1. Introduction
Tacit knowledge (TK) resides within human carriers in different units of a sustainable company and these units represent various stocks of TK [
1]. TK is an important strategic resource for sustainable development of a company [
2,
3]. The sustainable development of a company relies on TK acquisition and integration across these units (i.e., subsidiary-subsidiary, parent-subsidiary, and subsidiary-parent) [
1]. In general, a receiver acquires cross-unit TK by actively observing the role modeling of its carrier from other units through sustainable interactions based on apprentice-like relationship [
1]. During the apprentice-like interaction process, only the receiver who appreciates or likes (i.e., has positive affection toward) the carrier will have motivation to actively observe the carrier as a role model. The receiver collects more observed practices and cognitively processes them to absorb TK [
4]. For the receiver, the sustainable process of tacit knowledge acquisition (TKA) involves two interrelated mechanisms: affective mechanism that triggers the receiver’s observation motivation and cognitive mechanism that captures his information processing. However, most TK management studies [
5,
6] only focus on receivers’ cognitive mechanism of TKA but pay little attention to their affective mechanism of TKA. Accordingly, to improve TKA, important research questions are considered: “How can affective mechanism be improved to motivate a receiver to actively observe TK carrier across units in a sustainable organization?” and “How can cognitive mechanism be improved to enable a receiver to effectively absorb TK in the sustainable observation?”
Most extant studies have examined how interpersonal interaction-related factors promote TKA by affecting receivers’ cognitive mechanisms in a sustainable organization [
6,
7,
8]. Interpersonal interaction is only a desirable context that facilitates TK senders, like mentors, to show role modeling for receivers to transfer TK. In this light, the senders’ mentoring or role modeling embedded in interpersonal interaction might be more important than the interaction by itself in improving TKA in a sustainable company [
9]. However, few studies are concerned on these effects. Furthermore, extant studies of TKA have mainly identified socio-contextual factors as predictors of TKA in an offline context [
6,
7,
8,
10]. Little attention has been paid on the effects of IT-enabled factors or IT artifacts on TKA in a sustainable organization. To fill these gaps, this study examines how social media-enabled mentoring affects TKA of a sustainable company by affecting receivers’ affective and cognitive mechanisms.
Social media enabled-mentoring is a new type of e-mentoring that involves the use of IT (e.g., social media) to facilitate online interpersonal interactions in which knowledge senders provide support and role modeling to help receivers learn new knowledge [
11,
12,
13]. This type of mentoring widely exists in universities, sustainable enterprises, and virtual communities [
14,
15,
16]. For instance, teachers use social media, such as Facebook, as learning platforms to provide mentoring for students in distance sustainable learning [
14,
17,
18,
19]. In a sustainable company, cross-unit TK is difficult to be acquired as unit boundaries or geographical gap inhibits cross-unit interpersonal interaction [
4]. Social media is a network technology-based social tool (e.g., WeChat, Facebook, Twitter, micro-blog, and instagram), which facilitates sustainable social interaction in online context [
20]. Social media enables employees to have great flexibility in forming cross-unit mentorship [
12]. Employees can use social media to contact colleagues in other units as “friends.” By viewing one another’s profiles and posters, they select appropriate mentors (or mentees) among these “friends” [
13,
21]. Furthermore, social media enables a mentor to implement sustainable mentoring by using “voice message” or “text” to provide feedback or support and use “pictures” or “video calls” to synchronously show role modeling without geographic restrictions and unit boundaries [
13].
This study integrates cognitive and affective learning theory (CALT) [
22,
23] from the training evaluation field and media-dependent perspective (MDP) [
24] of the IS field to investigate how and why social media-enabled mentoring improves sustainable TKA in a company. CALT describes how training-related factors affect learning outcomes through learners’ sustainable cognitive learning (e.g., coping with cognitive states that are used to process learning content) and affective learning processes (e.g., coping with affective states that motivate learners to learn) [
22,
23]. We expect that social media enabled-mentoring affects employees’ TKA by improving mentees’ liking of mentors (i.e., affective learning) and shared mental model (SMM) (i.e., cognitive learning). According to MDP, the effectiveness of social media-enabled mentoring should depend on the fit between mentoring and information and communication technology (ICT) capabilities of social media [
24]. Therefore, we propose two key ICT capabilities of social media, namely, social presence and synchronicity [
21,
24,
25], to moderate the effects of social media-enabled mentoring on the two sustainable learning mechanisms (mediators) and TKA.
Our research suggests that the strengths of social media-mediated mentoring in facilitating TKA would not be practical offline. For instance, we have found that social media-enabled mentoring increases TKA through shaping SMM only when synchronicity is high. Synchronicity means that ICT supports fast information exchange [
24]. In the offline context, two employees from different units with different locations cannot synchronously exchange information; thus, mentoring is less likely to increase SMM and TKA of a sustainable company in the offline context. We also have found that social media-enabled mentoring increases TKA by improving liking (i.e., mentees’ liking of mentors) only when social presence is high. Social presence refers to the atmosphere of intimacy created by ICTs [
25]. In the offline context, a mentor with poor social and emotional expression skills often has difficulty in creating social presence when interacting with others. Under this context, his mentoring is less likely to improve the receiver’s liking and decrease their TKA. However, many ICT functions, such as emotion icons, can help him easily create social presence online [
26]. In a word, with the assistance of synchronicity and social presence, sustainable mentoring is more likely to increase TKA in the context of social media platform than in the offline context.
The remainder of this paper first introduces the theoretical background. Next, the research model and hypotheses are developed. Furthermore, sample, data collection, measure, data analysis are introduced. Finally, we discuss our findings, implications, and limitations.
4. Sample and Procedure
Data was gathered from 45 companies, 66% of the 68 randomly selected companies. We surveyed Chinese employees of the 45 companies. Survey instructions were mailed to human resource (HR) managers of these companies who helped with the survey logistics. Similar with many mentoring studies, we asked each employee to assess whether he/she has a closet mentor from other units within his/her company. Mentorship was split into informal and formal mentorship. This study focused on informal mentorship as mentors and mentees based on their own voluntary choice (i.e., informal mentorship), which were more active in interacting with each other than those assigned by their companies (i.e., formal mentorship) [
37]. All of these informal mentorships lasted for at least three months. Therefore, mentees had enough information to evaluate the mentoring activities. Participation was voluntary and all participants were offered gifts (RMB 40) as an incentive for participation in the online survey.
Three-wave longitudinal survey has been conducted to reduce common-method bias and provide evidence for the causal directions proposed [
50]. The same set of respondents was invited to finish the three waves of the online survey. Initially, 212 mentees of 53 companies completed the items for the control variables, social media-enabled mentoring, synchronicity, and presence. After six weeks, 178 individuals of 51 companies participated in the second-wave survey and completed the items for liking and SMM. After six weeks, a total of 156 mentees from 45 companies provided valid data and completed the items for TKA, resulting in a response rate of 73.6%.
Table 1 shows the demographics of the participants. The companies were distributed across various industries mainly including manufacturing, information technology services, high technology, wholesale and retail, real estate, business services, which improved sample representative.
We assessed the possibility of non-random attrition through multiple logistic regression analysis [
51]. A dichotomous variable (1 = attrition case, who responded in the first or first two waves; 2 = respondent) was regressed on the principal variables collected in the first wave. No significant regression coefficients were found, thus suggesting no serious non-random attrition problem.
5. Measures
The items for all studied variables were adopted from the English literature. The survey was executed in China, and thus, the questionnaire was translated into Chinese and back translated into English to ensure equivalence of meaning. The measures for the studied variables included 21 questions. All the items were measured by a seven-point Likert scale ranging from 1 = strongly disagree to 7 = strongly agree.
Specifically, to improve participants understanding of social media-enabled mentoring, its definition was introduced. Mentoring activities (i.e., work support, psychological support, role modeling) and social media platforms (e.g., Wechat, facebook, twitter, microblog, instagram, etc.) were enumerated, respectively. Every participant was asked to recall his most recent (in the past four weeks) experience of a mentoring mediated by the social media that is most frequently used by his mentor and himself. Each respondent was asked to fill out the instrument based on his most recent experience of the social media-enabled mentoring. The instrument explained to the respondents that all the questions related to “this social media” referred to the one most frequently used by his mentor when giving mentoring for him. Three items for the social media-mediated mentoring construct (i.e., social media usage for mentoring) were adapted from the scale of instant message usage at work in Ou and Davison’s (2011) study [
52]. Three items that were used to measure mentee liking of mentors were adapted from Jehn (1995)’s liking scale [
53]. Four items were adapted from Xiang et al. (2013) study to measure SMM [
54]. Following Ou et al. (2014), synchronicity was measured by using a two-item scale [
25]. Social presence was measured by a four-item scale adapted from Ou et al. (2014). TKA was measured using five items adapted from the TKA scale of Lyles and Salk’s (1996) study [
9]. After development of the measures, three Chinese mentees with experience of receiving social media-enabled mentoring reviewed the questions and provided feedback. A pilot test of 46 subjects was conducted to test the wordings of the instrument. Cronbach’s alpha of all studied scales ranged from 0.78 to 0.92, which indicated that the studied scales had acceptable reliabilities (>0.70). The main survey was then initiated.
Appendix A lists the final items used in the questionnaire.
6. Data Analysis
LISREL (version 8.70) [
55] and SPSS (version 17.0) were used to conduct the data analysis.
6.1. Measurement Model
The Cronbach’s alpha and composite reliability of all studied scales are higher than the acceptable value (>0.70) (see
Table 2). The results indicate that the reliability of the measures are satisfactory. We calculated the AVE (average variance extracted) of all studied variables to test convergent validity of all variables. The AVE for each variable was above the acceptable value (>50), indicating good convergent validity [
56]. Discriminant validity of all variables was tested by comparing the square root of AVE with the correlations of all studied variables. The square root of the AVE was larger than the correlations, indicating good discriminant validity for all variables [
56]
Second, CFA was conducted to assess the overall goodness of fit of the measurement model. We used six model-fit indices: the ratio of χ2 to degrees of freedom (df), Non-Normed Fit Index (NNFI), Normed Fit Index (NFI), comparative fit index (CFI), incremental fit index (IFI) and root mean square error of approximation (RMSEA). The results demonstrated that the measurement model containing six factors (i.e., Social media-enabled mentoring, social presence, synchronicity, SMM, liking, TKA) yielded a good fit (χ2/df = 1.84; CFI = 0.96; NFI = 0.93; NNFI = 0.96; IFI = 0.96; RMSEA = 0.07), and fit the data very well.
In addition, we adopted the chi-square difference test [
57] to compare the six-factor model with five alternative models that increase in complexity. The results in
Table 3 show that the fit of the six-factor model is significantly better than each of the five alternative models. This procedure demonstrates that our measures have good discriminant validity and exhibit a good fit with the data collected. Furthermore, the results minimize the possibility of common method bias, as a simple model does not fit the data as well as a more complex model [
50]. Moreover, we used “temporal separation of measures” method to create time lags between the predictors and dependent variables. This technique is particularly useful to reduce common method bias in the study of attitude-attitude relationships [
50].
Table 4 presents the means, standard deviations for all studied variables and VIF (variance inflation factor). The VIF (<3.33) for all the dependent variables are well below the levels that might indicate multicollinearity [
30].
6.2. Mediation Effect of Liking and SMM
We used multiple regression analysis of SPSS and followed the Baron and Kenny’s (1986) four-step approach to test whether liking and SMM mediate the relationship between social media-enabled mentoring and TKA [
58]. In addition to demographic variables (i.e., position, gender, age and education), cross-cultural training and international work experience were also included as control variables, as prior literature [
1,
6] suggested that the two variables will affect TKA among employees with international background.
Baron and Kenney (1986) posit four steps to test a mediation. First, independent variables (IVs) significantly affect dependent variables (DVs). Second, mediators significantly affect the DVs. Third, the IVs also significantly affect the mediators. Fourth, when the mediators enter into the regression equation, the relationships between the IVs and the DVs should be significantly weaken (partial mediation) or should become non-significant (full mediation). This study follows the four steps to test the mediation. First step, social media-enabled mentoring has a significant positive effect on TKA (β = 0.23, p < 0.01). Second step, social media-enabled mentoring has significant positive effects on liking and SMM, respectively (β = 0.27, p < 0.01; β = 0.23, p < 0.01). Third step, liking and SMM have significant positive effects on TKA (β = 0.36, p < 0.001; β = 0.27, p < 0.01). Fourth step, when controlling for liking and SMM, social media-enabled mentoring has no significant effect on TKA (β = 0.08, ns.), while liking and SMM still have significant and positive effects on TKA (β = 0.34, p < 0.001; β = 0.26, p < 0.01).This indicates that the positive effect of social media-enabled mentoring on TKA is fully mediated by liking and SMM. Moreover, the mediating effect of liking is stronger than that of SMM in the relationship between social media-enabled mentoring and TKA. Thus, H1 and H4 were supported. In total, the mediation model explains about 36% variance of TKA.
6.3. Moderating Effects of Social Presence and Synchronicity
We use moderated regression analysis in SPSS to test whether social presence moderates the effect of social-media enabled mentoring on liking; and whether synchronicity moderates the effect of social-media enabled mentoring on SMM. Results (see
Table 5) show that social presence significantly moderates the effect of social-media enabled mentoring on liking (β = 0.20,
p < 0.01). In addition, synchronicity significantly moderates the positive effect of social-media enabled mentoring on SMM (β = 0.23,
p < 0.01). Thus, H2 and H5 were supported. As explorative analyses, we found that synchronicity doesn’t moderate the effect of social-media enabled mentoring on liking (β = 0.02, ns.); and social presence doesn’t moderate the effect of social-media enabled mentoring on SMM (β = 0.12, ns.). To interpret the interactions, we plotted the interactions (see
Figure 2A,B) and examined the simple slopes. The positive relationship between social media-enabled mentoring and liking was significant only when social presence was high (b = 0.16,
p < 0.05) rather than when it was low (b= −0.09, ns.). The positive relationship between social media-enabled mentoring and SMM was significant when synchronicity was high (b = 0.26,
p < 0.01) rather than when it was low (b = −0.03, ns.)
6.4. Moderated Mediation Effects
In this study, the effect of social media-enabled mentoring (X) on SMM (M
1) is moderated by synchronicity (W), and the effect of social media-enabled mentoring (X) on liking (M
2) is moderated by social presence (V). As revealed by the regression Equations (1) and (2):
Subsequently, moderated mediation effects (also called “conditional indirect effects”) occur if the indirect effect of X (social media-enabled mentoring) on Y (TKA) through M
1 (SMM) is contingent on the level of W(synchronicity); and if the indirect effect of X on Y through M
2 (liking) is contingent on the level of V (social presence) [
59]. X to Y independent of M
1 and M
2 is not specified as moderated. If the indirect effect of X differs as a function of W or V, we can say that the mediation of X’s effect on Y by M
1 or M
2 is moderated by W or V—moderated mediation. As revealed by Equations (4) and (5), respectively.
Hayes’ (2013) bootstrapping approach (n boots =1000; 95% Bias corrected confidence interval) was used to test the conditional indirect effects (H3 and H6). Bootstrapping was found to be the most powerful method to detect conditional indirect effects [
59] and it has been used by many organizational behavior studies and the studies of social media [
60]. A confidence interval must not contain a zero to assume a significant mediation or conditional indirect effect [
59]. The bootstrapping analysis (see
Table 6) found that social media-enabled mentoring significantly improved TKA through liking only when social presence was high (+1 SD) (B = 0.09,
p < 0.05) [BC 95% CI; 0.02, 0.19] rather than when social presence was low (−1 SD) (B = −0.05, ns.) [BC 95% CI; −0.13, 0.02]. H3 was supported. As expected, social media-enabled mentoring significantly improved TKA through SMM only when synchronicity was high (+1 SD) (B = 0.12;
p < 0.05) [BC 95% CI; 0.04, 0.21] rather than when synchronicity was low (−1 SD) (B = −0.01, ns.) [BC 95% CI; −0.08, 0.05]. Thus, H6 was also supported.