Enterprise Social Media Use and Employee Innovation: The Role of Employee Capital and Empowering Leadership

Abstract

This study investigates the relationship between employees’ task-oriented and social-oriented use of enterprise social media (ESM) and their innovation performance through the accumulation of employees’ social, human, and psychological capital. Integrating Self-Determination Theory and Social Learning Theory, we propose a multiple-mediation framework in which ESM serves as a resource-building infrastructure that supports innovation indirectly by strengthening employee capital. We test the model using survey data from 613 employees in Chinese knowledge-intensive enterprises. Results show that both ESM use orientations are positively associated with all three forms of capital; however, neither orientation has a significant direct effect on innovative performance once the capitals are included. Instead, the ESM–innovation link is transmitted through these capitals, indicating an indirect-only mediation pattern. We further find that empowering leadership amplifies the extent to which human and psychological capital translate into innovative performance, whereas its moderation on the social capital–innovation relationship is comparatively weak. Overall, the findings position ESM as a digital infrastructure that enables a multi-capital pathway to employee innovation in contemporary work settings.

1. Introduction

Enterprise social media (ESM) has gradually become an indispensable digital infrastructure for internal communication and knowledge collaboration in contemporary organizations (Leonardi, 2015; Mäntymäki & Riemer, 2016; Zhou & Yin, 2025). Its affordances mainly lie in visibility, persistence, and efficient and immediate interaction. These features can effectively reduce the cost of information search, facilitate the precise location of professional knowledge, and promote cross-departmental and cross-boundary coordination and cooperation (Jia et al., 2022; Leonardi & Treem, 2012; Schötteler et al., 2023).

Prior ESM research has evolved along three interrelated streams. One stream emphasizes the affordance-based value of ESM, showing how visibility, persistence, and efficient, immediate interactivity facilitate knowledge sharing, expertise location, and collaboration across organizational boundaries (Pitafi et al., 2023; Y. Sun et al., 2023; Treem & Leonardi, 2013). A second stream links ESM use to employee creativity, innovative behavior, and work performance, suggesting that digital interaction can expand employees’ access to ideas and problem-solving resources (Kuegler et al., 2015; Kumar & Rani, 2024; Rasheed et al., 2023). A third stream, however, points to the darker side of ESM use, including information overload, distraction, work interruption, and cyberslacking. Taken together, these studies suggest that ESM is not uniformly beneficial or harmful (Nusrat et al., 2024; Y. Shang et al., 2023; Si et al., 2023). Rather, its consequences depend on how employees use it, what resources they develop through it, and whether the organizational context allows these resources to be translated into innovation.

Three limitations remain in the existing literature. These limitations are especially relevant in Chinese knowledge-intensive enterprises, where employees’ willingness to speak up, experiment, and mobilize workplace resources may be shaped by hierarchical norms and power-distance expectations. First, although ESM use is increasingly recognized as a multidimensional behavior, many studies still measure it as a general frequency or intensity of use. This approach obscures the fact that employees may use the same platform for different purposes. Consistent with the motivational lens of Self-Determination Theory, ESM use can be divided into task-oriented use (TESM), which focuses on work coordination and problem solving, and social-oriented use (SESM), which focuses on relationship maintenance and emotional communication (Chen & Wei, 2020; Zhu et al., 2024). This distinction matters because task- and social-oriented use may develop different resources, such as knowledge and skills on the one hand, and relational support or psychological encouragement on the other. This is particularly salient in Chinese enterprises, where instrumental task exchange and affective relationship building are deeply embedded in workplace interactions. Therefore, distinguishing TESM and SESM helps explain why ESM may support innovation in some conditions but not in others.

Beyond the oversimplified conceptualization of ESM use, a second key limitation is that previous studies have often examined single mediating mechanisms, such as knowledge sharing or social capital, but have rarely empirically tested the parallel mediating roles of multiple forms of employee capital within a unified analytical framework. From a resource-based perspective, social, human, and psychological capital can be understood as valuable intangible employee resources that support innovative performance. In this study, ESM is not treated as the focal strategic resource itself; rather, it is conceptualized as a digital context through which employees develop and mobilize these resources. Specifically, social capital (SOCAP) reflects access to relational resources embedded in workplace networks (Nahapiet & Ghoshal, 1998), human capital (HUCAP) captures employees’ knowledge and skills relevant to problem solving and creativity (Nonaka & Peltokorpi, 2006), and psychological capital (PSYCAP) represents positive psychological resources (e.g., efficacy, hope, optimism, resilience) that sustain effort and adaptability under uncertainty (Luthans et al., 2010). By bringing these mechanisms together in a multiple-mediation model and testing it in Chinese knowledge-intensive enterprises, we provide a more fine-grained account of the heterogeneous innovation outcomes associated with differentiated ESM use patterns.

Finally, even when differentiated ESM use and multi-dimensional mediating mechanisms are accounted for, extant research has largely overlooked the leadership conditions under which resources accumulated through ESM are effectively translated into innovative performance (INNO). Importantly, resource accumulation does not automatically translate into innovative performance. Particularly in hierarchical settings, speaking up or experimenting can be socially risky. We therefore consider empowering leadership (EL), defined as leader behaviors that delegate authority, encourage participation, and strengthen employees’ confidence, as a contextual condition under which accumulated resources are more likely to be mobilized and deployed. Empowering leaders grants autonomy, invites participation in decision making, and expresses confidence in employees’ capabilities, which can provide a “permission structure” for employees to mobilize what they have gained from ESM into concrete innovative actions (Ahearne et al., 2005; Amundsen & Martinsen, 2015; Kim et al., 2018). This boundary role is especially relevant in Chinese knowledge-intensive enterprises, where power distance norms may discourage proactive resource use and make empowerment function as a potential “cultural neutralizer” that legitimizes initiative and reduces perceived interpersonal risk (Y. Dai et al., 2022; Kwan et al., 2025).

Theoretically, this study integrates Self-Determination Theory (SDT) (Ryan & Deci, 2000), Social Learning Theory (SLT) (Bandura, 1977), and a complementary resource-based logic (Barney, 1991). Self-Determination Theory helps explain why employees engage in ESM use when digital interaction satisfies their needs for autonomy, competence, and relatedness. Social Learning Theory explains how employees learn from visible knowledge, peer interaction, and shared work practices on ESM. A resource-based logic further explains why the social, human, and psychological capital developed through ESM can serve as valuable employee resources for innovative performance. Together, these perspectives provide the theoretical basis for linking dual-oriented ESM use to innovative performance through employee capital and for examining EL as a condition that strengthens the conversion of employee capital into innovation.

This study makes three contributions to the ESM and employee innovation literature. First, it extends prior ESM research by distinguishing task-oriented and social-oriented ESM use rather than treating ESM use as a single undifferentiated behavior. This distinction helps explain why the same digital platform may generate different innovation-related outcomes across employees and contexts. Second, this study advances the mechanism-based understanding of ESM by integrating social, human, and psychological capital into a multiple-mediation model. By doing so, it shows how different forms of employee capital jointly explain the relationship between ESM use and innovative performance. Third, this study contributes to the leadership and digital work literature by identifying empowering leadership as a boundary condition for the capital-to-innovation relationship. This finding clarifies that employee resources accumulated through ESM are more likely to support innovative performance when leaders provide autonomy, participation, and confidence.

The remainder of this paper is organized as follows. Section 2 reviews the theoretical background and develops the hypotheses. Section 3 describes the research design, sample, measures, and analytical procedures. Section 4 presents the empirical results. Section 5 discusses the theoretical and practical implications of the findings. Section 6 concludes the paper by summarizing the main contributions, limitations, and directions for future research.

2. Theoretical Framework

2.1. Core Constructs

2.1.1. Enterprise Social Media

Enterprise social media (ESM) refers to organizationally embedded social technologies that allow employees to communicate, coordinate, and share both work-related and personal information within the firm ((Leonardi et al., 2013). Compared with public social media, ESM is designed for internal collaboration and has typical characteristics of visibility, persistence, and rapid interaction across functions and locations (Leonardi & Treem, 2012). These affordances make employees’ activities, expertise, and interactions more observable and retrievable, which can lower search costs and help employees locate “who knows what” when tasks require distributed expertise (Jia et al., 2022; Leonardi, 2015; Schötteler et al., 2023).

Empirical findings, however, are not uniformly positive. While many studies link ESM to improved communication quality, knowledge sharing, and coordination (Ellison et al., 2014; Kwahk & Park, 2016), others point to overload, conflict, work–life boundary blurring, and cyberslacking that may offset or even reverse performance benefits (Nusrat et al., 2023; Y. Sun et al., 2021). The above differences indicate that for the organization and its employees, the core issue is no longer “whether to adopt ESM”, but “how to effectively utilize this system”.

A motivation-based perspective, therefore, distinguishes task-oriented and social-oriented ESM use (Chen & Wei, 2020). Task-oriented ESM use (TESM) focuses on acquiring work information, coordinating tasks, and sharing documents to support problem solving and task accomplishment. Social-oriented ESM use (SESM) involves maintaining relationships, informal messaging, and socioemotional support that can cultivate trust and cohesion (Zhu et al., 2024). Treating these orientations separately provides a parsimonious way to explain heterogeneous outcomes and to theorize distinct resource-building pathways from ESM use to innovation.

2.1.2. Employees’ Three Capitals

To explain how ESM use becomes consequential for innovation, we focus on three employee-level resources: social capital, human capital, and psychological capital, which capture relational access, competence, and positive psychological resources, respectively. Although conceptually distinct, these capitals often co-develop in knowledge-intensive work and can jointly support idea generation, promotion, and implementation.

Social capital (SOCAP) refers to the actual and potential resources embedded in social relationships, including information, trust, reciprocity, and shared norms (Bourdieu, 1986; Coleman, 1988; Nahapiet & Ghoshal, 1998). At the employee level, it reflects the relational resources that employees can access and mobilize through workplace ties. Nahapiet and Ghoshal (1998) framework describes SOCAP as three dimensions with structural (connection patterns), relational (trust/reciprocity), and cognitive (shared language/vision). In ESM contexts, the visibility and persistence of interactions can amplify relational signals, shared awareness, and common narratives, which makes social capital a reasonable mechanism linking digital interaction to innovation ((Masood et al., 2023; Pekkala & van Zoonen, 2022; C. Wang et al., 2022).

Human capital (HUCAP) refers to the knowledge, skills, abilities, and other personal qualities that enable individuals to create value (Becker, 1964; Ployhart & Moliterno, 2011). Beyond formal documents or certificates, tacit human capital like problem-solving, creative thinking, and adaptive expertise, are critical in knowledge-intensive work (Nonaka & Takeuchi, 1996). These skills have the characteristics including value, scarcity and difficulty in complete transfer (Ployhart & Moliterno, 2011). HUCAP is an employee-embedded competence resource that supports idea generation and implementation. In ESM contexts, task discussions, seeking feedback and cross-functional contacts can drive the acquisition, refinement and reorganization of knowledge (Moqbel et al., 2020; R.-A. Shang & Sun, 2021).

Psychological capital (PSYCAP) is not a personality trait, but rather a set of malleable psychological resources—self-efficacy, hope, optimism and resilience (Luthans et al., 2007). It determines whether employees are willing to take on challenging tasks, remain committed, and recover from setbacks. Innovation often occurs in uncertain circumstances, and PSYCAP can be the fuel that ignites the internal drive (Abbas & Raja, 2015; Avey et al., 2011; Grover et al., 2018). We believe that the value of ESM goes beyond information transmission: peer recognition and the provision of a sense of belonging are important catalysts for activating these psychological resources (Nambisan & Baron, 2021; Peng et al., 2018).

2.1.3. Innovative Performance (INNO)

We define INNO as the overall ability of employees to develop new ideas from their inception to achieving practical results, rather than merely the quantity of creative outputs. This definition aims to emphasize that the difficulty of innovation lies not in the scarcity of inspiration, but in the fact that an idea must undergo three tests—its infancy at the time of conception, the doubts during its promotion, and the resistance upon its implementation (Janssen, 2000; Scott & Bruce, 1994). Therefore, when evaluating an employee’s INNO, we should not only count the number of suggestions they have made but also see if they can complete the entire cycle of idea generation—promotion—implementation. To complete this cycle, three completely different types of capital reserves are required: professional knowledge for refining the idea (C.-Y. Lin & Huang, 2021; Soomro & Soomro, 2024), psychological resilience for sustaining motivation and resilience in the face of setbacks (Liang et al., 2024; Luthans et al., 2007), and social relationships for finding allies (Bhatti et al., 2021). Our core argument is: The true potential of ESM lies precisely in its ability to simultaneously supply these three types of capital—by making knowledge more transparent, feedback more timely, and cross-disciplinary minds more likely to meet.

2.1.4. Empowering Leadership (EL)

EL refers to leader behaviors that enhance employees’ autonomy, participation in decision making, perceived competence, and a sense of work meaningfulness (Ahearne et al., 2005; Amundsen & Martinsen, 2015). Although it is related to transformational and distributed leadership, it has a more specific focus. Transformational leadership emphasizes inspiration, vision, and individualized support, whereas distributed leadership concerns the sharing of leadership roles across organizational members (Khan, 2023). By contrast, EL centers on delegating authority, encouraging participation, and strengthening employees’ confidence to act independently. Therefore, this study treats EL as a leadership behavior and contextual condition, rather than as psychological empowerment or empowerment as an individual state.

Employees may hesitate to innovate not because they lack ability, but because they are uncertain about the interpersonal risks of trying. In some workplaces, experimentation is costly and speaking up can be interpreted as stepping outside one’s role. Under these conditions, employees may withhold suggestions even when they have the expertise to contribute, simply because they are unsure how others will react. Over time, this “strategic silence” can become a practical barrier to innovation (Morrison, 2023).

This concern can be stronger in high power-distance settings, which are common in many Chinese organizations (Hofstede, 2011; Jing et al., 2022). When hierarchy is highly visible in everyday interactions, employees may become cautious about how their ideas will be received, for example, whether raising a concern will be seen as challenging a superior, or whether a failed trial will lead to blame. In such contexts, EL can help offset hierarchy-related concerns about speaking up and experimenting by signaling that proactive behavior is legitimate and supported (Chiang & Chen, 2021). However, the effects of empowerment are not guaranteed. Prior work suggests that empowerment may be less effective when employees strongly internalize hierarchical values (i.e., high power distance orientation), because they may be less ready to accept autonomy even when it is offered (Y. Dai et al., 2022; M. Sun et al., 2025).

Consequently, this study treats EL as a moderating condition rather than a direct antecedent. While prior research has linked empowerment to outcomes such as social relationships, learning behaviors, or innovation-related processes, these pathways are often examined separately (Fu et al., 2019; Pitafi, 2024; R.-A. Shang & Sun, 2021; Zhu et al., 2024). It therefore remains unclear how EL differentially facilitates the translation of SOCAP, HUCAP and PSYCAP into INNO.

2.2. Theories

2.2.1. Self-Determination Theory

Self-Determination Theory (SDT) suggests that the satisfaction of basic psychological needs such as autonomy, competence, and a sense of belonging by individuals are the key internal driving forces that motivate their autonomous motivation and positive work behavior (Deci & Ryan, 1985; Ryan & Deci, 2000). When these needs are supported, employees tend to show stronger engagement, more active learning, and better performance (Baard et al., 2004; Deci et al., 2017). SDT regards competence and belonging as the core sources of internal motivation. TESM continuously provides employees with a sense of competence by enhancing problem-solving efficiency and work transparency; SESM consistently offers a sense of belonging through interpersonal interaction and emotional communication (Chen & Wei, 2020). Thus, ESM is no longer merely an information channel but has become a dual space within the organization that combines both task functions and psychological functions. The satisfaction of these needs subsequently maintains employees’ willingness to participate and drives them to continuously engage in subsequent behaviors that contribute to resource accumulation.

2.2.2. Social Learning Theory

Motivation alone does not become capability without behavioral and social-cognitive mechanisms. Social Learning Theory (SLT) explains how individuals acquire behaviors and skills through observational learning, imitation, reinforcement, and self-efficacy processes (Bandura, 1977; Bandura, 1986). ESM is well suited for these mechanisms because interactions and work traces are visible and persistent. Employees can observe how others solve problems, communicate, and coordinate; imitate effective practices; and receive rapid feedback or recognition that reinforces participation (Majchrzak et al., 2013; Qi & Chau, 2018; Vuori & Okkonen, 2012). Over time, such learning cycles can build three key resources: learning and feedback strengthen HUCAP, interaction history and reciprocity strengthen SOCAP, and accumulated successful experiences enhance PSYCAP (Leonardi & Treem, 2012).

2.2.3. Integrating Theories

To explain how ESM becomes a resource engine, it is useful to separate willingness from mechanics. SDT clarifies why employees are motivated to engage—ESM affordances can help satisfy psychological needs and sustain participation. SLT specifies how participation turns into resources—through observation, reinforcement, and efficacy-building processes that translate engagement into accumulated social, human, and PSYCAP (Bandura, 1977; Bandura, 1986). Together, the two theories position ESM as both a motivational context that supports continued contribution and also a learning platform that makes resource accumulation more systematic and scalable.

A resource-based logic further explains why the forms of capital developed through ESM are relevant to innovative performance (Barney, 1991). In this study, RBV is not used to conceptualize ESM itself as the focal strategic resource. Rather, ESM is treated as a digital work context through which employees may develop and mobilize valuable resources. The focal resources are social capital, human capital, and psychological capital. Social capital provides access to relational ties, trust, and shared understanding; human capital reflects employees’ knowledge, skills, and problem-solving capabilities; and psychological capital captures positive psychological resources such as efficacy, hope, optimism, and resilience. These forms of employee capital support innovation because they help employees identify opportunities, combine knowledge, persist under uncertainty, and gain support for novel ideas. Thus, RBV complements SDT and SLT by explaining why the resources accumulated through ESM can serve as a basis for innovative performance.

2.3. Hypotheses Development

2.3.1. ESM Use and Employee Capitals (H1–H6)

ESM provides enterprises and the opportunity to have an infrastructure that supports internal communication, collaboration, and knowledge exchange by making work traces more visible and easier to reuse over time (Liang et al., 2021). Distinguishing task-oriented and social-oriented use (Pekkala & van Zoonen, 2022; Sæbø et al., 2020), this study argues that ESM used in both orientations can enhance employees’ SOCAP, HUCAP and PSYCAP.

With respect to SOCAP, TESM can enhance instrumental connections by making expertise, requests, and contributions more accessible to search, and by simplifying cross-functional coordination across formal boundaries (Suh & Bock, 2015; C. Wang & Cardon, 2019). An important instrumental attribute of ESM is the display of “who knows what” and “who knows who”. The history of communication and cooperation is typically continuously visible on the ESM platform, allowing employees to find reliable partners and continuously strengthen cooperation (Luqman et al., 2021). Compared to the more formal or assignment-related interactions of TESM, SESM is capable of cultivating relationship resources through informal communication and emotional interaction. These exchanges and interactions convey kindness and enhance trust, thereby activating weak ties and expanding the channels for obtaining support (Kamboj et al., 2017; Leonardi, 2015; Pekkala & van Zoonen, 2022).

The continuous accumulation of HUCAP requires two types of support: reusable experience and exchangeable perspectives. TESM makes implicit experience explicit and disorganized knowledge structured by recording problem-solving and feedback iterations, thereby forming retrievable organizational memory (Leyer et al., 2019). SESM introduces heterogeneity of cognition through informal interactions, preventing human capital from falling into path dependence (Lu & Pan, 2019). In the organizational context of China, this mechanism is highly dependent on interpersonal trust. Structural barriers that are difficult for formal processes to penetrate are often overcome through informal channels carried by “Guanxi”; these latter not only reduce the cost of seeking help but also provide a space for the circulation of intangible practical wisdom (J. Liu & Yan, 2021; Yu, 2014).

Finally, ESM use can improve PSYCAP by shaping employees’ motivational and affective states. Experiences of task accomplishment, recognition, and collective problem solving on ESM can strengthen efficacy and hope and help employees cope with challenges (Leonardi, 2015; Luthans et al., 2007; Nambisan & Baron, 2021). SESM can provide emotional support and belongingness that gradually builds optimism and resilience (Men et al., 2020; Nusrat et al., 2024). However, these psychological dividends are not linear; chronic cognitive overload or dysfunctional interpersonal exchanges can attenuate the positive effects on PSYCAP. This suggests that the accumulation of psychological resources is governed more by the relational valence and perceived manageability of the ESM environment than by quantitative usage metrics. Consistent with SLT and SDT, we therefore expect both orientations to contribute to the accumulation of employee capital.

Based on these arguments, this study proposes the following hypotheses:

H1. 

TESM positively affects employees’ SOCAP.

H2. 

TESM positively affects employees’ HUCAP.

H3. 

TESM positively affects employees’ PSYCAP.

H4. 

SESM positively affects employees’ SOCAP.

H5. 

SESM positively affects employees’ HUCAP.

H6. 

SESM positively affects employees’ PSYCAP.

2.3.2. Employee Capital and Employees’ Innovative Performance (H7–H9)

INNO requires both capability and motivation to generate, promote, and implement new ideas. SOCAP, HUCAP and PSYCAP capture complementary resources that can support these stages in different ways. Rather than assuming any single resource is sufficient, we argue that each capital contributes a distinct input that can enable innovation in knowledge-intensive work.

SOCAP can facilitate innovation by expanding access to diverse information, reducing coordination costs, and enabling cross-boundary collaboration—conditions that support idea recombination and the mobilization of support for implementation (Ganguly et al., 2019; Ko & Choi, 2019; C. Wang & Cardon, 2019). Trust and reciprocity further increase willingness to share tentative ideas and provide feedback, which can improve both the novelty and feasibility of innovations (Bhatti et al., 2021; Leonardi, 2015). A boundary condition is that not all networks are equally useful: dense, homogeneous ties may reinforce conformity, whereas more diverse ties are more likely to provide nonredundant knowledge, making the quality and diversity of SOCAP particularly consequential.

HUCAP can provide employees with the knowledge base and problem-solving skills needed to identify opportunities, integrate insights, and transform ideas into feasible solutions (Rivaldo & Nabella, 2023). Especially when the innovation process is filled with uncertainties, requires repeated trial and error, and is highly dependent on professional judgment, the importance of such knowledge and ability becomes even more prominent (Huang et al., 2015; Moqbel et al., 2020; Ren & Song, 2024). However, whether HUCAP can be effectively transformed into innovative performance largely depends on whether employees have sufficient autonomy. In an organization lacking autonomy and with rigid processes, even if employees have a high level of knowledge and ability, their innovative behaviors are difficult to fully exert.

PSYCAP helps employees keep going when early attempts do not work. With higher optimism and resilience, employees are more willing to search for alternative solutions after setbacks (Baig et al., 2021; Luthans et al., 2007). This may be especially relevant in digital task settings. Work is often more visible, feedback comes faster, and employees can learn quickly, but they may also feel more exposed to evaluation (Liang et al., 2024). PSYCAP may not translate into innovation if employees think that speaking up or trying new methods could lead to interpersonal trouble. In this case, a climate that supports exploration and treats reasonable trial and error as acceptable is more likely to allow PSYCAP to show its value.

Thus, this study proposes:

H7. 

Employees’ SOCAP positively affects INNO.

H8. 

Employees’ HUCAP positively affects INNO.

H9. 

Employees’ PSYCAP positively affects INNO.

2.3.3. ESM Use and INNO (H10–H11)

Although this study focuses on the capital-building pathway, ESM may also relate to innovative performance through day-to-day coordination processes. In routine work, ESM can make emerging problems visible (e.g., someone posts a bottleneck, a workaround, or a request for input), and colleagues can respond quickly with comments, edits, or follow-up questions. These exchanges may help teams align interpretations and refine partial solutions before they become formal proposals (Leonardi, 2015; Nambisan & Baron, 2021). At the same time, such immediate process benefits may be relatively modest once the broader resource channels (SOCAP, HUCAP, and PSYCAP) are taken into account, because many of these “quick-iteration” advantages operate by gradually building those same resources.

For TESM, a potential direct link to innovation comes from its ability to organize attention around concrete tasks. Compared with fragmented email chains, traceable task threads and searchable records of prior decisions can reduce coordination frictions and shorten problem-solving cycles (Sæbø et al., 2020; Wei et al., 2022). TESM can also provide shared reference points across units, which helps keep revisions, feedback, and next steps aligned as ideas move toward implementation (Ganguly et al., 2019; García-Sánchez et al., 2017).

SESM may support innovation through a different, more informal route. Casual exchanges can trigger quick feedback and lightweight sensemaking that helps employees adjust or reframe early ideas while they are still tentative. Because these interactions are typically lower-stakes and less formal, they may make it easier to voice unconventional thoughts at an early stage (Ma et al., 2021). In this way, SESM may contribute to observable innovative behaviors even before substantial resources are mobilized. This leads to the following hypotheses:

H10. 

TESM positively affects INNO.

H11. 

SESM positively affects INNO.

2.3.4. Moderating Effects of Empowering Leadership (H12a–H12c)

In our framework, EL is not treated as another “the more, the better” factor in the innovation process. What we want to emphasize is that EL can shape whether employees are able to use the resources they have already built up and turn them into visible innovative outcomes (Burhan & Khan, 2024). The reason is straightforward: even when employees have accumulated strong SOCAP, HUCAP and PSYCAP, they may still face interpersonal and “organizational politics” risks when they raise new ideas and try to put them into practice. In high power-distance settings (which are common in many Chinese workplaces), employees are often used to a clear hierarchy, and these concerns may be even stronger (Guo et al., 2022).

For this reason, the key issue is not only whether leaders delegate authority, but whether employees feel that their proactive actions are acceptable and will not be punished in an unfair way. In more realistic situations, when empowerment is provided in an appropriate way, EL can reduce the psychological burden of speaking up and trying new methods. Leaders can do this by supporting autonomy, encouraging participation in decisions, and allowing experimentation (Kim & Beehr, 2021; Naqshbandi et al., 2019). In this sense, EL offers employees a sense of “permission” and a certain level of “protection”. In high-power-distance contexts, EL may also function as a cultural neutralizer, reducing hierarchy-related hesitation and making proactive behavior feel more acceptable.

Based on this logic, we expect EL to change how strongly the three types of capital relate to innovative performance. With higher EL, employees may more readily leverage interpersonal networks, strengthening the SOCAP–innovation link (Adler & Seok-Woo, 2002; Monje-Amor et al., 2021; Phelps et al., 2012). Autonomy and encouragement of experimentation can help employees apply and recombine expertise proactively, strengthening the HUCAP–innovation link (Hoang et al., 2021; Khatoon et al., 2024; Rao Jada et al., 2019). An empowering climate may also allow employees with high PSYCAP to pursue novel ideas with lower perceived risk and trial-and-error costs, strengthening the PSYCAP–INNO link (Amundsen & Martinsen, 2014; Kim & Beehr, 2023).

Thus:

H12a. 

EL positively moderates the relationship between employees’ SOCAP and INNO, such that the relationship is stronger under high EL.

H12b. 

EL positively moderates the relationship between employees’ HUCAP and INNO, such that the relationship is stronger under high EL.

H12c. 

EL positively moderates the relationship between employees’ PSYCAP and INNO, such that the relationship is stronger under high EL.

2.3.5. Mediating Effects of Employee Capitals (H13)

Building on H1–H6 and H7–H9, this study proposes that TESM and SESM influence INNO primarily through employees’ accumulated social, human, and PSYCAP. Digital collaboration technologies often affect innovation by reshaping relational, cognitive, and psychological resources rather than operating only through direct effects (Leonardi, 2015; Nambisan & Baron, 2021).

First, TESM and SESM can build SOCAP via visible collaboration and informal relationship building, respectively (Kamboj et al., 2017; Pekkala et al., 2022; C. Wang & Cardon, 2019; Wei et al., 2022). SOCAP then supports innovation by enabling access to diverse knowledge and collaboration (Singh et al., 2021; X. Zhao et al., 2022). Second, both orientations can enhance HUCAP via task-based learning and informal knowledge exchange (Leonardi, 2015; Leyer et al., 2019; Lu & Pan, 2019; Riemer et al., 2010), and HUCAP is linked to INNO (Hanifah et al., 2022; Z. Wang et al., 2021). Third, ESM use can cultivate PSYCAP through recognition, support, and belonging (Men et al., 2020; Nusrat et al., 2024), and PSYCAP is consistently associated with INNO in knowledge-intensive settings (Baig et al., 2021; Liang et al., 2024).

Accordingly, this study advances a multiple mediation framework in which the three capitals jointly transmit the effects of TESM and SESM on INNO (Specific indirect effects are concluded in Supplementary Material S6, Table S19):

H13 (overall mediation).

Employees’ SOCAP, HUCAP, and PSYCAP mediate the relationships between ESM use orientations (TESM and SESM) and INNO.

Based on the above arguments, the overall conceptual framework of this study is presented in Figure 1.

3. Materials and Methods

3.1. Research Design

This study adopts a cross-sectional quantitative research design. Given the complexity of the hypothesized model—featuring parallel mediation mechanisms and moderating effects—and the prediction-oriented aim of explaining INNO (J. Hair & Alamer, 2022), we employed Partial Least Squares Structural Equation Modeling (PLS-SEM) for data analysis. All analyses were conducted using SmartPLS4 (SmartPLS GmbH, Monheim am Rhein, Germany), which is well suited for estimating complex structural models and assessing predictive performance. To enhance the credibility of the findings, we incorporated both procedural and statistical remedies to minimize potential common method variance (CMV).

3.2. Sampling and Data Collection

Data were collected from June to August 2025 using the Wenjuanxing online survey platform. The target population consisted of full-time employees working in three knowledge-intensive sectors in China (information technology, financial services, and manufacturing). To ensure structural diversity, we applied quota targets across industry and functional role (R&D, marketing, operations, and HR). Survey access was facilitated through organizational contacts (e.g., HR departments or senior managers) in 26 firms located across five provincial-level regions/municipalities (Beijing, Shanghai, Guangdong, Jiangsu, and Sichuan), who distributed the survey link to eligible employees within their organizations. Participation was voluntary, and respondents were assured of anonymity and confidentiality. A small token incentive was offered to encourage participation, while a pre-established response-quality protocol was implemented to mitigate careless responding (see Appendix C.1).

3.3. Measures

All constructs in this study were measured using mature multi-item scales, and a five-point Likert scale (1 = strongly disagree; 5 = strongly agree) was employed. To adapt to the context of ESM, we slightly modified some of the wording (full items are presented in Appendix A), and all scales were translated into Chinese and back-translated.

TESM and SESM items were adapted from Zhang et al. (2019). TESM consists of 3 questions (work coordination, task management, work information exchange); SESM consists of 4 questions (informal interaction, relationship maintenance, emotional support).

SOCAP items were adapted from Chiu et al. (2006) and were based on the framework of Nahapiet and Ghoshal (1998). Question 14 cover the three dimensions of structure, relationship and cognition.

HUCAP adopts the 14-item self-assessment scale proposed by Dahiya and Raghuvanshi (2022), covering dimensions such as capability, leadership and motivation, creativity, etc. “Leadership and motivation” refer to the self-driven/active nature of employees, rather than the authorization behavior of managers.

PSYCAP was measured using 14 items selected from the Psychological Capital Questionnaire (PCQ-24) (Luthans et al., 2007), representing the four core dimensions: self-efficacy, hope, optimism, and resilience.

EL adopts the 12-item empowerment leadership behavior scale (LEB) proposed by Ahearne et al. (2005).

INNO was based on the innovation work behavior scale developed by Janssen (2000). To control the length of the questionnaire, we adopted a simplified 6-item version, covering three dimensions: idea generation, idea promotion, and idea realization (2 items per dimension); the specific items are presented in Appendix A.

Modeling choice for capitals. Although SOCAP, HUCAP, and PSYCAP are conceptually multidimensional, we modeled each as an overall first-order reflective construct to keep the structural model parsimonious and aligned with the study’s mediation-focused design.

Control variables included gender (dummy-coded), age, education level, tenure, and job level (the latter four were ordinal-coded). Detailed category ranges for age and tenure are provided in Appendix A and Appendix C.1.

3.4. Common Method Bias and Reliability Precautions

Given that this study relies on a single, self-reported survey, CMV cannot be completely ruled out. Procedurally, respondents were assured of anonymity and confidentiality and were informed that there were no right or wrong answers. Statistically, Harman’s single-factor test indicated that a single factor did not account for the majority of the variance. In addition, we assessed collinearity-based diagnostics (inner VIF and full collinearity VIF) to check for potential method-related inflation. Overall, the diagnostics suggest that severe CMV is unlikely to materially bias the main estimates, although it cannot be entirely excluded. Consistent with recent methodological reviews, we treat these diagnostics as suggestive rather than definitive evidence and therefore interpret the results with appropriate caution given the single-source survey design (Podsakoff et al., 2003).

3.5. Data Analysis Strategy

Partial least squares structural equation modeling (PLS-SEM) was employed using SmartPLS4 to assess both the measurement and structural models. PLS-SEM is appropriate for this study because (1) the research model includes multiple latent constructs and mediation–moderation relationships, (2) the sample size aligns with recommended criteria in terms of statistical power and model complexity, and (3) PLS-SEM performs robustly under non-normal data conditions (J. Hair & Alamer, 2022; J. F. Hair et al., 2021; Ringle et al., 2023).

The measurement model was evaluated through reliability (Cronbach’s α, composite reliability), convergent validity (AVE), and discriminant validity (Fornell–Larcker and HTMT). The structural model was tested using a bootstrapping procedure with 5000 resamples. Mediation paths were evaluated following X. Zhao et al. (2010). Moderation was estimated using SmartPLS4’s built-in moderation procedure, which implements a two-stage approach. Standardized latent variable scores were used to form the interaction terms in the second stage (default setting). Notably, under this moderation routine, SmartPLS automatically includes the main effect of the moderator on the dependent variable (EL → INNO) in the estimation output, even if this path is not explicitly drawn in the structural model; this effect is therefore reported in Appendix B together with the focal interaction results. Control variables were included in all regression paths predicting INNO. To focus on the hypothesized boundary role of empowering leadership (EL), we conceptualized EL primarily as a moderator of the capital → INNO relationships and therefore did not theorize a direct EL → INNO hypothesis in the main model. However, consistent with SmartPLS4’s two-stage moderation estimation, the EL → INNO main effect is included by default in the estimation output and is reported in Appendix B to demonstrate that the focal moderation inferences remain substantively unchanged when the moderator’s main effect is present.

Model evaluation followed established PLS-SEM guidelines, including assessment of measurement quality (reliability, convergent and discriminant validity), structural model evaluation (path significance, R² and effect sizes f², collinearity), and approximate model fit (SRMR). In addition, PLSpredict was conducted using k-fold cross-validation with SmartPLS default settings, reporting Q2_predict and prediction errors (RMSE/MAE), and comparing them with a linear model benchmark (Ringle et al., 2024).

3.6. Ethical Consideration

This study was approved by the Ethics Committee in Human Research. All participants were adults. Informed consent was obtained from all participants included in the study. Because the study involved minimal risk and was conducted through an anonymous online survey, the Ethics Committee granted a waiver of signed documentation of informed consent. Prior to accessing the questionnaire, respondents were presented with a written informed consent statement on the first page of the survey, which explained the purpose of the study, the voluntary nature of participation, and their right to withdraw at any time before submission. Participants provided written informed consent electronically by voluntarily proceeding with and submitting the questionnaire. No identifiable personal information was collected, and the data were used solely for academic research.

4. Results

This chapter presents the empirical results: sample characteristics followed by an assessment of the measurement model; the structural model evaluation—both its explanatory and predictive performance—and reports the hypothesis tests, including direct, moderating, and mediating effects.

4.1. Respondent Profile and Sample Overview

Of the 800 invited employees, 689 completed the questionnaire. Following the pre-established data quality screening protocol (see Appendix C.1), 613 responses were retained for analysis.

Table 1. Sample characteristics.

Variables	Categories	Respondents
		Frequency	Percentage
Gender	Male	318	51.88%
	Female	295	48.12%
Age	18–25	110	17.9%
	26–30	197	32.1%
	31–40	126	20.6%
	41–50	112	18.3%
	51–60	68	11.1%
Education	High school/Technical secondary	16	2.6%
	College	165	26.9%
	Bachelor degree	262	42.7%
	Master’s degree	155	25.3%
	Doctor degree	15	2.4%
Work Tenure	1–5 years	185	30.2%
	6–10 years	244	39.8%
	11–20 years	144	23.5%
	21–30 years	31	5.1%
	31–40 years	9	1.5%
Job Level	General staff	285	46.5%
	Junior management	208	33.9%
	Supervisor	75	12.2%
	Senior Executive	45	7.3%

summarizes the sample characteristics (e.g., industry, functional role, and key demographic profiles), indicating that respondents were predominantly early to mid-career professionals with relatively high educational attainment. The final sample size exceeds our initial target of 500 and provides adequate statistical power for estimating the proposed model and conducting predictive assessments (e.g., PLSpredict). The final dataset consists of respondents from multiple knowledge-intensive enterprises across different industries and organizational functions. The sample shows balanced representation across industry–function quotas and includes participants from several regions of China. Detailed information on quota attainment, organizational coverage, regional distribution demographics, and descriptive statistics of the key constructs is reported in Supplementary Material S1 (see Tables S1–S5).

4.2. Measurement Model Assessment

We assessed the reflective measurement model prior to estimating the structural paths, drawing on standard PLS-SEM evaluation criteria (J. F. Hair et al., 2019). Convergent validity was then evaluated by reviewing indicator loadings alongside AVE. All retained indicators met the minimum loading criterion (≥0.50), with the majority exceeding 0.70. AVE values ranged from 0.536 to 0.808, surpassing the 0.50 threshold and supporting convergent validity. Internal consistency reliability was satisfactory, with Cronbach’s α values ranging from 0.858 to 0.945 and composite reliability (CR) values between 0.894 and 0.951 (see

Table 2. Assessment of the Measurement Model.

	Cronbach’s Alpha	Composite Reliability (Rho_a)	Composite Reliability (Rho_c)	Average Variance Extracted (AVE)
HUCAP	0.943	0.944	0.950	0.576
INNO	0.858	0.859	0.894	0.584
EL	0.937	0.943	0.946	0.599
PSYCAP	0.933	0.934	0.942	0.536
SESM	0.877	0.878	0.916	0.731
SOCAP	0.945	0.946	0.951	0.582
TESM	0.882	0.882	0.927	0.808

Note: TESM = Task-oriented enterprise social media use; SESM = Social-oriented enterprise social media use; SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance; EL = Empowering leadership.

).

Discriminant validity was assessed using both Fornell–Larcker criterion and the heterotrait–monotrait ratio (HTMT). The Fornell–Larcker criterion was chosen because it is an established and intuitively interpretable benchmark for reflective measurement models in PLS-SEM. It evaluates whether each construct shares more variance with its own indicators than with other constructs. However, because this criterion may lack sensitivity in detecting subtle discriminant validity problems, it was not treated as the sole standard. We therefore complemented it with HTMT, which is widely regarded as a more stringent diagnostic. This dual-assessment approach aligns with best-practice recommendations to report both criteria for a more robust assessment of discriminant validity (J. F. Hair et al., 2019). The square root of AVE for each construct exceeded its correlations with other constructs, and all HTMT values remained below the conservative cut-off of 0.85, confirming adequate discriminant validity (see

Table 3. Fornell–Larcker criterion.

	TESM	SESM	SOCAP	HUCAP	PSYCAP	EL	INNO
TESM	0.899
SESM	0.518	0.855
SOCAP	0.436	0.459	0.763
HUCAP	0.395	0.470	0.451	0.759
PSYCAP	0.435	0.438	0.479	0.532	0.732
EL	−0.033	−0.046	−0.134	−0.118	−0.097	0.774
INNO	0.314	0.363	0.459	0.458	0.445	0.567	0.765

Note: TESM = Task-oriented enterprise social media use; SESM = Social-oriented enterprise social media use; SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance; EL = Empowering leadership.

and

Table 4. Discriminant validity (HTMT).

Path	Heterotrait–Monotrait Ratio (HTMT)
INNO ↔ HUCAP	0.507
EL ↔ HUCAP	0.127
EL ↔ INNO	0.630
PSYCAP ↔ HUCAP	0.565
PSYCAP ↔ INNO	0.496
PSYCAP ↔ EL	0.108
SESM ↔ HUCAP	0.515
SESM ↔ INNO	0.417
SESM ↔ EL	0.058
SESM ↔ PSYCAP	0.481
SOCAP ↔ HUCAP	0.477
SOCAP ↔ INNO	0.507
SOCAP ↔ EL	0.147
SOCAP ↔ PSYCAP	0.508
SOCAP ↔ SESM	0.503
TESM ↔ HUCAP	0.432
TESM ↔ INNO	0.360
TESM ↔ EL	0.040
TESM ↔ PSYCAP	0.478
TESM ↔ SESM	0.588
TESM ↔ SOCAP	0.476

Note: TESM = Task-oriented enterprise social media use; SESM = Social-oriented enterprise social media use; SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance; EL = Empowering leadership.

). Additional measurement model outputs, including indicator cross-loadings, HTMT matrices, Fornell-Larcker statistics, and indicator-level collinearity diagnostics, are reported in Supplementary Material S3 (See Tables S8–S10).

Notably, EL exhibited only negligible bivariate associations with ESM use orientations and employee capitals in the correlation matrix (r = −0.03 to −0.13), suggesting that it is not simply another “more-is-better” resource aligned with ESM use. Importantly, our theorizing and tests focus on the interaction effects, such that the boundary role of EL is evaluated through moderation in the structural model rather than inferred from zero-order correlations.

Since this study relied on a single, self-reported survey, common method variance (CMV) could not be entirely ruled out. To mitigate method bias, procedural remedies were implemented, including anonymity assurance, neutral questionnaire instructions, and the separation of construct measures across questionnaire sections (see Supplementary Material S4, Table S12). Ex post diagnostics provided only tentative indications that CMV is unlikely to substantially bias the estimates: Harman’s single-factor test indicated that the first factor accounted for 27.985% of the variance (Podsakoff et al., 2003) (see Supplementary Material S4, Table S13). Collinearity diagnostics indicated no multicollinearity concerns in the structural model. Inner VIF values were low (max = 2.870), including the interaction terms (VIFs = 1.410–1.571), suggesting that collinearity is unlikely to bias the path estimates (see Supplementary Material S4, Table S15). Full collinearity VIFs ranged from 1.025 to 4.344 (see Supplementary Material S4, Table S14). Overall robustness checks are summarized in Supplementary Material S4 (see Table S16). Although the Full VIF for INNO exceeded the conservative 3.3 guideline, it remained below 5; therefore, severe collinearity-related inflation is unlikely, while CMV cannot be completely ruled out due to the single-source survey design.

Model fit was evaluated using the Standardized Root Mean Square Residual (SRMR) as a supplementary approximate fit diagnostic. In PLS-SEM, model evaluation does not rest primarily on global fit indices in the same confirmatory manner as in covariance-based SEM; instead, greater emphasis is placed on measurement quality, collinearity diagnostics, explanatory power, predictive performance, and the significance of structural paths. SRMR was selected because it is a commonly reported and readily interpretable approximate fit index in PLS-SEM, reflecting the standardized discrepancy between observed and model-implied correlations. While other measures such as NFI, d_ULS, and d_G can also be computed, they are generally treated as supplementary diagnostics that can be sensitive to model complexity and sample characteristics and are not regarded as decisive stand-alone cut-offs for model adequacy. Accordingly, SRMR served as a complementary diagnostic check, with the obtained value of 0.034 falling below the conservative threshold of 0.08, indicating acceptable approximate model fit (see Appendix B,

Table A3. Model performance summary (SmartPLS default moderation estimation).

Metric	Model Fit
(R2) for INNO	0.770
Q2_predict for INNO (PLSpredict)	0.488
SRMR	0.034
PLSpredict summary: RMSE wins/total	46/48
PLSpredict summary: MAE wins/total	47/48

).

4.3. Structural Model Assessment

The structural model was evaluated using bootstrapping with 5000 resamples to test the significance of path coefficients. Multicollinearity was again examined at the structural level; all VIF values were below 5, indicating no serious collinearity concerns.

4.3.1. Explanatory Power

Model explanatory power was assessed using R² values (see

Table 5. R-squared and Adjusted R-squared coefficient.

Variables	SOCAP	HUCAP	PSYCAP	INNO
R-squared	0.264	0.252	0.251	0.770
Adj. R-squared	0.262	0.250	0.249	0.765

Note: SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance.

). The three mediators—SOCAP (R² = 0.264), HUCAP (R² = 0.252), and PSYCAP (R² = 0.251)—showed moderate explanatory power. INNO exhibited substantial explained variance (R² = 0.770), suggesting that the combined predictors (ESM use orientations, the three capitals, and interaction terms) jointly explained a large proportion of variance in INNO.

Effect sizes (f²) were calculated to assess substantive contribution (see

Table 6. Effect sizes.

Path	f-Square
TESM → HUCAP	0.042
TESM → PSYCAP	0.079
TESM → SOCAP	0.073
SESM → HUCAP	0.129
SESM → PSYCAP	0.083
SESM → SOCAP	0.101
SOCAP → INNO	0.322
HUCAP → INNO	0.224
PSYCAP → INNO	0.120
EL × SOCAP → INNO	0.006
EL × HUCAP → INNO	0.009
EL × PSYCAP → INNO	0.013

Note: TESM = Task-oriented enterprise social media use; SESM = Social-oriented enterprise social media use; SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance; EL = Empowering leadership.

). SOCAP showed the largest incremental explanatory power for INNO (SOCAP → INNO: f² = 0.322), followed by HUCAP (HUCAP → INNO: f² = 0.224) and PSYCAP (PSYCAP → INNO: f² = 0.120). The interaction terms exhibited small effect sizes (f² = 0.006–0.013), which is typical for moderation effects in behavioral research.

4.3.2. Predictive Performance

Out-of-sample predictive relevance was examined using PLSpredict (see

Table 7. PLSpredict.

Construct	Q2 Predict	RMSE	MAE
HUCAP	0.246	0.871	0.708
INNO	0.488	0.717	0.564
PSYCAP	0.245	0.871	0.715
SOCAP	0.258	0.863	0.713

Note: SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance.

). All endogenous constructs produced positive Q²_predict values (SOCAP = 0.258; HUCAP = 0.246; PSYCAP = 0.245; INNO = 0.488), indicating meaningful predictive capability. For INNO, the overall prediction error metrics were RMSE = 0.717 and MAE = 0.564.

To benchmark predictive performance, prediction errors from the PLS-SEM model were compared with those from a linear regression (LM) benchmark model. Across the 48 indicators of the four predicted constructs (SOCAP = 14, HUCAP = 14, PSYCAP = 14, and INNO = 6; k-fold cross-validation; SmartPLS default settings) (Ringle et al., 2024), the PLS-SEM model produced lower RMSE values than the LM benchmark for 46 indicators (95.8%) and lower MAE values for 47 indicators (97.9%), indicating better overall predictive accuracy. To address potential concerns regarding the omission of a theorized direct effect of the moderator, we clarify a modeling detail in SmartPLS4’s built-in two-stage moderation estimation: the software automatically includes the moderator’s main effect on the dependent variable of the moderated relationship (i.e., EL → INNO) even if this path is not explicitly drawn in the structural model. Accordingly, we report this main effect as part of the moderation estimation output. The main effect is strong and positive (EL → INNO: β = 0.698, p < 0.001) (see Appendix B,

Table A4. Main effect of EL → INNO (bootstrapped; standardized coefficient).

Path	β	t	p	95% CI	Decision
EL → INNO	0.698	27.564	<0.001	[0.648, 0.747]	Significant

), and the focal moderation inferences remain substantively unchanged and robust in its presence (see Appendix B for detailed estimation settings and stability results).

4.4. Hypothesis Testing

This section reports on the hypothesis testing results for the proposed theoretical model, including direct effects, the moderating effect of empowering leadership, and the mediating effects of employee capitals. The full structural model results with standardized coefficients are visualized in Figure 2, core hypothesis tests are summarized in

Table 8. Summary of Structural Paths.

Hypothesis	Path	β	95% CI	p	Result
H1	TESM → SOCAP	0.272	[0.187, 0.361]	<0.001	Supported
H2	TESM → HUCAP	0.207	[0.130, 0.291]	<0.001	Supported
H3	TESM → PSYCAP	0.285	[0.210, 0.365]	<0.001	Supported
H4	SESM → SOCAP	0.318	[0.229, 0.405]	<0.001	Supported
H5	SESM → HUCAP	0.363	[0.277, 0.443]	<0.001	Supported
H6	SESM → PSYCAP	0.291	[0.210, 0.367]	<0.001	Supported
H7	SOCAP → INNO	0.324	[0.271, 0.378]	<0.001	Supported
H8	HUCAP → INNO	0.277	[0.224, 0.333]	<0.001	Supported
H9	PSYCAP → INNO	0.205	[0.150, 0.263]	<0.001	Supported
H10	TESM → INNO	−0.008	[−0.055, 0.040]	0.744	Not supported
H11	SESM → INNO	0.029	[−0.017, 0.073]	0.200	Not supported
H12a	EL × SOCAP → INNO	0.041	[0.000, 0.084]	0.060	Marginally supported
H12b	EL × HUCAP → INNO	0.056	[0.005, 0.108]	0.031	Supported
H12c	EL × PSYCAP → INNO	0.064	[0.016, 0.112]	0.009	Supported

Note: TESM = Task-oriented enterprise social media use; SESM = Social-oriented enterprise social media use; SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance; EL = Empowering leadership.

, and detailed mediation results are presented in

Table 9. Mediation Analysis Details.

Path	β (Indirect)	95% CI	β (Direct)	Total (Σab)	Mediation Type
TESM → SOCAP → INNO	0.088 ***	[0.059, 0.123]	−0.008 (n.s.)	0.204	Indirect-only mediation
TESM → HUCAP → INNO	0.057 ***	[0.035, 0.083]	−0.008 (n.s.)	0.204	Indirect-only mediation
TESM → PSYCAP → INNO	0.058 ***	[0.038, 0.084]	−0.008 (n.s.)	0.204	Indirect-only mediation
SESM → SOCAP → INNO	0.103 ***	[0.070, 0.138]	0.029 (n.s.)	0.264	Indirect-only mediation
SESM → HUCAP → INNO	0.101 ***	[0.071, 0.133]	0.029 (n.s.)	0.264	Indirect-only mediation
SESM → PSYCAP → INNO	0.060 ***	[0.038, 0.084]	0.029 (n.s.)	0.264	Indirect-only mediation

Note: *** p < 0.001; n.s. = not significant. TESM = Task-oriented enterprise social media use; SESM = Social-oriented enterprise social media use; SOCAP = Social capital; HUCAP = Human capital; PSYCAP = Psychological capital; INNO = Innovative performance. Indirect effects were tested via bootstrapping with 5000 resamples. Indirect-only mediation interpretation follows X. Zhao et al. (2010), supported by significant indirect effects and non-significant direct effects.

.

4.4.1. Direct Effects (H1–H11)

Both forms of enterprise social media use significantly predicted the three forms of employee capital. TESM had positive effects on SOCAP (β = 0.272, p < 0.001), HUCAP (β = 0.207, p < 0.001), and PSYCAP (β = 0.285, p < 0.001), supporting H1–H3. SESM also significantly enhanced SOCAP (β = 0.318, p < 0.001), HUCAP (β = 0.363, p < 0.001), and PSYCAP (β = 0.291, p < 0.001), supporting H4–H6.

The three capitals were positively related to INNO, supporting H7–H9. SOCAP showed the strongest effect (β = 0.324, p < 0.001), followed by HUCAP (β = 0.277, p < 0.001) and PSYCAP (β = 0.205, p < 0.001).

In contrast, TESM (β = −0.008, 95% CI [−0.055, 0.040], p = 0.744) and SESM (β = 0.029, 95% CI [−0.017, 0.073], p = 0.200) did not show significant direct effects on INNO.

Control variables (gender, age, education, tenure, and job level) showed no significant effects on INNO.

4.4.2. Moderation Effects (H12a–H12c)

As specified in the research model, EL was modeled as a moderator of the capital-INNO links rather than as a direct predictor of INNO. The interaction effects were positive for all three capitals. Specifically, the interactions with HUCAP (β = 0.056, p = 0.031) and PSYCAP (β = 0.064, p = 0.009) were significant. Whereas the interaction with SOCAP was marginal (β = 0.041, p = 0.060; 95% CI [0.000, 0.084]), suggesting preliminary but not conclusive evidence under the 0.05 criterion. Overall, the results suggest that EL tends to strengthen the effects of employee capital on INNO, with stronger evidence for human and PSYCAP. Because the PSYCAP × EL interaction showed the strongest significant moderation effect, Figure 3 presents this relationship as an illustrative plot. Interaction plots, simple slope analyses, and Johnson-Neyman analyses are presented in Supplementary Material S5 (see Tables S17 and S18).

As a sensitivity check, we also examined the moderation results after explicitly including the main effect of EL on INNO. The moderation conclusions remained substantively unchanged, as reported in Section 4.5 and Appendix B.

4.4.3. Mediation Effects (H13)

Indirect effects were tested using bootstrapping with 5000 resamples, and the key mediation results are summarized in Table 9. To unpack H13, we report the specific indirect effects via SOCAP/HUCAP/PSYCAP for both TESM and SESM. Supporting the overall mediation hypothesis (H13), the orientation-specific indirect effects via each of the three capitals were positive and statistically significant, as their bootstrapped 95% confidence intervals excluded zero (see Supplementary Material S6, Table S19). For TESM, the indirect effects operated through SOCAP (TESM → SOCAP → INNO: β = 0.088, 95% CI [0.059, 0.123]), HUCAP (β = 0.057, 95% CI [0.035, 0.083]), and PSYCAP (β = 0.058, 95% CI [0.038, 0.084]). For SESM, significant mediation was observed through SOCAP (SESM → SOCAP → INNO: β = 0.103, 95% CI [0.070, 0.138]), HUCAP (β = 0.101, 95% CI [0.071, 0.133]), and PSYCAP (β = 0.060, 95% CI [0.038, 0.084]). The summed (total) indirect effects were also significant for both orientations (TESM → INNO: Σab = 0.204, 95% CI [0.155, 0.260]; SESM → INNO: Σab = 0.264, 95% CI [0.208, 0.319]) (see Supplementary Material S6, Table S20). Importantly, the corresponding direct effects were not significant (TESM → INNO: c′ = −0.008, p = 0.744; SESM → INNO: c′ = 0.029, p = 0.200), while the indirect effects were significant, supporting an indirect-only mediation interpretation (X. Zhao et al., 2010) (see Supplementary Material S6, Table S21).

Taken together, these results suggest that TESM and SESM relate to INNO primarily through SOCAP, HUCAP, and PSYCAP, consistent with an indirect-only mediation interpretation (X. Zhao et al., 2010).

4.5. Sensitivity Analyses and Robustness Checks

To ensure the findings are not contingent on specific analytical choices, we conducted three sensitivity checks (see

Table 10. Summary of sensitivity analyses.

Sensitivity Check	Purpose	Main Conclusion
SmartPLS moderation specification	To clarify whether the moderator’s main effect was included in the moderation estimation	EL → INNO was automatically included in SmartPLS moderation estimation; focal moderation results remained stable
Control-variable specification	To examine whether results depend on the treatment of demographic/work-related controls	Main path directions and significance decisions remained substantively unchanged
Bootstrap resampling setting	To examine whether inference depends on resampling settings	Results based on 10,000 bootstrap resamples were consistent with the original 5000-resample results

). First, we verified the model specification for moderation. While SmartPLS4’s two-stage procedure automatically accounts for the moderator’s main effect (EL → INNO), re-estimating the model with this path explicitly drawn yielded identical estimates, confirming the stability of the interaction effects. Second, we tested the model’s sensitivity to control variables. Re-estimating the structural model without the demographic controls (gender, age, education, tenure, and job level) showed that the path coefficients and significance levels of the core hypotheses (H1–H9 and H12–H13) remained substantively unchanged. Third, to ensure statistical stability, we increased the bootstrapping resamples from 5000 to 10,000. The results remained consistent, with no changes in hypothesis support. These combined checks, detailed in Appendix B, provide convergent evidence that our conclusions are robust across alternative modeling and estimation settings.

First, for the moderation specification check, we examined whether the moderation conclusions remained stable after explicitly reporting the main effect of EL on INNO. This check is relevant because SmartPLS4’s two-stage moderation procedure automatically includes the direct effect of the moderator on the dependent variable of the moderated relationship. EL showed a strong positive main effect on INNO (β = 0.698, p < 0.001; 95% CI [0.648, 0.747]). Importantly, the focal moderation conclusions remained substantively unchanged: EL significantly strengthened the effects of HUCAP and PSYCAP on INNO, while the interaction with SOCAP remained marginal. Detailed results are reported in Appendix B.

Second, for the control-variable specification check, we re-estimated the model without gender, age, education, tenure, and job level. The directions and significance decisions of the main theoretical paths remained substantively unchanged. Third, increasing the number of bootstrap resamples from 5000 to 10,000 did not alter the hypothesis-support decisions.

5. Discussion

5.1. Summary of Key Findings

The results are consistent with the proposed multi-capital mediation framework. In the structural model, both TESM and SESM were positively associated with employees’ SOCAP, HUCAP, and PSYCAP. When all three capitals were included simultaneously, the direct paths from ESM use orientations to INNO were not statistically supported. This pattern is consistent with an indirect-only mediation interpretation (X. Zhao et al., 2010), suggesting that ESM use orientations relate to INNO mainly through the relational, skill-based, and psychological resources captured in the model rather than through an additional residual direct effect.

Among the three mediators, SOCAP showed the strongest association with INNO, while HUCAP and PSYCAP were also positively related but with smaller effects. This ordering is reasonable in Chinese knowledge-intensive enterprises, where innovation often requires cross-unit coordination and timely access to feedback and support. The model explained a substantial share of variance in INNO (R² = 0.770), indicating that the three-capital pathway captures a large portion of the ESM–innovation linkage in this context.

The interaction effects indicate that leadership shapes how resources translate into INNO. EL strengthened the HUCAP → INNO and PSYCAP → INNO relationships, whereas the moderation for SOCAP → INNO was weaker. This pattern suggests that EL matters most at the “conversion” stage—from resources to action. This may be because empowering signals reduce the perceived risk of experimentation and give employees more discretion when outcomes are uncertain.

5.2. General Discussion

5.2.1. Interpretation of TESM and SESM

TESM and SESM were both positively linked to SOCAP, HUCAP and PSYCAP, but they appear to do so through different interaction patterns. For SOCAP, TESM tends to strengthen connections by making work-related expertise more visible. When discussions are organized around task requests, proposed solutions, and peer feedback, it becomes easier for employees to locate “who knows what” and coordinate across roles (Pitafi et al., 2023). Repeated problem-focused exchange can also reinforce collaborative ties and embeddedness in the organizational network over time (Caya & Mosconi, 2023). SESM, in contrast, supports SOCAP more through informal communication. Casual exchanges can reduce psychological distance, build interpersonal trust, and increase perceived support, which makes help-seeking and collaboration feel easier (Ma et al., 2021; Wu et al., 2021).

For HUCAP, TESM is more directly tied to work learning. Persistent feedback loops in task threads and searchable solution archives can help employees refine domain knowledge and compare alternative methods (Berraies et al., 2020; Xiong & Sun, 2022). This is consistent with evidence that ESM-enabled knowledge acquisition is associated with innovation-relevant competence (J. Dai et al., 2024). That said, the learning benefit depends on interaction quality: if task threads are dominated by shallow updates rather than substantive problem analysis, TESM may add distraction with limited skill development. SESM may contribute to HUCAP in a different way by exposing employees to cross-department perspectives and informal learning, which can broaden problem framing (Sharma et al., 2023). Social motivation may also sustain participation in knowledge-sharing communities (Kalra & Baral, 2019). When informal exchanges drift too far from work, they can dilute attention and weaken skill-related gains.

A similar contrast appears for PSYCAP. TESM can reduce uncertainty by clarifying task ownership and making progress visible, which may support confidence when employees face new challenges (Pekkala et al., 2022). Documented contributions and constructive feedback can also reinforce perceived competence and autonomy, strengthening hope and resilience (Y. Sun et al., 2023). SESM contributes more through socio-emotional support: a sense of belonging and informal encouragement may foster optimism and resilience (Bodhi et al., 2023; Feng et al., 2024; Pitafi, 2024). Negative interactions, exclusion, or conflict can also undermine these gains, highlighting the importance of norms that support respect and psychological safety in informal spaces.

5.2.2. The Role of SOCAP

The positive SOCAP–INNO relationship suggests that innovation in knowledge-intensive work is often supported by relational resources. When employees have stronger connections, information and feedback can travel faster, it becomes easier to ask for help, and it is more feasible to bring the right people together to refine and implement an idea (Lyu et al., 2022; Singh et al., 2021). In this sense, SOCAP can reduce the coordination frictions that often separate idea generation from implementation. However, not all networks support innovation in the same way. Dense and homogeneous ties can facilitate smooth coordination, but they may limit novelty if employees repeatedly encounter similar viewpoints (Al-Omoush et al., 2022). More bridging and diverse connections are more likely to provide nonredundant knowledge and new perspectives that support recombination and reframing. A practical implication is that organizations may benefit from balancing cohesive internal collaboration with opportunities for boundary-spanning exchange.

The Chinese cultural context may also help explain why SOCAP showed the strongest association with INNO. In collectivist workplace settings, employees often place strong value on relational obligation, interpersonal harmony, and trusted group-based exchange. As a result, social capital may be particularly important because innovative ideas often require support from colleagues, access to nonredundant knowledge, and coordination across formal boundaries. In this context, ESM-enabled relational embeddedness can help employees seek feedback, mobilize allies, and reduce uncertainty when promoting new ideas. At the same time, this interpretation should be treated as contextual rather than conclusive, because cultural values such as collectivism were not directly measured in this study.

5.2.3. The Role of HUCAP

The significant HUCAP→INNO relationship suggests that employees’ expertise and learning capacity provide a practical basis for generating feasible ideas and moving them toward implementation (Muñoz-Pascual & Galende, 2020). This is particularly relevant for iterative innovation, where employees need to revise assumptions and reorganize existing knowledge under uncertainty (Ullah et al., 2021; S. Zhao et al., 2020). In ESM-enabled work, HUCAP may be reflected in the quality of task-thread contributions—for example, explaining reasoning, comparing alternatives, and integrating feedback—so that ideas are less likely to stall at the discussion stage.

Expertise alone is not enough when employees have little discretion to experiment; in such settings, competence may translate primarily into efficiency rather than innovation. This is consistent with the moderation results, suggesting that leadership can shape whether employees feel able to use their expertise in more exploratory ways.

5.2.4. The Role of PSYCAP

PSYCAP can support innovation by shaping how employees appraise challenges and persist under uncertainty. Employees with higher PSYCAP tend to tolerate ambiguity, recover from setbacks, and continue investing effort when early attempts do not work out (Hu et al., 2023; Y. Liu et al., 2023). Prior research similarly links PSYCAP to creativity and innovation across organizational contexts (Alshebami, 2021).

These psychological resources are less likely to show up as innovation when interpersonal risk is high and speaking up is perceived as unsafe. In this context, supportive peer interaction on ESM may help, and empowering signals from leaders may further reduce the perceived risk of experimentation. Together, these conditions can make it more likely that employees’ optimism, hope, and resilience show up in observable innovative behaviors.

5.2.5. The Boundary Role of EL

The moderation results suggest that the role of EL differs across capital types. EL strengthened the HUCAP → INNO and PSYCAP → INNO relationships, indicating that empowerment may help employees translate expertise and psychological resources into innovative behavior. This may be because autonomy-supportive signals reduce the perceived risk of experimentation and give employees more discretion when outcomes are uncertain.

For SOCAP, the interaction was weaker. In Chinese firms, the mobilization of relational resources may rely more on peer norms and routine coordination than on formal leadership signals. Another possibility is partial substitution: because ESM already increases visibility and connectivity, there may be less additional room for leadership behaviors to amplify the innovation value of existing social ties. This pattern suggests that the SOCAP-related moderation may be contingent on factors such as network diversity or task interdependence. Overall, consistent with SDT, empowering signals that support autonomy may be especially consequential when innovation depends on employees’ internal resources (HUCAP and PSYCAP) rather than on access through social connections (SOCAP).

5.2.6. Theoretical Implications

First, this study responds to the inconsistent findings of prior research on the ESM-innovation linkage. Existing studies have reported mixed positive, non-significant, and even negative effects of ESM on employee innovation (Nusrat et al., 2021; Y. Shang et al., 2023). The indirect-only mediation results suggest that ESM use does not exert a robust residual direct effect on innovation after employees’ social, human, and psychological capital are considered. Instead, its innovation value appears to be realized primarily through the accumulation of these employee resources. This explains why studies that only focus on the direct effect of ESM usage frequency often obtain unstable results.

Second, this study extends the application of multi-capital theory in the digital collaboration context. Most existing studies treat employee capital as a single dimension or only examine a single capital mediator (Dahiya & Raghuvanshi, 2022; Saraf et al., 2023; Y. Wang et al., 2012). We demonstrate that social, human, and psychological capital form a bundled mediation mechanism, among which social capital plays the strongest driving role, providing a more complete explanation for the micro-mechanism of ESM enabling innovation.

Third, this study integrates Self-Determination Theory (SDT) and Social Learning Theory (SLT) to construct a coherent theoretical framework. We verify that ESM meets employees’ autonomy, competence, and relatedness needs (SDT) through task and social interaction, and provides observable learning channels through interactive feedback and behavioral modeling (SLT), which together promote the accumulation of three types of capital. This theoretical integration makes up for the deficiency of a single theory in explaining the ESM-innovation mechanism.

Finally, this study clarifies the nuanced boundary role of empowering leadership. Existing studies mostly regard EL as a universal positive moderator (Burhan & Khan, 2024; Khatoon et al., 2024; M. Lin et al., 2020), but we find that EL only significantly strengthens the conversion of internal resources (HUCAP and PSYCAP) to innovation and has no significant moderating effect on the SOCAP-INNO link. This boundary condition refines the applicable scenario of EL theory in digital collaborative innovation.

5.3. Practical Implications

5.3.1. Implications for Organizations Using ESM

The findings suggest that organizations should not treat ESM implementation as a simple matter of increasing platform adoption or user activity. Since TESM and SESM were related to INNO mainly through SOCAP, HUCAP, and PSYCAP, organizations should evaluate ESM less by surface indicators such as login frequency, number of posts, or general activity volume, and more by whether the platform helps employees build usable resources for innovation. Relevant indicators may include cross-unit interaction, access to nonredundant knowledge, reuse of archived solutions, constructive feedback, and psychological safety. In this sense, ESM should be managed not merely as a communication tool, but as a resource-building infrastructure for innovation.

The results also suggest that organizations should support both task-oriented and social-oriented ESM use. Given the positive relationship between TESM and HUCAP, organizations can strengthen task-oriented ESM functions by building standardized project channels, searchable solution repositories, technical discussion threads with clear ownership, and expert skill directories. These designs may help turn task-oriented ESM use into cumulative learning, rather than leaving task channels as spaces for shallow status updates with limited developmental value. At the same time, social-oriented ESM should not be overcontrolled or treated simply as a distraction. Informal spaces, peer-support groups, and cross-department communities can help employees build trust, reduce psychological distance, and obtain emotional support. However, organizations should also establish clear norms for respectful communication so that informal interaction remains supportive rather than becoming a source of conflict or information overload.

Because SOCAP showed the strongest relationship with INNO, organizations should pay particular attention to boundary-spanning collaboration. ESM can be used to create cross-functional problem-solving spaces where employees from different departments, roles, and business units exchange knowledge and combine diverse perspectives. Recognizing cross-unit knowledge sharing and collaborative support as part of broader team contribution systems may also encourage employees to build bridging ties rather than relying only on familiar contacts within their immediate work groups.

5.3.2. Implications for Managers and Team Leaders

The moderation results provide practical guidance for managers and team leaders. EL strengthened the effects of HUCAP and PSYCAP on INNO, while its moderating effect on the SOCAP–INNO relationship was weaker. This suggests that managers should not only encourage employees to accumulate resources through ESM but also help them convert these resources into innovative behavior.

For HUCAP, managers can give employees greater discretion to apply their expertise in exploratory ways. Employees who gain knowledge from ESM discussions may still hesitate to propose new solutions if experimentation is perceived as risky. Managers can reduce this risk by supporting knowledge-based suggestions, allocating limited time or resources for trials, and treating minor failures as learning opportunities. These actions can help employees transform knowledge and skills into actual innovation attempts.

For PSYCAP, managers can use ESM to provide constructive and autonomy-supportive feedback. Recognition of thoughtful suggestions, persistence, and responsible experimentation can strengthen employees’ confidence, optimism, and resilience. By creating a psychologically safe communication climate, managers can help employees translate psychological resources into observable innovative behavior.

For SOCAP, the weaker moderation result suggests that managers may not need to intervene heavily in every relational exchange on ESM. Social capital may often be mobilized through peer norms, routine collaboration, and everyday knowledge exchange. Therefore, the managers’ role is less to control these interactions directly and more to maintain the conditions under which they can function well. This includes encouraging respectful participation, preventing knowledge silos, and ensuring that useful expertise and connections remain visible across the organization.

Managers can also use ESM activity patterns to diagnose collaboration problems. Repeated questions, inactive groups, low participation from certain departments, or unresolved discussion threads may indicate that knowledge is not being captured or circulated effectively. In such cases, simply asking employees to “use the platform more” is unlikely to solve the problem. More targeted actions may include improving tagging rules, assigning responsibility for knowledge curation, creating cross-unit projects with clear ownership, and making high-quality contributions easier to find and reuse.

5.3.3. Implications for Employees and Work Teams

The findings also suggest that employees should not view ESM only as an administrative or communication channel, but as a tool for building the social, human, and psychological resources that support innovative performance.

First, employees can use task-oriented ESM to strengthen HUCAP. By participating in problem-solving threads, asking for feedback, sharing technical explanations, comparing alternative solutions, and reusing archived knowledge, employees can improve their professional competence and contribute to collective learning. High-quality participation is more valuable than frequent but shallow posting. Employees who explain their reasoning and document useful lessons can make their knowledge more visible to others while also strengthening their own learning.

Second, employees can use social-oriented ESM to develop SOCAP and PSYCAP. Informal interaction can help employees build trust with colleagues outside their immediate work unit, maintain weak ties, and gain access to diverse perspectives. These connections may be especially useful for innovation because new ideas often require nonredundant knowledge and cross-boundary support. Social interaction on ESM can also provide encouragement and emotional support, which may help employees remain confident and resilient when innovation attempts are uncertain or difficult.

Finally, employees can actively seek empowering support when attempting to turn their resources into innovation. For example, they can use ESM to share early-stage ideas, explain problems encountered during experimentation, request feedback from colleagues, and seek managerial support for promising solutions. By making their learning process and innovation attempts visible, employees may be more likely to receive autonomy, encouragement, and resources from leaders. In this way, ESM can help employees not only accumulate resources but also mobilize those resources toward concrete innovative behavior.

6. Conclusions

Overall, this study shows that ESM creates innovation value primarily by building employees’ social, human, and psychological capital, and that EL helps translate these types of capital into innovative performance. By distinguishing task-oriented and social-oriented ESM use and integrating a multi-capital mechanism with a leadership boundary condition, the findings offer a resource-centric and context-sensitive explanation of how digital collaboration tools can foster employee innovation. The following sections summarize this study’s main contributions and outline its limitations and future research directions.

6.1. Contributions

6.1.1. Theoretical Contributions

This study makes three incremental contributions to the literature on ESM and employee innovation:

• It distinguishes between task-oriented and social-oriented ESM use and shows their differentiated relationships with employees’ social, human, and psychological capital. This distinction addresses a gap in prior studies that have often focused on overall ESM use while paying less attention to the heterogeneous effects of different use orientations.
• It develops and tests a multi-capital indirect mediation framework, thereby clarifying the “black box” between ESM use and employee innovation. By examining SOCAP, HUCAP, and PSYCAP simultaneously, this study provides a more integrated explanation for the mixed findings in prior ESM–innovation research.
• It identifies the differentiated boundary role of EL in the resource-to-innovation process. The results suggest that EL is especially important for helping employees convert HUCAP and PSYCAP into innovative performance, while its moderating role in the SOCAP–INNO relationship is weaker. This extends the contingent view of EL in the context of digital collaborative innovation.

6.1.2. Practical Contributions

This study also provides targeted practical guidance for digital collaborative innovation management in knowledge-intensive enterprises:

• For organizations, the findings suggest that ESM should be positioned as a resource-building infrastructure for innovation rather than merely as a communication tool. ESM evaluation should therefore focus not only on surface activity indicators, but also on whether the platform supports the accumulation of SOCAP, HUCAP, and PSYCAP.
• For managers, the findings clarify the practical focus of EL. Managers should reduce the perceived risk of experimentation and support employees in applying their knowledge and psychological resources to innovation rather than over-intervening in everyday social interaction and collaboration on ESM.
• For employees, the findings suggest a strategic path for ESM use. Employees can support their innovative performance by combining high-quality task-oriented participation with appropriate social-oriented interaction, thereby accumulating the social, human, and psychological resources required for innovation.

6.2. Limitations and Future Research

This study has several limitations that provide clear opportunities for future research.

First, the data were collected through a single-source self-report cross-sectional survey, which may limit causal inference and raise concerns about common method bias. Although we adopted procedural and statistical remedies to alleviate this concern, future research could adopt a time-lagged longitudinal design to track the dynamic process of TESM/SESM promoting the gradual accumulation of SOCAP, HUCAP, and PSYCAP over time. Meanwhile, future studies can combine objective behavioral data from enterprise ESM systems (such as the content, frequency, and interaction quality of TESM/SESM posts) with supervisor-rated innovation performance to further strengthen causal inference.

Second, the sample was drawn from knowledge-intensive enterprises in China. This context is theoretically appropriate because innovation in such firms depends heavily on knowledge exchange, cross-boundary collaboration, and employee autonomous initiative. However, the generalizability of the findings to other institutional environments, industries, and organizational types needs to be further verified. Future studies could examine whether the dual-oriented ESM framework and the moderating role of EL remain valid in more individualistic cultural contexts, labor-intensive industries, or small- and medium-sized enterprises with low digital maturity.

Third, this study focused on the overall orientations of ESM use rather than the specific platform affordances and governance mechanisms that shape different use orientations. In practice, ESM platforms differ in functional designs such as content persistence, searchability, visibility control, and recommendation algorithms, which may have heterogeneous effects on TESM and SESM. Future research could examine how these specific platform affordances differentially facilitate the accumulation of SOCAP, HUCAP, and PSYCAP. It is also valuable to explore additional boundary conditions, such as task interdependence, team psychological safety, and network diversity, to further clarify when ESM-based resources are more likely to translate into innovative performance.

Despite these limitations, this study suggests that the value of ESM lies not simply in platform use itself, but in the employee resources that such use helps to build. By shifting attention from activity-based use to resource-building processes, the findings provide a more nuanced explanation of ESM-enabled innovation and offer practical guidance for designing digital collaboration practices that better support employees’ innovative performance.