SIMILARITY-ATTRACTION THEORY AND FEEDBACK-SEEKING BEHAVIOR AT WORK: HOW DO THEY IMPACT EMPLOYABILITY?

Employees’ feedback-seeking networks at work are important for employees’ learning and employability. Earlier studies often neglected the specific characteristics of the different relationships an individual employee has at work. We conduct social network analyses in seven samples to study inter-individual differences in feedback-seeking relationships in detail. We investigate 2,058 feedback-seeking relationships of 118 employees to study how similarity-attraction affects the composition of feedback-seeking networks at work and how the composition of these feedback-seeking networks influences employees’ employability. This research study aims to contribute by taking into account both the mechanisms that shape feedback-seeking networks and the effects of this on employability. The results show that similarity-attraction affects feedback-seeking in the workplace and that having a largely homogeneous feedback-seeking network has detrimental effects on employability.


Similarity-attraction and feedback-seeking at work: How do they impact employability?
Employability has recently received a lot of political and academic attention (Lu et al., 2015;Peeters et al., 2017). How can the competences that empower individuals to maintain, create, or obtain adequate jobs be developed ? Previous research studied various forms of learning activities and how they impact employability (e.g., Froehlich et al., 2015;Van der Heijden et al., 2009;Van der Heijden & Bakker, 2011). In this literature, feedback-seeking stands out as a major contributor (Anseel et al., 2013).
To derive meaningful interventions for practice, we need to comprehend the specific feedback-seeking relationships individuals engage in (cf. Wu et al., 2013) and how these feedback-seeking networks impact individuals' employability. So far, research has focused on the broad relationship between engagement in (social) learning activities and employability (Froehlich et al., 2019;Froehlich & Beausaert, 2014;Gerken et al., 2016;Lecat et al., 2018;Van der Heijden & Bakker, 2011) or, for instance, job (in)security (Van Hootegem & De Witte, 2019). An investigation of the specific network ties is often more useful than studying relationships at only a very general level (Reagans et al., 2004). This is especially true for social settings marked by heterogeneity, where individual differences in terms of pluralistic backgrounds, competences, or perspectives are potential catalysts for reflection and growth. Therefore, we pose the following research question: How does the composition and usage of individuals' feedback-seeking networks affect the feedback-seekers' employability?
The text makes two major contributions. First, by building on similarityattraction theory (Byrne, 1971), we take into account both the mechanisms that shape feedback-seeking networks at work and the effects of these on employability. This information complements previous research findings on how social informal learning translates into employability  and how interactions may help to get a job after higher education (Batistic & Tymon, 2017;Chen, 2017). Second, we argue that previous research's focus on the quantity of feedback-seeking is limiting, as other strands of research suggest that the quality of the feedback received and the attributes of the source of feedback also matter (e.g., London, 1995). The data gathered for this study, which includes 2,058 feedback-seeking relationships of 118 employees in seven complete organizational networks from three countries, make it possible to study the contributions of specific configurations of feedback-seeking networks in an international context.

Theoretical background
Competence-based employability In this study, we view employability as an individual's ability to continuously fulfill, acquire, or create work through the optimal use of competences . How can competences be used to fulfill, acquire, or create work? At least two theoretical lenses offer an explanation. First, human capital theory (Becker, 1993) looks at an individual's competences as malleable resources that determine the individual's capacity and quality of work and hence also as important resources to offer on the labor market. Therefore, time spent learning and in education are useful investments. The strong relationship between human capital and (perceived) employability also received empirical support (Berntson et al., 2006). The resource-based view of a firm (Barney, 1991) emphasizes an employer's view on the same matter: individuals who possess important competences are regarded as valuable resources for the company (Van der Heijde & Van der Heijden, 2006). Thus, they should be invested in. Both perspectives indicate that individuals competent in areas that are in demand are very attractive to the labor market. This emphasizes the role of competences in individual employability.
What are these in-demand competences? In line with other research (De Vos et al., 2011;Froehlich et al., 2018), we see employees as employable if they not only have relevant technical knowledge, but also are attentive to changes in the work context and adequately adapt to changes imposed on them. This means we focus on : occupational expertise (technical domain knowledge); anticipation and optimization (proactive, self-initiated screening and preparation for potential changes in job and career requirements and conditions); and personal flexibility (reactive adaptation to change).

Developing competence-based employability via seeking feedback
Knowing about these competences is one thing. But how are they to be developed and maintained? Previous research on workplace learning suggests that seeking feedback from others-the inquiry for information targeted at evaluating and reflecting upon work processes and the self (Anseel et al., 2007)-is especially important for this. Indeed, feedback-seeking has been related to many positive outcomes, such as goal attainment (Ammons, 1956), managerial effectiveness (Ashford & Tsui, 1991), performance (Kluger & DeNisi, 1996), a more accurate self-view (Ashford et al., 2003), career success (Cheramie, 2013;Van der Rijt et al., 2012a), and employability .
A useful lens to understanding learning from others is offered by social capital theory (Burt, 2005). Connecting to other people gives access to their human capital and other resources, such as their personal networks, which 80 can be used to develop one's own competences. Put differently, through seeking feedback individuals may acquire developmental resources to enhance their own human capital and, hence, their employability.
However, Ilgen et al. (1979) and Vancouver and Morrison (1995) note that it is difficult to separate the effects of feedback-seeking from the source of feedback-seeking. A network of contacts represents potential resources-an individual needs to become active and ask for feedback to actually access these resources (Ashford & Cummings, 1985;Frieling & Froehlich, 2017;Harwood & Froehlich, 2017). One important mechanism here is homophily-the tendency to build relationships with similar others (McPherson et al., 2001;Mello & Delise, 2015). Previous research offers two theoretical explanations for this tendency. Social categorization theory posits that employees categorize their colleagues and themselves based on observable characteristics such as gender (Ibarra, 1992;Louch, 2000) or age (Tsui & O'Reilly, 1989;Zenger & Lawrence, 1989). Similarity-attraction theory adds that people tend to interact with similar others in an attempt to reduce the potential for discomfort (Byrne, 1971;Standifer et al., 2013). Consequently, personal networks are often rather homogeneous in terms of these demographic characteristics. Empirical evidence for such effects has been found, for instance, in performance evaluation situations (Ferris et al., 1991) or when reporting job satisfaction (Peccei & Lee, 2005). Given homophily's evident influence on the development of interpersonal networks in the workplace (Carpenter et al., 2012), we hypothesize that homophily also affects feedback-seeking. Aligned with the previous studies on homophily and feedback-seeking mentioned above, the focal attributes of this study are age, tenure, function, and gender.
Hypothesis 1a: Employees seek more feedback from others who are similar to themselves in terms of age, tenure, function, and gender.
While many studies have found positive outcomes of feedback-seeking, Mulder and Ellinger (2013) have observed conflicting findings. They attribute this to varying conceptualizations of what constitutes good quality feedback. This statement is in line with the findings of Van der Rijt et al. (2012b), which show that the usefulness of the feedback received had a much greater impact than the quantity of feedback on perceived career development in their sample of financial experts. This evidence suggests that the usefulness of the feedback received needs to be considered in addition to the frequency of feedbackseeking. This is especially true given that feedback can have devastatingly negative effects if not delivered in constructive ways (Wang et al., 2015).
As mentioned before, similarity-attraction theory explains homophily via the reduced potential to experience discomfort or conflict in an interaction between similar individuals (Byrne, 1971). Relatedly, empirical research linked homophily to increased credibility (Wright, 2000), trust (Winter & Mitesh, 81 2013), and ease of communication (Zenger & Lawrence, 1989). More generally, this suggests that the potential for common understanding between two similar individuals is greater. This, in turn, makes the feedback received more likely to be perceived as more useful.
Hypothesis 1b: Employees rate feedback received from others who are similar to themselves in terms of age, tenure, function, and gender as more useful.
While the positive effects of seeking useful feedback are prevalent in the literature, much less focus has been put on the role of similarity-attraction. We posit, however, that similarity-attraction is an important constraint when considering feedback-seeking and its outcomes. This is because the process of feedback-seeking can be modeled as a process of experiencing uncertainty (e.g., a gap in one's knowledge), gathering additional information, and processing this information to reduce uncertainty and reach the respective goals (cf. Ashford et al., 2003). This raises the question about the nature of information that can be accessed in one's feedback-seeking network. We assume that the knowledge in one rather homogeneous group based on age, tenure, function, and gender is potentially more similar than in heterogeneous groups. Similar arguments have been brought forward elsewhere (e.g., Froehlich & Messmann, 2017;Granovetter, 1973;Phillips et al., 2006). From this perspective, individuals who are heavily influenced by similarity-attraction constrain themselves to only a limited pool of resources. In other words, what differentiates employees with heterogeneous feedback-seeking networks from employees with homogeneous feedback-seeking networks is that the former have greater potential to access information that is instrumental in reducing uncertainty and contributing to reaching a set goal. Hence, we hypothesize that individuals learn less for their own employability if their feedback-seeking network is characterized by similarity-attraction. Similar arguments have been advanced in other fields, such as in the context of top management teams' advice-seeking to make strategic decisions (Alexiev et al., 2010).
Hypothesis 2: Homophilous feedback-seeking networks relate negatively to employability in terms of occupational expertise, anticipation and optimization, and personal flexibility.

Sample
We test the model in a population of knowledge-intensive workers. The major reason for the choice of this population is that the dynamic nature of knowledge-intensive work (Froehlich, 2017) may make the need to update competences even more pronounced. Given the sensitivity of social network analysis to missing data (Froehlich & Brouwer, 2021), we sent out questionnaires to a convenience sample of seven complete organizational entities (see Table 1). To keep the findings generalizable across sectors and countries, the sample includes research units (aca.at, aca.nl ), consulting firms (con.in, con.at), a team on an innovative project at a bank (ban.at), and a startup (soc.at) in Austria, the Netherlands, and India (Note: The consulting teams are project teams of a large IT implementation program that work on-site in Austria-con.at-and off-site in India-con.in-for the same client.). After sending two reminders to the potential participants, we achieved a sufficient response rate of 92%. In total, the data of 2,058 feedback-seeking relationships, or ties, and 118 employees were available for analysis.  (10) 40 (12) 34 (4) 34 (8) 33 (10) 43 (10) 37 (12) Experience on job in months (Mean + SD) 127 (115) 106 (99) 11 (15) 30 (27) 37 (24) 127 (111) 21 (10) Instruments We collected both psychometric and sociometric data via the survey. For the sociometric data, we presented each respondent with a roster of all employees (including supervisors) in the respective workgroup or project team.
We focus on feedback-seeking via inquiry, as it is "a more explicit behavioral choice" (Ashford et al., 2003, p. 788), and asked for the frequency of feedbackseeking (1 = (almost) never, 7 = several times a day). Each person also estimated how often they were approached by another person for feedback. We used the mean of these two values as a more reliable estimate of the frequency of feedbackseeking for all relationships and in both directions (raters diverged only 0.5 points on average). Furthermore, we asked for the usefulness of the feedback received from every other person in the network (1 = (almost) not useful, 5 = very useful ).
We assessed the similarity of each pair of employees based on the demographic and work-related characteristics of the respondents: age, tenure, function, and gender. For each tie, we evaluated whether the employees are of similar age (+/-one SD of the group), have a similar tenure (+/-one SD of the group), work in similar functions (based on their self-reported role description, which is not linked to their departmental affiliation), or identify as the same gender. The SD was chosen to capture the differences in terms of group composition across samples. This information (0 = dissimilar, 1 = similar) was used to create networks based on similarity for each attribute under study. We then extracted two feedback-seeking networks: one that includes the frequencies of feedback-seeking only between similar employees for each respective characteristic and one that includes all frequencies of feedback-seeking. We calculated the weighted out-degree centrality for all nodes in both networks. Weighted out-degree centrality denotes the number of outgoing ties each node has, multiplied by the respective frequency of interaction. We then set the weighted out-degree centrality based on similar persons in relation to the total weighted out-degree centrality of the individual in terms of feedback-seeking frequency for each characteristic. The result is the ratio of sources of feedback that are similar to the employee weighted by the frequency of feedback-seeking. This figure ranges from zero (all sources of feedback are different) to one (all sources of feedback are similar).
We gauged employability using three validated subscales by Van der Heijde and Van der Heijden (2006): occupational expertise (Sample item: "I consider myself competent to engage in in-depth, specialist discussions in my job domain."), anticipation and optimization (Sample item: "I take responsibility for maintaining my labor market value."), and personal flexibility (Sample item: "I adapt to developments within my organization."). Respondents answered on a 5-point Likert-scale (1 = almost never, 5 = very often). The scales achieved satisfactory Cronbach's alphas (α = 0.92, 0.79, and 0.71, respectively).

Analyses
To test Hypothesis 1, we used the Multicollinearity Robust Quadratic Assignment Procedure (MRQAP) to estimate regression coefficients. MRQAP is a standard procedure to analyze autocorrelated structures, such as social network data (Dekker et al., 2003). After calculating the standardized regression coefficients (β), standard errors, and p-values for each sample, we performed a meta-analysis to estimate the effects shown in the MRQAPanalyses across all samples. For this purpose, we calculated three statistics: the Q-statistic (which assesses heterogeneity across the samples; Krackhardt & Kilduff, 1999), the summary effect (β + ), and the overall level of significance.
We performed multiple regression analyses to test Hypothesis 2, using chronological age, tenure, gender, and sectoral background (dummy coded) as control variables. To aid the presentation of the results, we also highlight the relationships significant at an α-level of 0.10. This, however, will not change the interpretation of the general pattern of effects found.

Results
In this section, we present the analyses that answer our research question in three steps. First, we provide summary statistics across all samples that relate feedback-seeking to demographic characteristics. Second, we explore these relationships more deeply by also considering the results of individual samples. Last, we investigate the relationship between individuals' feedback-seeking networks and their competence-based employability.
We hypothesized that employees seek more feedback from colleagues and supervisors who are similar to themselves in terms of age, tenure, function, and gender. Table 2 shows the results of the meta-analysis. The Q-statistics, which indicate heterogeneity across samples, show that heterogeneity was found only for the relationship between same function and feedback-seeking frequency (Q = 21.54, p < 0.01). Put differently, for this relationship, the results vary across the samples. Additionally, we found a strong summary effect on feedback-seeking frequency of same function (β + = 0.30, p < 0.01). This means that employees in all samples were more likely to seek feedback from colleagues and supervisors who work in a similar function. This supports Hypothesis 1a only partially. Concerning the perceived usefulness of the feedback received, the effect sizes were homogeneous across all independent variables. We found a strong summary effect of same function (β + = 0.19, p < 0.01). This supports Hypothesis 1b for function.  Table 3 shows the results of the individual samples. It shows that employees seek feedback more often from colleagues and supervisors who work in the same function (five samples; from β = 0.23, p < 0.05 to β = 0.39, p < 0.01) and are of the same gender (two samples, from β = 0.17, p < 0.05 to β = 0.29, p < 0.01). Conversely, employees addressed colleagues and supervisors of a different age more readily in the startup (β = 0.14, p < 0.10). For colleagues and supervisors with a similar tenure, we found both positive (con.at: β = 0.08, p < 0.05) and negative (con.in: β = −0.15, p < 0.10) relationships with feedbackseeking frequency. The independent variables explained up to 20% of the variance in feedback-seeking frequency. Table 3 Standardized

results of the Matrix Multiple Regression via Dekker Double Semi-Partialing Multiple Regression Quadratic Assignment analyses on feedback-seeking frequency and perceived usefulness of feedback
Sample aca.at aca.nl con.in con.at con.nl ban.at soc.at The patterns of effects on the perceived usefulness of feedback are similar. Employees perceive feedback as more useful from colleagues and supervisors who work in the same function (six samples; from β = 0.17, p < 0.05 to β = 0.45, p < 0.01) and who have a similar tenure (two samples; from β = 0.18, p < 0.10 to β = 0.19, p < 0.05). We found mixed results for colleagues and supervisors of a similar age (soc.at: β = 0.17, p < 0.05; con.at: β = −0.13, p < 0.10) and the same gender (aca.nl: β = 0.16, p < 0.01; con.nl: β = 0.20, p < 0.01; con.at: β = −0.16, p < 0.10). The independent variables explained up to 23% of the variance in perceived usefulness of feedback (in aca.at).
We were also interested in the relationships between feedback-seeking and employability. In the regression analyses (Table 4), we found that similarityattraction in terms of function relates negatively to occupational expertise (β = −0.18, p < 0.10), anticipation and optimization (β = −0.23, p < 0.05), and personal flexibility (β = −0.27, p < 0.05). For the composite measure of employability, a similar effect-size was noted (β = −0.27, p < 0.05). Similarityattraction in terms of age relates negatively to occupational expertise (β = −0.29, p < 0.10). These findings suggest that feedback-seeking networks largely composed of people of similar ages and who work in similar functions relative to the feedback-seeker may be less helpful for improving their employability. Therefore, Hypothesis 2 is partially confirmed. Conversely, similarity-attraction in terms of gender relates positively to occupational expertise (β = 0.24, p < 0.05). As concerns the control variables, we note negative effects of chronological age on occupational expertise (β = −0.45, p < 0.05) and differences between the participating organizations. In total, the model explains 20% of the variance in the employability scales (R ² = 0.15 to 0.21).

Discussion
We set out to investigate how similarity-attraction relates to feedback-seeking in the workplace and employees' employability. We hypothesized that employees seek feedback from colleagues and supervisors like themselves. We proposed that the feedback received from similar sources is perceived to be more useful. In this section, we discuss the findings in light of their contributions to (a) similarity-attraction and feedback-seeking research and (b) employability and competence development research.

Contributions to similarity-attraction and feedback-seeking research
Our meta-analysis across seven organizations shows that employees seek more feedback from colleagues and supervisors who work in a similar function and that the feedback received from these sources is perceived as more useful. This is in line with previous theorizing about similarity-attraction (Byrne, 1971;Lazarsfeld & Merton, 1954;McPherson et al., 2001) and empirical research about homophily (Tsui & O'Reilly, 1989;Wexley et al., 1980). Our data agrees with earlier findings that network ties are very much structured by intraorganizational boundaries such as functional organizational structures (Feld, 1981;Kleinbaum et al., 2013). Like earlier studies, the results show different effects for the different variables used to test similarity (Goldberg et al., 2010).
This part of the study also revealed an interesting secondary finding about the relationship between feedback quantity and feedback quality. Unlike previous research (e.g., Van der Rijt et al., 2012b), the patterns of findings for feedback-seeking frequency and perceived usefulness of feedback received were remarkably similar. This suggests the (very rational) behavior that employees seek feedback more often from colleagues and supervisors from whom they have received useful feedback in the past. This leads to a convergence of the two concepts.

Contributions to employability and competence development research
The data also shows that having a homogeneous feedback-seeking network has detrimental effects on employees' employability and the competences it contains. We find a negative relationship between homogeneous networks in terms of function and all three dimensions of employability tested. This suggests that if employees constrain themselves to similar colleagues and supervisors, they do not make use of the full potential of the existing inter-individual differences. Colleagues or supervisors different in terms of their demographic or functional background may be precious sources of feedback as well-people with different backgrounds have rather different ways of thinking (Chua, 2018). Here, we contribute to the literature of developmental networks. For example, Higgins and Thomas (2001) called for further examination of diversity within developmental relationships such as feedback-seeking relationships.
The relationship between feedback-seeking network composition and employability is less pronounced for occupational expertise. One reason for this might be that colleagues outside the same function have access to different information. Therefore, a feedback-seeking network that consists mainly of colleagues in the same function has limited potential to access information outside the main domain of work (e.g., upcoming trends). When it comes to the technical expertise needed for the tasks at hand, this information is less relevant. The sociometric data reveal that the employees themselves judge the feedback received from their colleagues in the same function as more useful. This contradiction may result from the different motives for seeking feedback. Ashford et al. (2003) distinguish instrumental, ego-based, and image-base motives, of which instrumental feedback-seeking is most conducive to developing competences. However, employees may seek feedback from similar sources for other motives, which may explain the diverging results.
We found a negative relationship between homogeneous feedback-seeking networks in terms of age and occupational expertise. One possible reason for this is that knowledge and skills are different but compatible across age groups. For instance, younger employees may contribute important knowledge of how to use IT more efficiently, while older employees may give helpful advice for social relationships and general work conduct. Eventually, all these competences are needed to remain employable. Therefore, seeking feedback from colleagues and supervisors of other age groups can be beneficial (Mannix & Neale, 2005). In any case, the findings need to be cautiously interpreted, as age has been found to moderate the relationship between learning and employability ( Van der Heijden et al., 2016).
As concerns homogeneous networks in terms of gender, we found a positive effect on occupational expertise. This finding is not in line with our hypothesis and stands in contrast with the other empirical evidence we have generated. But it still can be explained, given that homogeneous relationships also have advantages. For example, interaction with similar partners is associated with lower transaction and communication costs, as both parties may share similar experiences and have more common ground. This may explain the positive effect between homogeneous networks in terms of gender and occupational expertise. Future research may investigate these competing mechanisms (similarity lowers communication costs but also potential value, as outlined above).
The results complement the findings about the antecedents and consequences of advice-seeking relationships prevalent in social network research (Borgatti & Cross, 2003). This is interesting, as seeking feedback is conceptually very different from seeking advice and we cannot just assume similar antecedents and outcomes of different types of networks. For instance, feedback-seeking focuses on the past, while advice-seeking centers on upcoming challenges. The irreversible nature of past actions, however, may leave the feedback-seeker even more vulnerable to others' judgments and might lead to a more intimate, more closed-off network.

Limitations and directions for future research
The meta-analytic strategy applied in this research made it possible to find effects across the samples. The MRQAP analyses of individual samples, however, show varying patterns of effects in the samples. This exploratory research cannot fully explain why these differences exist. Specifically, while there are many potential differences between the groups-for instance, in terms of climate concerning diversity (Chrobot-Mason & Aramovich, 2013) or learning )-our research does not capture all of these variables. Interestingly, while our model explained a significant part of the variance in the MRQAP analyses in six samples, the effects were less pronounced for the banking sample. One apparent difference between this particular sample and the other samples is the content of work: ban.at is a steering committee with its members spread across organizational units. Consequently, their work is less focused on technical details, but more on coordinating and managing. Also, banks operate in a highly regulated environment and may face less uncertainty than the more entrepreneurial organizations taking part in this study. This warrants other sets of competences, ways of collaborating, and making decisions. Future research may consider different sectors, types of tasks, and ways of working.
Also, we cannot assess the role of feedback-seeking relationships that employees may have outside their organizational entity. This may represent a problem given our conceptualization of employability, which also includes the more generic competences of personal flexibility and anticipation and optimization. However, social network analysis demands that boundaries are set that define who is in the network and who is not, and focusing on workgroups is a common approach to that (Cummings & Cross, 2003). According to focus theory (Feld, 1981), individuals who share the same focus (in this case, the same organizational entity) are more likely to interact with each other. This is even more true given the constraining nature of this particular focus (i.e., individuals mostly have no choice who else is part of their respective organizational entity). Therefore, we decided to focus on the feedback-seeking relationships within organizational entities-capturing all relevant interactions is impossible, but we at least collect information about the relationships that are most likely to be of major influence. The strong effects of homogeneity in terms of function on the frequency and perceived usefulness of feedback-seeking empirically support this assumption. Nevertheless, future research may also study feedback-seeking across organizational boundaries. For example, acquaintances who work at different departments or organizations, members of the same sports club, and family members may also be good sources for feedback. In our case, confining data collection to the respective organizational units was also important in terms of limiting confounding effects of proximity. Proximity has received some attention in previous research (e.g., Ibarra & Andrews, 1993;Rice & Aydin, 1991), especially when linked to homophily (Monge & Contractor, 2003;Skerlavaj et al., 2010). However, in this study, the participating organizational entities are small and all employees are relatively close to each other. Also, for SIMILARITY-ATTRACTION THEORY AND FEEDBACK-SEEKING BEHAVIOR ... two (additional) Belgian samples that were used for pilot testing, we studied the office plans to see whether proximity should be taken into accountwith negative results. In sum, this gives us confidence in stating that proximity does not confound the results pervasively. However, future research may also make use of approaches that are more liberal with respect to network boundaries-for instance, ego-social network analyses or qualitative (Hollstein & Straus, 2006) and mixed methods approaches to social network analysis Froehlich & Van Waes et al., 2020).
The measurement of constructs used presents another challenge. The self-reported scores of feedback-seeking and employability may raise concerns about common method variance (Podsakoff & Organ, 1986). However, selfappraised employability scores are well supported by theory and empirical evidence (Dries et al., 2014). Also, both the source and the target were asked about the frequency of the feedback-seeking relationship. Nevertheless, future research may use different methods and different sources, such as observations or tracking devices. After all, previous research found employees to rate themselves higher than their employers did (Van der Heijde & Van der Heijden, 2006) and higher than their labor market success would suggest (Patrickson & Ranzijn, 2003).

Practical implications
The results indicate that similarity affects employees' feedback-seeking networks and, subsequently, their employability. This is important to consider, especially because managers often attempt to assemble heterogeneous teams in the hopes of increasing the group's potential to innovate (Miura, 2004). To accompany such measures, this research suggests that it may be advisable to educate the employees about the potential pitfalls of homophily by offering training and sessions for reflection. The aim here could be to encourage employees to use the full potential of their network and reduce the insecurities of contacting "other" people. For instance, this may include assigning tasks to pairs of previously unrelated colleagues with different backgrounds or a general awareness training about one's social network in the workplace. Mentoring or job shadowing programs are also useful interventions. In that respect, the use of social network analysis may be considered as an apt managerial tool to diagnose existing social networks in the workplace and to design specific interventions (Palonen & Froehlich, 2020).

Conclusion
While earlier research has investigated the effects of feedback-seeking on employability, we extend the body of literature by considering both the sources of feedback-seeking and the composition of employees' feedback-seeking networks. The findings suggest that employees are more likely to ask colleagues and supervisors similar to themselves for feedback. However, having a highly homogeneous feedback-seeking network is not always beneficial for one's employability.