration patterns and the scale of labor demand. The results show that high AI adoption intensity generally has a negative impact on the scale of labor demand. Specifically, the first two configuration patterns—while promoting AI adoption intensity—are associated with a reduction in the scale of labor demand. By contrast, the “R&D investment and risk-taking dominant” configuration increases AI adoption intensity without exerting a significant negative effect on labor demand.
This study offers both theoretical and practical contributions. Theoretically, it (1) advances understanding of AI adoption intensity by examining the synergistic effects of TOE dimensions from a configurational perspective, thereby addressing the limitations of linear, single-factor approaches; (2) extends research on the mechanisms through which AI adoption intensity influences the scale of labor demand in manufacturing enterprises; and (3) advances beyond prior paradigms that focused solely on antecedents of AI adoption intensity or solely on its scale of labor demand consequences, by integrating antecedent configurations and outcome effects into a unified causal system, thereby forming a logical loop more consistent with real-world scenarios. Practically, the findings offer guidance for manufacturing enterprises in selecting appropriate AI adoption pathways and optimizing labor management strategies, while also providing policy insights for governments seeking to balance AI-driven transformation with employment stability.

Against the backdrop of the digital intelligence era which is characterized by rapid shifts in technological paradigms and intensifying market competition, the fleeting nature of opportunity windows imposes unprecedented demands on the growth and survival of startups. As nascent and inherently vulnerable organizations, startups grapple with resource scarcity, limited legitimacy, and underdeveloped organizational capabilities, all of which hinder their ability to elevate their ecological niche. Effectively harnessing artificial intelligence (AI) to overcome these vulnerabilities and advance their niche position has thus become a critical strategic imperative. While existing research widely acknowledges AI's transformative potential in unlocking data value, optimising decision-making efficiency, and reshaping value creation pathways, the black-box mechanism through which AI concretely translates into sustainable competitive advantages for resource-constrained startups remains under-explored. Furthermore, according to the theory of technological affordance, the realisation of a technology's value potential is not automatic but depends on internal organisational behaviour and strategic choices. This creates a complex, dynamic interplay between technological characteristics, strategic orientation, and organisational behaviour.
In response, this study integrates technological affordance theory and organizational behavior theory to develop a unified theoretical model centered on AI affordance, scenario-driven innovation, and the entrepreneurial ecological niche. Specifically,the study conceptualize AI affordance along two distinct dimensions: autonomous AI affordance and interactive AI affordance. The former captures AI's capability for independent task execution and automated decision-making, while the latter reflects its potential to enable human-machine collaboration and facilitate information exchange. This study addresses these core questions: Firstly, Can, and if so, how do the distinct dimensions of AI affordance enhance the ecological niche of vulnerable startups? And, as a novel paradigm of the digital era, can scenario-driven innovation serve as a pivotal mediating mechanism in this process? Secondly, how does expectation gap, as an internal performance feedback signal, moderate the intensity with which start-ups leverage AI affordance to drive scenario-driven innovation?
This study focuses on technology-based startups and collects data through rigorously designed questionnaires administered to managerial personnel with comprehensive knowledge of their firms' operations and AI applications. A total of 500 questionnaires were distributed over a four-month period, yielding 315 valid responses after stringent screening. Empirical analysis yields three principal conclusions: First,autonomous AI affordance strengthens niches via operational automation, while interactive AI affordance enhances resource acquisition and external collaboration.Second, scenario-driven innovation plays a crucial mediating role between AI affordance and ecological niche. It provides concrete contextual frameworks for unlocking AI's potential, enabling startups to bridge the gap between technological promise and market performance by reconfiguring resources and creating new opportunities through integrating technology with specific business scenarios. Third, expectation gaps exert differentiated moderating effects. Both persistence and scope expectation gaps positively moderate the relationship between autonomous AI affordance and scenario-driven innovation, indicating that performance pressures prompt managers to leverage AI's autonomous capabilities for problem solving. However, expectation gaps do not significantly moderate the relationship between interactive AI affordance and scenario-driven innovation. This suggests that under performance gap pressures, the vast information and complex interactions required for AI affordance may divert managers' limited attention, disrupting technology optimisation processes within specific scenarios.
This study offers both theoretical and practical insights. Theoretically, it advances the literature in three ways: (1) by applying a technological affordances lens to reveal how AI interacts with resource-constrained startups; (2) by positioning scenario-driven innovation as a critical bridge between AI capabilities and entrepreneurial growth, thereby extending its relevance beyond design or service contexts into strategic technology deployment; and (3) by introducing the expectation gap as a behavioral moderator to broaden the boundaries of AI's influence on corporate development. Practically, the findings urge startups to move beyond passive AI adoption. Instead, they should deliberately co-create application scenarios that align AI affordance with their strategic resource needs. Moreover, managers under performance pressure should prioritize autonomous AI solutions when attention resources are scarce.

Facilitating the transfer of tacit knowledge from users to artificial intelligence (AI) (referred to as "human-machine tacit knowledge transfer") is crucial for enterprises to address persistent challenges in AI implementation,including data scarcity,algorithmic constraints,and limited computational resources.This process is thus essential for unlocking the strategic value of AI and gaining a competitive edge.Consequently,how to promote human-machine tacit knowledge transfer has become an important issue that enterprises urgently need to address.
Although research on human-machine tacit knowledge transfer has been increasing,it is still scattered and lacks an effective framework.In addition,existing research emphasizes technological rationality,focusing on the technical implementation of human-machine tacit knowledge transfer from the perspective of AI as the knowledge transfer object,while largely overlooking the role of users as active knowledge contributors.This not only leads to an imbalance in research perspectives,but also results in a dual rupture of human-machine tacit knowledge transfer in terms of subject (users) and object (AI),as well as between behavior and technology,thereby hindering practical guidance for organizations seeking to facilitate effective knowledge transfer.
Therefore,this study investigates how human-machine tacit knowledge transfer unfolds from the user's perspective.Drawing on the theory of AI socialization,the study specifically examines how AI capabilities influence human-machine tacit knowledge transfer.It develops a research model in which AI capability serves as the independent variable,human-machine tacit knowledge transfer as the dependent variable,new professional role identity as a mediator,and perceived value and perceived threat as moderators.On this basis,this study collected data from 321 enterprises through a questionnaire-based survey and conducted empirical analysis and testing of research hypotheses using methods such as structural equation modeling and hierarchical regression analysis.
The results show that AI capabilities positively affect human-machine tacit knowledge transfer; furthermore,AI capabilities positively influence users' new professional role identity,through which AI capabilities positively influence human-machine tacit knowledge transfer.This also indicates that AI capabilities can not only directly affect human-machine tacit knowledge transfer,but also indirectly affect it through the mediating role of users' new professional role identity.The results also show that perceived value positively moderates the influence of AI capabilities on users' new professional role identity; whereas perceived threats do not hinder the influence of AI capabilities on users' new professional role identity.This also indicates that perceived value can indirectly enhance the impact of AI capabilities on human-machine tacit knowledge transfer.
This study contributes to the theory of human-machine tacit knowledge transfer:first,it promotes the diversification of research perspectives on human-machine tacit knowledge transfer by introducing a user behavior perspective; second,it provides a behavioral mechanism for the theory of human-machine tacit knowledge transfer,promoting the connectivity of research content in this field; third,it constructs an analytical framework of "capabilities (AI capabilities) - cognition (users' new professional role identity)-behavior (human-machine tacit knowledge transfer)" to study the behavioral mechanism of human-machine tacit knowledge transfer,providing effective theoretical support and a logical framework for subsequent research.
The findings offer actionable insights for organizations deploying AI on how to obtain the expected value of AI.Specifically, companies should prioritize building users' AI capabilities to facilitate tacit knowledge transfer between humans and machines. Additionally, aligning users' AI competencies with their evolving professional identities can further enhance this knowledge exchange. Organizations should also actively communicate the value of AI to encourage users to fully leverage these capabilities.The results also suggest that enterprises should not be fixated on whether to eliminate users' perceived threat of AI.Instead,they can guide users to turn perceived threats into motivation,thereby encouraging them to use their AI capabilities to carry out new work and transfer tacit knowledge to AI.

In the new technological revolution,technology diffusion underpins major countries' innovation strengths. China's super-large-scale market,synergized with data factors,uniquely enables industries to break the S-shaped curve saturation of traditional technology diffusion and reshape it into a J-shaped trajectory,creating latecomer catch-up conditions. Supported by this integrated diffusion model,sectors including new energy vehicles and 5G communications have achieved efficient technology diffusion and technological catch-up. This study thereby explores whether China's super-large-scale market advantage,empowered by data factors,can overcome the saturation of conventional technology diffusion,and how to translate such market strengths into novel industrial innovation advantages.
This study finds that the linear modeling of the super-large-scale market in existing studies only captures its stretching effect on the S-shaped curve of traditional technology diffusion,with a theoretical mechanism indistinguishable from that of an ordinary large-scale market,and generally ignores the synergistic activation role of data factors. This study argues that China's super-large-scale market gives rise to a large number of marginal segmented markets under the overlapping effects of institutional unity,spatial heterogeneity and market multi-hierarchy,and data factors can precisely activate this potential market capacity,breaking through the static market assumption of traditional technology diffusion. From a micro-firm perspective,this paper systematically explains the internal mechanism by which data factors activate marginal segmented markets and synergize with the super-large-scale market to improve enterprises' technology diffusion efficiency and drive technology diffusion to break through saturation constraints. Meanwhile,it analyzes the core action channels of the synergistic effect between the super-large-scale market and data factors from three dimensions:artificial intelligence penetration,innovative knowledge integration and innovation collaboration networks.
In accordance with the above analysis,this study constructs a theoretical model for the super-large-scale market potential to break through enterprises' technology diffusion saturation constraints and realize the reconstruction of technology diffusion trajectories through the collaboration of data factors,and conducts an empirical test of the model using data of listed manufacturing companies from 2009 to 2024. The empirical results show that (1) the super-large-scale market advantage has the potential effect of breaking through enterprises' technology diffusion saturation constraints,reconstructing diffusion trajectories and promoting continuous technological iteration; (2) the technology diffusion breakthrough effect of the super-large-scale market highly relies on collaboration with enterprises' utilization of data factors,and data factors are the core premise for activating market potential; (3) artificial intelligence penetration,innovative knowledge integration and innovation collaboration networks are important channels for the synergistic effect between the super-large-scale market and data factors,and policy practices to promote the upgrading of technology diffusion can be precisely implemented through these three channels. The empirical analysis verifies the internal mechanism of transforming the super-large-scale market advantage into a new type of industrial innovation advantage under the collaboration of data factors,and provides theoretical support and policy reference for promoting the reconstruction of technology diffusion trajectories through the collaboration of the super-large-scale market and data factors,realizing the transformation of major-country scale advantages into industrial innovation advantages,strengthening the competitiveness of industrial technology diffusion,and serving the construction of a new development pattern and the shaping of international competitive advantages.
This study shifts the market advantage literature from static size effects to structural activation. By dissecting how institutional unity,spatial heterogeneity,and market multi-hierarchy generate marginal segmented markets,it reveals the micro-foundation through which super-large-scale markets transcend traditional diffusion ceilings. Integrating data collaboration into the diffusion model,it demonstrates that AI penetration,knowledge integration,and collaboration networks constitute the channels that transform latent demand into sustained technological iteration. This framework explains how latecomer firms bypass saturation constraints in the intelligent technology era,extending theoretical boundaries beyond linear market-size effects toward data-empowered trajectory reconstruction.

As digitalization continues to advance, the emergence of new technologies such as cloud storage and converged storage has eliminated the limitations on personal storage space, making digital hoarding behavior a gradually widespread social phenomenon. As a unique product of the digital age, digital hoarding behavior manifests as excessive accumulation of digital resources, and this behavior is particularly pronounced and typical in the context of work. However, the majority of existing studies have focused on the impact of digital hoarding behavior within the social media, with relatively limited exploration of the underlying triggers in the context of work. Digital hoarding behavior in the workplace not only poses psychological health risks to individuals and reduces employees' work efficiency, but also triggers a series of severe consequences for businesses, such as decreased productivity and cybersecurity threats. Therefore, exploring the influencing factors of digital hoarding behavior in the workplace is essential for guiding employees in digital resource management and enhancing organizational data utilization efficiency.
Drawing on attribution theory, this study selects six key conditional factors from two levels, internal and external factors, to construct a causal configuration framework for digital hoarding behavior in the workplace. Specifically, this study takes job insecurity, perceived usefulness, work responsibility, artificial intelligence usage, time pressure, and industry competitive pressure as the core antecedent conditional factors. Then it uses NCA and fuzzy set qualitative comparative analysis (fsQCA) to conduct configuration analysis on 286 samples of enterprise employees, and explores how internal and external factors affect employees' digital hoarding behavior in the workplace.
Through dual testing using the NCA method and the fsQCA method, this study finds that factors including job insecurity, perceived usefulness, work responsibility, artificial intelligence usage, time pressure, and industry competitive pressure cannot individually constitute necessary conditions for high digital hoarding behavior or non-high digital hoarding behavior, nor can they individually constitute sufficient conditions for digital hoarding behavior. This finding indicates that employees' digital hoarding behavior in the workplace is not driven by a single factor, but rather the result of the synergistic linkage and interaction of multiple antecedent conditions. The study reveals that there are four paths contributing to employees′ high-level digital hoarding behavior, namely technology-internal-driver, pressure-internal-driven, external-driven, and internal-external synergistic. Additionally, there are three paths leading to employees′ non-high-level digital hoarding behavior, including the motivation-deficiency type, value-absence type and pressure-insufficiency type. This study adopts a configurational perspective to analyze different combinations of antecedent conditions underlying employees′ digital hoarding behavior, thereby deepening the understanding of the joint effects of the antecedent conditions of digital hoarding behavior.
The theoretical contributions of this study are manifested in three aspects. First, this study clarifies the various factors influencing employees' digital hoarding behavior in the workplace, analyzes the inducements of digital hoarding behavior from the employee perspective, expands and complements the research perspective on digital hoarding behavior, and enriches the understanding of the influence mechanism of digital hoarding behavior. Second, this study constructs a framework for the influence of digital hoarding behavior from the perspective of attribution theory, which expands the application context and scope of attribution theory. Third, this study explores the configurational paths of digital hoarding behavior, further revealing the underlying mechanisms influencing employees' digital hoarding behavior.
This study puts forward practical implications for business managers and employees in the following three aspects. First, it is necessary to manage employees' personal digital resources well, clarify job responsibilities, and avoid mental health problems caused by excessive sense of responsibility. Second, employees should improve their ability to identify the value of digital information, filter useful information, and enhance their information processing and application capabilities. Third, employees need to plan their work schedules reasonably, and business managers should also provide a relaxed working environment for employees, which will help businesses eliminate employees' digital hoarding behavior and achieve a win-win situation for both businesses and employees.

Data elements refer to the conceptualization of "data" when discussing productivity and production relations in economics, emphasizing the role of data in promoting production value. The Fourth Plenary Session of the 19th Central Committee of the Communist Party of China proposed the goal of "market-determined pricing, autonomous and orderly flow, and efficient and equitable allocation" for the construction of a data element market system. This has initiated the 2.0 era of a unified, professional, and orderly development of a data element trading market ecosystem. Data element trading policies serve as guidelines for the construction of a data element trading market, and the lack of a comprehensive compliance framework has led to frequent cases of data leakage and illegal transactions, thereby restricting the deep participation of individuals and enterprises in the market. Therefore, the construction of the data element trading market cannot be separated from policy regulation of transactional behaviors. However, what factors affect the formulation and implementation of data element trading policies ?What are the internal relationships among them?These are issues that need to be addressed. Addressing these problems has important practical significance for building China's data element trading market and implementing the national big data strategy.
The interaction between data element trading policies and the external environment is intense. In existing research,isolated system construction and analysis of influencing factors make it difficult to effectively capture the status of the policy ecosystem. Integrating multi-source heterogeneous information, despite their information heterogeneity, asynchrony and redundancy, and capitalizing on the unique characteristics of each source enables the extraction of more comprehensive and accurate insights.This approach significantly enhances the accuracy of empirical investigation and provides stronger support for exploring policy ecosystem factors. Therefore, based on the integration of multi-source information, this research follows the procedural logic of grounded theory and takes the ecological factors of data element trading policies as the core issue. It derives theories from multi-source information, analyzes and refines concepts, gradually links ecological factors, and determines their hierarchical structure and attributes, ultimately constructing a data element trading policy ecosystem factor model.
The research finds that the ecological factor system of China's regional data element trading policies is centered around regional policies such as those concerning data tiering, data pricing, data ownership, and post-transaction maintenance, and is further integrated with environmental factors,including political, economic, cultural, technological, and risk-related dimensions as well as policy stakeholders, including policymakers, implementers, and recipients. Together, these components constitute a complex and dynamic system of multi-source collaboration. The diverse ecological factors and their intricate symbiotic relationships drive the ecological evolution of the data element trading policy system from "disorder" to "order" in an iterative and helical manner. From a policy ecology perspective, optimizing the ecological factor system of data element trading policies hinges on three core principles: First, environmental adaptability: cultivating a supportive policy ecological environment to lay the foundation for effective implementation. Second, policy dynamism: establishing a closed-loop mechanism of "evaluation-feedback-adjustment". Third, stakeholder collaboration: constructing a "multi-party governance" ecosystem network for data element trading.
Within the data element trading policy ecosystem, by drawing an analogy to the basic components of a natural ecosystem,such as environment, agents, and objects,different ecological factors occupy distinct ecological niches. This framework reflects, from multiple perspectives, the interdependence and mutual adaptation among stakeholders, between stakeholders and the environment, between stakeholders and policies, among policies themselves, and between policies and the environment. From the perspective of policy ecology, optimizing data element trading policies not only broadens the scope of research on such policies but also enriches and refines the ecological map of the data element trading policy system. This approach not only endows the policy system with "systemic wisdom" but also facilitates the scientific assessment of policy applicability and helps maintain the stability of the data element trading policy ecosystem.

In the context of building a unified national market, deepening market-oriented reforms in capital factors characterized by autonomous and orderly inter-regional capital flow helps improve resource allocation efficiency, unleash market potential, and promote coordinated regional development. However, the "siphon effect" of capital in China's developed regions remains prevalent. Affected by market segmentation, local protectionism, and other factors, inter-provincial capital flows face obstacles and inefficient allocation. The question of how to promote the smooth and efficient flow and allocation of capital over a broader range has become a hot topic in both policy practice and theoretical research. The rapid development of digital technologies has given rise to the rise and maturation of platforms, with corporate embedding in platform ecological creating broader space and potential for cross-region capital flows. However, existing studies mainly focus on the impact of platform ecological embedding on innovation behavior, digital transformation, supply chain coordination, and value chain upgrading. Few scholars have explored the specific role and mechanism of platform ecological embedding in cross-region capital flows from an investment perspective.
In order to explore the impact of platform ecological embedding on cross-regional capital flows and its underlying mechanisms, this study draws on a 36 252 firm-year panel of Shanghai- and Shenzhen-listed companies(2010-2023),and employs a two-way fixed effects model. It follows a four-stage research protocol. First, platform ecosystem embeddedness was quantified as the ratio of 82 platform-centric keywords extracted from the “Management Discussion & Analysis” section via Python text-mining to total MD&A words. Second, subsidiaries reported in CSMAR were manually geocoded, the cross-regional capital-flow variable(Cinvest) was constructed as the natural log of the number of non-local subsidiaries plus one, and addresses are validated with Baidu Maps. Third, an OLS two-way fixed-effects model was estimated, incorporating controls for fourteen firm-level financial, governance, and regional-level institutional variables. Finally, the baseline estimate was subjected to an eight-fold robustness sequence: 2SLS with industry-mean and lag-2 instruments, PSM 1:2 nearest-neighbor matching, digital-transformation split-sample, alternative embeddedness and investment metrics, high-dimensional fixed effects, and sample exclusions for COVID-19 and IT industries.
The results show that platform ecological embedding significantly promotes cross-regional capital flows. Mechanism tests indicate that platform ecological embedding facilitates inter-regional capital flow by mitigating corporate uncertainty perception and promote high-quality development of enterprise. Heterogeneity analysis reveals that the effect of platform ecological embedding on promoting capital flow across regions is more significant for private enterprises, firms with short-sighted management and higher financing constraints, and the enterprises in areas with backward development of digital economy. Furthermore, platform ecological embedding effectively suppresses the positive impact of inter-regional capital flow on operational cost ratios. Therefore, the government should upgrade digital infrastructure(data centers, industrial internet, 5G), offer fiscal/talent incentives, and dismantle regional protection via unified market rules and interoperable credit data. For platform firms, managers should treat platform embedding as a dynamic capability by proactively co-evolving with complementors, upgrading data analytics competences, and institutionalizing cross-unit knowledge-sharing routines to transform ecosystem affordances into sustainable cross-regional competitive advantage.With regard to the incumbents, it is essential to strategically embed in third-party or self-built platforms, co-create value with ecosystem partners, and leverage digital tools to synchronize headquarters with geographically dispersed subsidiaries, thereby enhancing responsiveness to heterogeneous market demands and institutional environments.
By integrating platform ecological embedding and cross-regional capital flows into a single framework, this study provides evidence for a direct causal link and simultaneously extends the drivers of inter-regional investment and the economic consequences literature on platform ecosystems. It further examines uncertainty-mitigation and developmental-spillover effects, and dissects the micro-mechanism through which platform ecological embedding channels capital across regions, thereby completing the theoretical chain linking embedding to inter-regional capital flows. The findings provide theoretical support and empirical basis for understanding the enabling mechanism of platform ecological embedding to enable cross-regional capital flows in the context of digital transformation, and also provides practical inspiration for optimizing cross-region capital allocation and accelerating the construction of a national unified market.

Currently, China's economy has entered a new normal after a period of rapid development, with market demand gradually becoming saturated and the limitations of exploitative innovation increasingly evident. The transition from exploitative innovation to exploratory innovation, known as an innovation leap, is considered a crucial strategy for enterprises to escape from existing innovation routines, identify future technological trajectories, and seize R&D opportunities. If enterprises lack the capability to achieve an innovation leap from exploitative innovation to exploratory innovation, they risk falling into the “exploitative innovation trap” due to their inability to cross the innovation gap in shifting innovation models. Consequently, enterprises urgently need to upgrade and transform their ambidextrous innovation models to break free from the trap. Particularly in the current context of restrictions on key core technologies, enhancing innovation openness has become a critical pathway for enterprises to access external resources and achieve innovation leap. However, previous research has rarely examined the antecedent mechanisms of innovation leap, and there is a lack of exploration into innovation leap in the digital context. To address these research gaps, this study analyzes how innovation openness influences enterprise innovation based on innovation recombination theory.
This study constructs a pathway for achieving enterprise innovation leap based on innovation recombination theory. Using a panel data set of 330 Chinese A-share listed companies spanning the period 2007 to 2024, it empirically examines how innovation recombination differences, manifested in knowledge flexibility and knowledge complexity triggered by the breadth and depth of innovation openness, affect enterprises' innovation leap, as well as the moderating role of enterprise digitalization level in the above relationship.
The study results indicate that the innovation openness breadth positively facilitates the innovation leap from exploitative innovation to exploratory innovation by enhancing knowledge flexibility, while the innovation openness depth exhibits an inverted U-shaped effect on the innovation leap due to increased knowledge complexity. Furthermore, enterprises' digitalization level positively moderates the relationship between innovation openness breadth and innovation leap, and further strengthens the inverted U-shaped relationship between innovation openness depth and innovation leap.
The theoretical contributions of this paper are presented as follows. First, departing from previous studies which have narrowly focused on the consequences of enterprise innovation leap, this paper shifts the focus to the influence mechanism of innovation openness on innovation leap, expanding the knowledge-based antecedents of innovation leap, answering the question of “how innovation leap occurs”. Second, based on innovation recombination theory, this paper distinguishes between the breadth and depth of innovation openness and thoroughly investigates their differential effects on the innovation leap through the distinct recombination mechanisms of knowledge flexibility and knowledge complexity. Thereby, it not only provides a novel theoretical perspective for research on innovation leap but also extends the application boundaries of innovation recombination theory. Finally, this paper introduces the moderating role of enterprise digitalization level in the impact of innovation openness on innovation leap. It delves into the mechanism and impact of digitalization level on changes in enterprises' ambidextrous innovation strategy decisions. The findings not only broaden the research scope of enterprise digitalization but also enrich and deepen the theoretical understanding of ambidextrous innovation strategic management in a digital context.
To foster a virtuous cycle between exploitative innovation and exploratory innovation, enterprises should adopt open innovation strategies while balancing the breadth and depth of openness. Expanding openness breadth requires actively collaborating with suppliers, universities, research institutions, and other external partners to integrate diverse knowledge resources, which fuels sustained innovative breakthroughs. For openness depth, enterprises need moderately close, long-term stable partnerships with collaborators to facilitate effective knowledge sharing and conversion. They should avoid over-reliance on narrow knowledge domains to maintain an optimal level. Additionally, advancing digital transformation is critical. Enterprises can leverage cloud computing, big data, and AI to build a flexible, efficient innovation ecosystem. When enterprises strengthen their capabilities in data analysis and opportunity identification, they can achieve innovation leaps at lower transition costs and enhance the scientific rigor and adaptability of their innovation strategy decisions.

Against the backdrop of the accelerated formation of global innovation strategic networks, enterprises increasingly find it difficult to pursue “go-it-alone” innovation. As a vital component of open innovation and an advanced stage of internationalization, R&D internationalization is crucial for enterprises in emerging economies such as China to enhance innovation performance and international competitiveness. However, when constructing overseas R&D networks, these enterprises often face the dual challenges of substantial information gaps and high entry barriers, which in turn increase host-country organizational adaptation costs, reduce the efficiency of cross-border resource allocation, and lead to redundant R&D investment. At the same time, digital transformation has become a core strategic initiative for enterprises to capture market opportunities and respond to environmental challenges. This raises several key questions: Can digital transformation help Chinese enterprises leverage digital capabilities to serve as a “springboard” for R&D internationalization? What heterogeneity characterizes its effects? How can the theoretical mechanism between the two be explained? What roles do host-country information accessibility and market entry level play in this process?
Drawing on the new OLI theoretical paradigm, this study employs data on Chinese listed firms and their overseas investments from 2007 to 2023, and applies a fixed-effects panel model to empirically examine the impact of digital transformation on the level of R&D internationalization. Further, it investigates heterogeneous effects based on ownership (state-owned vs. non-state-owned) and executives' overseas backgrounds. In addition, mechanism tests are conducted to explore whether digital transformation enhances enterprises' ability to exert the new ownership-location-internalization (OLI)advantages of open resources, linkage, and integration by improving host-country information accessibility and market entry level, thereby overcoming the inherent barriers to R&D internationalization. From a theoretical perspective, the new OLI framework provides an important analytical basis and reference for outward international investment in the digital era. Digital transformation strengthens these three advantages, shrinking information gaps and entry barriers, cutting risk/cost and boosting R&D efficiency so firms expand and refine global R&D networks. Moreover, host-country information accessibility and market entry level mediate this process: digital tools reduce information asymmetry and allow real-time monitoring of regulatory change, lowering the liability of foreignness and sustaining R&D internationalization.
Empirical results demonstrate that digital transformation significantly enhances the level of R&D internationalization, and this conclusion remains robust across a series of tests. Heterogeneity analysis further shows that the effect is more pronounced in state-owned enterprises and in firms with executives holding overseas backgrounds. Mechanism testing additionally confirms that digital transformation effectively improves host-country information accessibility and market entry level, which enables enterprises to realize the new OLI advantages, thereby alleviating information asymmetry and entry-barrier problems and ultimately advancing their R&D internationalization.
Building on these findings, this study proposes that,first, enterprises should accelerate digital transformation, strengthen their capability to embed in global R&D networks, deploy core digital technologies, and digitally reengineer R&D processes to enhance real-time insight into host-country markets, technologies, and talent information. Second, differentiated guidance should be implemented: state-owned enterprises should be encouraged to increase digital investment and closely link it to overseas R&D performance, while small and medium-sized enterprises should receive targeted "diagnosis-assistance" support to address specific challenges such as information asymmetry and shortages of international talent. Policy support should further incorporate the international experience and digital literacy of executive teams into its design.Third, host-country market access and information service guarantees should be reinforced, including building an authoritative and dynamic host-country R&D environment information database, promoting mutual recognition and institutional alignment in areas such as cross-border data flows and market entry, strengthening cooperation on intellectual property protection, and simplifying procedures for establishing overseas R&D institutions. In addition, building industry- or region-level digital R&D collaboration platforms can foster the sharing of innovation resources, reduce information asymmetry, and thus create a global innovation cooperation ecosystem that advances collaborative R&D.

Digital innovation is currently reshaping the economic system for innovative development at an unprecedented speed and scale. It has emerged as a pivotal engine for accelerating the high-quality development of the digital economy and for driving the evolution of new productive forces. However, constrained by inadequate digital resources and innovation capabilities, traditional enterprises are encumbered by such predicaments as ambiguous digital innovation pathways and ineffective realization of digital innovation value. As a result, they find themselves in a “well-intentioned but underpowered” dilemma when pursuing digital innovation.Studies indicate that integrating into digital platform ecosystems and engaging in dynamic interactions with multiple participants constitutes a critical pathway for enterprises to achieve digital innovation. Nevertheless, the existing literature has yet to uncover the specific digital-innovation pathways available to ecosystem-participating enterprises.
　Grounded in the complex adaptive systems theory, this study integrates micro-level cognitive factors, meso-level organizational capabilities, and macro-level ecosystem attributes to construct a theoretical analytical framework of "managerial cognition-organizational capabilities-ecosystem characteristics". Employing the fuzzy-set qualitative comparative analysis (fsQCA) method, it analyzes 263 sets of data to explore the configurational effects of managerial cognitive flexibility, digital perception capability, digital resource synergy capability, ecosystem richness, and ecosystem innovativeness on digital innovation among ecosystem-participating enterprises, while elucidating the underlying mechanisms and pathways of such innovation.
　The findings reveal that (1) no single core condition constitutes a necessary prerequisite for high-level digital innovation; however, managerial cognitive flexibility, as a core condition, exerts a relatively universal influence; (2) the coupling of diverse antecedent conditions forms two configurational pathways, corresponding to two types of configurations for achieving high-level corporate digital innovation: "cognition-opportunity driven" and "cognition-ecosystem interactive". Specifically, in the cognition-opportunity driven pathway, managerial cognitive flexibility, digital perception capability, and ecosystem innovativeness are present as core conditions, ecosystem richness is absent as a core condition, and digital resource synergy capability functions as a peripheral condition. In the cognition-ecosystem interactive pathway, managerial cognitive flexibility, digital resource synergy capability, ecosystem richness, and ecosystem innovativeness are present as core conditions, while digital perception capability serves as a peripheral condition; (3) three configurational pathways are identified, forming two types of configurations that lead to non-high-level digital innovation: "cognition-capability deficient" and "cognition deficient". The research conclusions remain valid after robustness tests.
This study makes three significant contributions. First, from the perspective of ecosystem participants, it identifies the key antecedents and operational mechanisms of digital innovation within participating enterprises, addressing the insufficient attention given to their digital innovation in existing research. Second, moving beyond the prevalent focus on linear relationships between isolated influencing factors and corporate digital innovation, this study draws on complex adaptive systems theory to innovatively construct a "managerial cognition-organizational capabilities-ecosystem characteristics" analytical framework. Using the fsQCA method, it explores the driving mechanisms of multi-factor linkage and synergy in promoting digital innovation among participating enterprises, thereby filling a gap in the literature regarding the unclear pathways of such innovation. Finally, the study innovatively incorporates the richness and innovativeness of digital platforms into the research framework, extending the analytical perspective to the ecosystem level and offering a new theoretical lens and framework for examining the relationship between platform ecosystem characteristics and corporate innovation capabilities.

This study addresses a persistent challenge in patent analytics and innovation management: how to identify potentially innovative inventions at an early stage, before ex post indicators such as forward citations, market transactions, or long-term legal outcomes become observable. In many real-world contexts, including research evaluation, technology strategy, and patent portfolio management, decision-makers must assess patent applications under conditions of high uncertainty and limited information. Conventional indicators of patent value or innovation often suffer from substantial time lags and limited interpretability, which constrains their usefulness for early decision making. To overcome these limitations, this study conceptualizes patent examination as an institutionalized expert evaluation process and exploits examiner citations, particularly their citation types, as early and structured signals of novelty assessment. Within the European Patent Office (EPO) system, grant outcomes are treated as a practical proxy for early-stage innovation.
An empirical dataset was constructed using IncoPat, focusing on EPO biomedical patents classified under IPC prefix A61 and published between 2005 and 2025. To ensure the reliability of supervised labels, only applications with final examination decisions and complete examiner citation records were retained. The resulting dataset contains 183 600 A-type citations, and mainly represent background references, and 139 132 X-type citations, and indicate strong novelty or inventive-step challenges. After patent identifier normalization, textual preprocessing of titles, abstracts, and first claims, and citation edge de-duplication, a large-scale patent knowledge graph was built in Neo4j, containing 364 471 patent nodes and 341 123 typed citation edges. The graph exhibits sparsity and heavy-tailed degree distributions, motivating the use of relational graph learning. To avoid information leakage, a subgraph-level data split was adopted so that each focal patent application appears in only one of the training, validation, or test sets, which were divided in a 7:1:2 ratio.
This study proposes an A/X citation-guided multi-task learning framework (AXMLM) that jointly optimizes two interrelated tasks. The first task is examiner citation relation prediction, which infers whether a pair of patents is connected by an A-type or X-type citation. The second task is innovation prediction, operationalized as the probability that a focal patent application will be granted. Patent text is encoded using SciBERT to capture domain-specific technical semantics, while structural information is modeled through a Relational Graph Convolutional Network (R-GCN) that explicitly distinguishes citation types during message passing. In addition to parameter sharing across tasks, a citation-aware dynamic adjustment strategy is introduced, which modifies innovation scores based on the relative prevalence of A-type and X-type citations. This design aligns model outputs with examination logic and improves interpretability by making the influence of examiner signals explicit.
Experimental results demonstrate the effectiveness of integrating semantic and relational information. For citation relation prediction, the SciBERT-based R-GCN model achieves an accuracy of 0.754 5 and an F1 score of 0.744 4, outperforming text-only baselines and a prompt-based large language model baseline. For innovation prediction, AXMLM attains an accuracy of 0.769 5 and an F1 score of 0.748 5, with particularly high recall, indicating strong capability for early screening of potentially innovative applications. Sensitivity analysis on task-weight coefficients shows that emphasizing the citation prediction task further improves innovation prediction performance, supporting the hypothesis that examiner citation types carry meaningful early evaluation signals.Ablation studies confirm the importance of key design components. Replacing R-GCN with a standard GCN leads to a substantial performance drop, highlighting the necessity of modeling relation types. Removing the graph component reduces performance to the level of text-only models, while disabling the dynamic adjustment strategy results in a moderate decline, suggesting that the strategy enhances robustness and interpretability. Additional analyses reveal heterogeneity in citation patterns and grant rates across examiners, IPC subclasses, inventors, and applicants. Robustness checks indicate that discrepancies between predicted and observed grant rates are mainly attributable to small-sample effects rather than systematic bias.
Overall, this study contributes an interpretable and decision-oriented framework for early-stage patent innovation assessment by integrating examiner citation types, domain-specific semantics, and relational learning. The proposed approach provides practical support for early patent screening, portfolio prioritization, and innovation management, and offers a foundation for future extensions to other technological domains, languages, and patent systems.

Global climate change and energy crises present urgent and complex challenges to human society, driving the need for innovative solutions that can enhance sustainability and energy security. In recent years, the integration of digital technologies to facilitate the transition toward smart energy systems has become a focal point of research and policy initiatives. This exploration aims to leverage the capabilities of digital technologies to optimize energy systems, improve operational efficiencies, and support sustainable development. Technological convergence is central to this transformation, defined as the process of integrating elements from different technological domains to create new functionalities and value. In the digital age, the convergence of digital technology with energy technology is particularly significant as it promotes novel solutions critical for smart energy transitions.
Collaboration plays a crucial role in enabling technological convergence. The rapid pace of technological evolution and the increasing complexity of interdisciplinary challenges make partnerships with external organizations an essential strategy. Establishing collaborative networks helps overcome disciplinary boundaries, promotes the exchange of knowledge, and accelerates the convergence process. However, despite the importance of collaboration in driving convergence, the existing literature on digital and wind power technology integration, especially within the context of China, remains sparse. Although policies supporting this integration have been implemented, comprehensive studies from a patent-based perspective that detail the convergence trends and collaborative dynamics are limited. Furthermore, research has often focused primarily on technological elements, overlooking the role of collaborative actors and the evolution of their relationships. This oversight may hinder a comprehensive understanding of the mechanisms that drive convergence.
To address these research gaps, this study aims to utilize patent co-classification data to explore the convergence landscape of digital and wind power technologies in China and to uncover the dynamic evolution of the collaboration networks underpinning this process. By focusing on collaborative patent data from 2005 to 2022, sourced from the IncoPat database, the study applies social network analysis to construct and analyze technological convergence networks, organizational collaboration networks, and two-layer networks that combine both dimensions.
The findings reveal several key insights: (1) The technological convergence process exhibits distinct evolutionary stages, marked by an accelerating pace over time. This is evidenced by the increasingly prominent small-world characteristics within the convergence network, suggesting enhanced connectivity and efficiency in knowledge flow across technological domains. The core technological areas of convergence have shifted over the study period, evolving from traditional wind turbine technologies to power supply, distribution, and energy storage technologies, and more recently, to digital data processing technologies. This progression highlights the growing role of digital solutions in the wind power sector, underscoring their importance for future innovation. (2) The organizational collaboration network has shown increasingly evident scale-free properties, indicating that a few dominant organizations account for the majority of collaborative activities. State-owned enterprises, such as the State Grid Corporation of China, have emerged as pivotal actors within this network, forming extensive internal collaboration networks. However, these networks demonstrate limited external openness, which may restrict broader knowledge transfer and the incorporation of diverse perspectives. (3) The expansion of the two-layer network has been accompanied by an increase in the technological breadth managed by organizations. In recent years, technology clusters centered on G06 (computing,calculating or counting) have gained prominence as hotspots for inter-organizational technological cooperation.
This research makes several contributions to the existing literature. By systematically revealing the trends of digital and wind power technology convergence through patent network analysis, it provides empirical insights that fill a significant research gap. The study also enhances the understanding of how collaborative relationships evolve over time, offering a comprehensive view of the interaction between technological and organizational factors. These insights are useful for policymakers, industry leaders and researchers, as they provide practical guidance for strategic planning and management practices aimed at fostering deeper integration of digital and renewable energy technologies. Ultimately, the findings support the development of targeted strategies to strengthen collaboration, encourage innovation, and facilitate the transition to more sustainable and efficient energy systems in China.

Family firms, as a vital component of China's private sector, play a significant role in driving economic growth, contributing tax revenue, alleviating employment pressure, and have become a key force within the national innovation system. The presence of multiple heirs may exert a complex influence on the innovation investment of family firms. On the one hand, the multiple-heir scenario can act as a restraining factor for corporate innovation. On the other hand, it may also present opportunities for promoting corporate innovation, as healthy competition among heirs could stimulate a stronger willingness to innovate. This inherent paradox between inhibitory effects and promotional opportunities leaves the impact of multiple heirs on family firms' innovation investment in a theoretical gray area. It is neither possible to simply conclude that it harms innovation nor to directly affirm its positive role, necessitating large-sample empirical research to clarify the true relationship between the two. However, existing studies have not sufficiently revealed the specific impact or the underlying mechanisms through which multiple heirs affect corporate innovation investment.
Building on this background and the need for theoretical clarification, this study examines the fundamental impact and mechanisms of multiple heirs on innovation investment by utilizing a sample of A-share listed family firms in China from 2004 to 2023, integrating Resource-Based Theory, Socioemotional Wealth Theory, and Agency Theory. After matching data on the number of heirs with corporate innovation investment records, and after setting variables and constructing a panel fixed-effects model, this paper empirically tests how the presence of multiple heirs influences corporate innovation investment and identifies its transmission channels. Furthermore, by incorporating multi-dimensional contextual factors—such as internal governance structures, industry attributes, and regional characteristics—this study reveals how the relationship varies across different internal and external contexts through heterogeneity analysis. Finally, focusing on family firms that have completed intergenerational succession, the paper investigates the long-term effects of having had multiple heirs on innovation persistence and innovation efficiency.
The findings demonstrate that the presence of multiple heirs significantly inhibits innovation investment in family firms, and this conclusion remains robust after a series of rigorous robustness checks. Mechanism tests reveal that this effect operates primarily through channels such as exacerbating financing constraints, reducing the firm's willingness to take risks, and weakening managerial incentives. Heterogeneity analysis indicates that the inhibitory effect of multiple heirs on innovation investment is more pronounced under the following conditions: deeper family involvement in management, lower industry competition intensity, stronger Confucian cultural influence, or the presence of non-statutory heirs. Further analysis shows that among family firms that have completed intergenerational succession and originally had multiple heirs, the successors tend to increase both the level and persistence of innovation investment, yet they suppress improvements in innovation efficiency.
Compared with prior literature, this study contributes in the following ways. First, it provides new empirical evidence for understanding the determinants of innovation investment in family firms. By focusing on the often-overlooked scenario of multiple heirs and analyzing how internal tensions and constructive competition among them shape innovation decisions, this research enriches the theoretical dimensions of family firm innovation and extends the existing analytical framework. Second, it broadens the perspective on intergenerational succession in the context of family firms with multiple heirs. Centering on the pre-succession stage—when multiple heirs coexist—this study examines its impact on innovation investment and uncovers the underlying mechanisms, thereby offering incremental insights into the succession-innovation nexus. Third, it carries important practical and policy implications. For policymakers, the findings inform the design of more effective measures to sustain the innovation momentum of family firms. For business families, the results highlight the unique challenges posed by multi-heir succession, enabling them to proactively formulate succession strategies and innovation roadmaps, ultimately supporting the long-term sustainability and prosperity of their enterprises.

In the era of the digital-intelligence paradigm, the deep integration of data, algorithms, and computing power is propelling human society into a new stage of development. As a strategic frontier technology reshaping the global economic landscape, artificial intelligence (AI) is driving a new round of scientific and industrial revolutions. In response to this trend, China has placed technological innovation at the core of its national development strategy, focusing on cultivating future industries represented by AI to foster new quality productive forces and achieve high-quality growth. As core actors of these future industries, AI enterprises are characterized by cross-disciplinary integration, rapid iteration, and collective innovation. However, beneath this technological dynamism lies a persistent challenge: many enterprises demonstrate strong technological exploration capability but struggle to convert scientific achievements into sustainable market value. This structural disconnection between technology exploration and commercial transformation highlights the need to establish an efficient linkage across the entire innovation chain of knowledge-technology-market.
Compared with traditional factor-driven industries, AI enterprises rely more heavily on intelligent factors that are fluid, self-evolving, and highly uncertain. These characteristics make innovation activities more complex and riskier, and render innovation performance increasingly dependent on entrepreneurs′ capacity to judge technological trajectories and identify market opportunities. In this context, entrepreneurial spirit is no longer merely an individual trait but serves as a vital coordinating mechanism that bridges strategic vision and technological execution, directly influencing the efficiency of innovation resource allocation and innovation output. Nevertheless, existing research has primarily focused on external resources or organizational structures, paying insufficient attention to the internal cognitive and behavioral mechanisms through which entrepreneurial spirit shapes innovation performance. Particularly within the dynamic and complex technological environment of AI enterprises, the underlying mechanism remains insufficiently studied and theoretically underdeveloped.
To better understand this issue, from a science-entrepreneurship ambidextrous perspective, this study adopts the innovation value chain theory as its analytical framework and constructs a two-stage model comprising knowledge R&D and technology valorization. The research sample consists of AI enterprises listed on China′s Science and Technology Innovation Board (STAR Market), which represent the frontier of the nation′s future industries. First, an LDA topic model is employed to identify four dimensions of entrepreneurial spirit—exploration, responsibility, resilience, and integration—which are further categorized into two subsystems: scientific-attribute spirit and entrepreneurial-attribute spirit. These dimensions are quantitatively measured through text analysis combined with a Word2Vec model. Second, a two-stage network Data Envelopment Analysis (DEA) model is used to evaluate the efficiency of each stage in the innovation value chain and the overall performance formed by their coupling. Finally, multiple regression analysis is conducted to test the effects of entrepreneurial spirit on stage-specific innovation efficiency and overall innovation performance.
The results reveal three main findings: (1) Scientific-attribute spirit primarily influences the knowledge R&D stage, where exploration spirit significantly enhances R&D efficiency, and responsibility spirit exhibits an inverted “U-shaped” pattern of “first inhibiting, then promoting”; (2) Entrepreneurial-attribute spirit mainly affects the technology valorization stage, where resilience spirit positively promotes commercialization efficiency, while integration spirit demonstrates an inverted-U effect; (3) The synergy between scientific and entrepreneurial spirits significantly enhances the overall performance of the innovation value chain, with the effect being stronger among entrepreneurs possessing research backgrounds.
Drawing on the innovation value chain perspective, and grounded in a science-entrepreneurship ambidextrous framework, this study systematically explores the mechanism through which entrepreneurial spirit drives innovation in AI enterprises. The framework reveals both the stage-specific and synergistic effects of multidimensional entrepreneurial spirit, and clarifies the mechanisms and pathways through which knowledge R&D and technology transformation achieve coordination. The findings refine the process perspective of the innovation value chain, expand the theoretical boundary of entrepreneurial spirit in hard-technology contexts, and provide valuable insights for promoting the sustainable innovation of China′s future industries.