Politics-based knowledge legitimation model: power exercise in organizational learning

Purpose – The purpose of this study is to propose a model of knowledge legitimation in organizational learningfocusing ontherelationship between power politics andlegitimacy. Design/methodology/approach – This studyadoptstheapproachofa conceptual discussion. Findings – This study developed an organizational learning model that explains how actors exercise their power and how knowledge is legitimated through politics. The author identi ﬁ ed various factors that shape the politics; these factors trigger, enhance, facilitate and inhibit power exercise. This study also identi ﬁ ed which type of power (in ﬂ uence, force,domination anddiscipline) leads to whichtype of legitimacy (pragmatic, moral and cognitive). Furthermore, this study found that power politics and organizational learning are interrelated; actors ’ powers bestow legitimacy on knowledge, and knowledge enhances the power of related actors. Originality/value – This study identi ﬁ ed the set of factors that shape actors ’ power exercise in organizational learning as well as their associated mechanism and illustrated how they lead to knowledge legitimation. The author also revealed the relationships between actors ’ power and legitimacy of knowledge. Finally, this study elaborated onthe ﬁ ndingsof priorstudiesconcerning politics oforganizationallearning.


Introduction
Organizational learning is not a simple process of incorporating "beneficial knowledge" (knowledge that benefits an organization). Power politics within an organization may intervene in the learning process (Dekker & Hans en, 2004). Knowledge is interpreted and filtered through political process. As a result, beneficial knowledge is not always integrated into an organization. Thus, politics that rule organizational learning process are the key factors that determine the trajectory of organizational change in the long term.
However, many organizational learning studies have not paid sufficient attention to this issue. Whereas some scholars (Karatas -Özkan & Murphy, 2010;Lawrence, Mauws, Dyck & Kleysen, 2005) claim that organizational learning is a political process, further examination is needed to understand how knowledge is advanced by key actors in the political process.
Through theoretical discussions based on the existing assertions made in the organizational learning literature, this paper examines the mechanisms of politics of organizational learning. In particular, we focus on the "interpretation" of knowledge, as it is at the core of organizational learning and determines the direction of the entire learning (Daft & Weick, 1984). Furthermore, we recognize that the legitimacy of knowledge is a requisite for interpretation, following Burgoyne and Jackson's (1997) claim that knowledge regarded as legitimate is maintained in an organization. Hence, we explore the mechanism through which knowledge is legitimated for learning.
The remainder of this article is organized as follows. First, to establish a baseline for the discussion, we review the basic perspectives of organizational learning. Then, we discuss legitimacy, which is expected to be the key for organizational learning. After presenting the research gap and our research questions, we elaborate on our politics-based knowledge legitimation model in organizational learning. Finally, we provide our conclusions.

Organizational learning
What is organizational learning? Huber (1991, p. 89) explains that "an entity learns if, through its processing of information, the range of its potential behavior is changed." He regards both explicit and implicit behavioral change as organizational learning. From Huber's (1991) conception, organizational learning is defined as the change of organizations' potential behavior through information processing. Furthermore, paying attention to unintended behavioral change of organizations, Levitt and March (1988) claim that "Learning does not always lead to intelligent behavior" (p. 335). Thus, learning may become suboptimal, and the behavior formed through learning may be undesirable for an entire organization (Christianson, Farkas, Sutcliffe & Weick, 2009;Örtenblad, 2002;Valentine, 2018).

Interpretation in organizational learning
In organizational learning, to implement new knowledge, organizations must interpret it (Daft & Weick, 1984). Even for an initial action, some interim interpretation is needed, and then, the result of the action is fed back to refine the interpretation. Daft and Weick (1984) regard an organization as an interpretation system. They propose a model, in which organizations scan for knowledge, interpret it and learn. In this model, the interpretation is the core process, as it determines the scope and mode of the entire learning process.
Interpretation is emphasized by researchers (Akinci & Sadler-Smith, 2019;Crossan, Lane, & White, 1999;Huber, 1991;Örtenblad, 2002). As storing knowledge in organizational routine comes at a cost, knowledge is stored selectively (Levitt & March, 1988). Before such selective storage can occur, new knowledge must go through the interpretation process to be given meaning. We recognize the importance of interpretation because focusing on it matches our view that knowledge is legitimated by the interpretation formed through politics, which we discuss next.

Power politics of organizational learning
During the establishment of consensus on interpretation, conflict arises among organizational members. As a result, only knowledge that has acquired legitimacy is accepted by members and can survive (Burgoyne & Jackson, 1997). Thus, organizational learning involves dynamics that include various actors. Hence, more organizational learning researchers tend to view organizations as collective entities (Huzzard & Östergren, 2002). Specifically, Burgoyne and Jackson (1997) and Lawrence et al. (2005) view an organization as an arena consisting of self-interested actors. Dekker and Hans en (2004) assert that "organizational learning is a dynamic and interactive process, which takes place at multiple levels" (p. 220). In this vein, this study is based on the conception of arena (Burgoyne & Jackson, 1997;Lawrence et al., 2005), as it provides us with a lens to analyze the intraorganizational dynamics of organizational learning.
To analyze the organizational learning dynamics, Lawrence et al. (2005) introduce the notion of power, that is "the ability to get others to do what you want them to, if necessary, against their will" (Hardy, 1996, p. S7). Lawrence et al. (2005) classify types of power in organizational learning in two dimensions: whether the power is targeted to subject (actors) or object (e.g. organizational practices, structure and rule) and whether the power is exerted episodically (based on actors' ad-hoc discretion) or systemically (based on on-going organizational practices). Thus, Lawrence et al. (2005) identify four types of power: influence (toward subject episodically), force (toward object episodically), domination (toward object systemically) and discipline (toward subject systemically). Actors influence others by manipulating their cognition and behavior on an ad hoc basis. By force, actors construct circumstances that restrict options in others' behavior. For example, this is achieved by agenda setting and excluding opponents when necessary. In contrast, by domination, actors systemically restrict the options by shaping organizational routine and structure. Discipline refers to changing others' cognition and behavior through socialization and training.
In relation to power, researchers also note the importance of addressing politics -"the dynamics of power in organizations" (Lawrence et al., 2005, p. 180). Coopey and Burgoyne (2000) stress that organizational learning is influenced by the political process among actors. Moreover, organizational learning is "a political process in organizations [. . .] which allows certain groups to fulfil their objectives by dominating others" (Karatas -Özkan & Murphy, 2010, p. 460). Easterby-Smith, Crossan, and Nicolini (2000), noting that power and politics are fundamental dimensions of learning, assert that "learning processes do not happen in a vacuum and, on the contrary, take place in a landscape of interests and differential power positions and relations" (p. 793).
With such notion of politics, organizational learning is not as simple as "selecting the most beneficial piece of knowledge." For example, where one faction within an organization may wish to implement some new knowledge, another may oppose (Akgün, Byrne, Lynn & Keskin, 2007). Knowledge holders support and advance knowledge that meets their purpose; they use their knowledge to display their ability for influence (Willem & Scarbrough, 2006). Managers control the organizations' learning process (Örtenblad, 2002), and in extreme cases, they may interfere with the learning of groups who oppose management's interest (Field, 2011). Thus, power relations influence the learning process through the control of access to learning opportunities (Berends & Lammers, 2010), and the knowledge supported by more powerful groups is adopted and learned by an entire organization. As a result, the learning process is often irrational (Vince, 2001). Hence, an examination of organizational learning requires addressing the concept of power and politics (Lawrence et al., 2005;Vince, 2001), incorporating factors such as the impact of actors' interests (Field, 2011) and pressure from external audiences (Dekker & Hans en, 2004). Thus, since the 2010s, organizational learning theory has considered politics (Ganz, 2018), emphasizing the social perspective rather than the technical process (Vince, 2001).

Organizational learning as knowledge legitimation
Hence, one of the important topics of organizational learning research is exploring how knowledge is interpreted through such political processes. To understand such process, Burgoyne and Jackson (1997, p. 62) provide the important suggestion that organizational Politics-based knowledge legitimation model learning is "the locus in which non-standard thoughts are given some legitimacy." Namely, organizations adopt knowledge that achieved legitimacy (MacLean & Behnam, 2010;Tregaskis, 2003); otherwise, knowledge is disregarded (Drori & Honig, 2013). According to Meyer and Rowan (1977), for example, formal bureaucratic organization was diffused not for its rationality but because it was regarded as legitimate. Hence, it is beneficial to address legitimacy in examining the dynamics of organizational learning (Chandler & Hwang, 2015).
To examine the legitimacy and knowledge legitimation, we need to first know more about the concept of legitimacy, which we expect to be key for understanding the politics of organizational learning. To this end, we focus on institutional theory for the following reasons.
First, legitimacy is one of the central topics of institutional theory (Tost, 2011). Institutional theory analyzes why and how organizations adopt new practices in the pursuit of legitimacy (Greenwood, Oliver, Lawrence & Meyer, 2017). Thus, institutional theory and organizational learning theory both aim to clarify the mechanism of change (or lack thereof) to an organization's practice. Some empirical studies integrate both perspectives, such as Cheng's (2010) analysis of the implementation of new practices by Taiwanese manufacturers and Sanders and Tuschke's (2007) study on the adoption of stock options by German firms. Hence, institutional theory will provide useful insight into the discussion of the politics of organizational learning with its findings on the influence of legitimacy.
Second, institutional theory provides the definition and classification of legitimacy (Suchman, 1995). These conceptions enable us to refine our discussion on the relationship between power and legitimacy by, for example, providing a framework to discuss "which type of power leads to which type of legitimacy." Third, institutional theory sheds light on both the reactive and proactive aspects of organizational learning (Daft & Weick, 1984). Upon examining such facets of organizational learning, the perspective from institutional theory that organizations react under environmental pressure (DiMaggio & Powell, 1983) is useful. Hence, researchers find it beneficial to integrate and apply both theories when analyzing organizational learning (Burgoyne & Jackson, 1997;Chandler & Hwang, 2015). For these reasons, we explore the concept of legitimacy in institutional theory in the next section.

Legitimacy in institutional theory
What is legitimacy? Among the most cited definitions of legitimacy is Suchman's (1995, p. 574) proposition that legitimacy is "generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions." Legitimacy is granted to not only organizations and groups but also various objects including acts, rules, procedures, routines and practices (Johnson, 2004). Modifying Suchman's (1995) definition to suit the focus of this study, we define the legitimacy of knowledge as "generalized perception or assumption that given knowledge is desirable, proper, or appropriate within some socially constructed system of norms, values, beliefs, and definitions." Suchman (1995) classifies legitimacy into three categories, which are widely accepted by researchers (Brown & Toyoki, 2013). Pragmatic legitimacy is based on an audience's evaluation of their own interests. Pragmatic legitimacy is granted when audiences recognize that the attributes of an object fit their own interests. In this case, the criterion for evaluation is an audience's interest. Moral legitimacy is granted based on an evaluation of the morality of an object's attributes. Here, the criterion is social desirability. For example, a food delivery service for older adults in the community is likely to achieve moral legitimacy, as audiences see that it contributes to improving the lives of these people. Cognitive legitimacy is granted when audiences take an object's attributes for granted without deep scrutiny. The criterion is taken-for-grantedness. For example, a new business model, such as an internet auction platform, achieves cognitive legitimacy once societies recognize it as a common form of business.
External legitimacy and influence of audiences Institutional theory emphasizes external audiences, as legitimacy depends on their evaluation (Hampel & Tracey, 2017). An audience judges legitimacy based on whether an object meets their expectations of interests, desirability and taken-for-grantedness.
Organizational learning is not independent of the influence of audiences. Organizations alter their behavior to satisfy audiences' expectations (DiMaggio & Powell, 1983). Thus, organizations may, in pursuit of legitimacy, constrain their own learning activities to meet the expectations of audiences (Burgoyne & Jackson, 1997). For example, Dekker and Hans en (2004) found that learning by government organizations was encouraged or even prevented under the influence of politicians and media.

Internal legitimacy
As discussed above, institutional theory focuses on legitimacy granted by external audiences (external legitimacy), but we must also consider internal legitimacy (Drori & Honig, 2013) to discuss organizational learning. That is, for new practices to be adopted by an organization, they must achieve internal legitimacy, which is granted by organizational members (MacLean & Behnam, 2010). Organizational learning is, thus, a process in which new knowledge achieves internal legitimacy (Burgoyne & Jackson, 1997).
This then begs the question of how knowledge achieves internal legitimacy. Undoubtedly, any practical benefit brought by the knowledge will lead to internal pragmatic legitimacy. However, organizations also incorporate new knowledge that does not result in practical benefit (Huber, 1991). For example, Guler, Guill en and Macpherson (2002) claim that the International Organization for Standardization (ISO) 9000s quality management system (QMS) was diffused not for technical benefit but as a response to institutional pressure. In this case, coercion by laws and regulations, as well as mimetic pressure, triggered learning; organizations implemented QMS to achieve moral and cognitive legitimacy.

Research questions
As discussed so far, previous studies have examined the mechanism of organizational learning, addressing interpretation, politics and legitimacy. Nevertheless, there are still some research gaps. First, the mechanism of knowledge legitimation remains unexplored. According to Burgoyne and Jackson (1997) and Tregaskis (2003), organizational learning is a process through which power politics bestow legitimacy on knowledge. However, examination of its underlying mechanism is insufficient (Guler et al., 2002). For example, how power contributes to internal legitimacy is not explored.
Second, analysis of the mechanism of power exercise in organizational learning is insufficient. Lawrence et al. (2005) elaborate on the types of power. However, for example, the search for factors that motivate actors to exercise power and alter power relations among them remains an issue.
To bridge these gaps, this study aims to identify the mechanism of politics through which knowledge is legitimated. Hence, we set our research questions (RQs) as follows: Politics-based knowledge legitimation model

RQ1. How do power politics bestow internal legitimacy on new knowledge?
RQ2. What shapes power politics of organizational learning?
Answering these questions will improve our knowledge on the dynamics of organizational learning as well as the mechanism of suboptimal learning. In particular, we seek to explain the underlying mechanism behind the proposition of Lawrence et al. (2005) that power politics drive organizational learning. Furthermore, we seek to elaborate on the assertion of Burgoyne and Jackson (1997) and Tregaskis (2003) that organizational learning is a process of knowledge legitimation. In our analysis, we focus on interpretation among the three subprocesses in Daft and Weick's (1984) model. This is because interpretation is the phase, wherein the knowledge is internally legitimated and meaning is given to the knowledge (Daft & Weick, 1984). Moreover, according to Vince's (2001) notion that "power relations directly mediate interpretation process" (p. 1329), the interpretation phase is the primary locus of power exercise.

Method
As an initial step to explore the mechanism of power exercise in organizational learning, we synthesized the claims of prior studies. To this end, we referred to the method of "narrative review" (Post, Sarala, Gatrell & Prescott, 2020) rather than conducting a quantitative systematic review. Although a systematic review has benefits, it requires clear criteria to identify the scope of the literature for investigation. Unfortunately, it is difficult to apply this method to organizational learning theory, though not impossible. This is because there are various domains and studies that do not advocate the term "organizational learning" while targeting equivalent phenomena such as "knowledge management" studies. Therefore, while recognizing the limitations, we conducted "qualitative discussions" (Post et al., 2020, p. 371) that are not bound by objective criteria.
The literature extraction referred to Webster and Watson's (2002) "structured approach" as follows. First, we reviewed the literature in major journals of organizational theory (Academy of Management Journal, Academy of Management Review, Administrative Science Quarterly, and Strategic Management Journal) with "organizational learning," "knowledge management" and "knowledge transfer" as search terms. In addition, we extracted articles that cited Burgoyne and Jackson (1997) and Coopey and Burgoyne (2000), which are seminal papers that focus on politics of organizational learning at the early stage. This is what Webster and Watson (2002) call the "go forward" method. We further explored articles cited in these articles, which Webster and Watson (2002) call the "go backward" method. From the literature extracted through these methods, we explored factors that shape the politics of organizational learning. In addition, we supplemented the discussion with other seminal works in institutional theory and organizational learning theory. However, our method was not exhaustive, and thus, the model presented here is of tentative nature.

Politics-based knowledge legitimation model
In this section, we present a politics-based knowledge legitimation model (Figure 1), which focuses on the interpretation of new knowledge (Daft & Weick, 1984). This model is summarized as follows. Actors exercise four types of power (Lawrence et al., 2005)influence, force, dominate and disciplinewhich are triggered, enhanced, facilitated and inhibited by various factors. Each type of power leads to bestowing different type of legitimacypragmatic, moral and cognitiveon knowledge. New knowledge recognized as legitimate is adopted and implemented by an organization. Furthermore, the learning results are fed back to the next learning cycle. The following subsections elaborate on this model.

Determinants of power exercise
Power politics of organizational learning is shaped by the following factors: estimated practical benefit; estimated cost; reliability, validity and urgency of knowledge; compatibility with existing templates; belief and interests of influential actors; and external legitimacy. In overview, these factors shape politics in the following manner. First, some factors trigger power exercise. For example, actors' belief triggers their power exercise. Second, some factors enhance power exercise. Actors will enhance exerting their power if they recognize that the learning contributes to their own interest. Third, some factors facilitate power exercise or even make it unnecessary. If some new knowledge is compatible with an organization's identity (Brown & Starkey, 2000), then the knowledge is likely to be accepted by members, where less power is needed. Forth, in contrast, some factors inhibit power exercise. For example, the cost of learning inhibits actors' power exercise because of its negative impact on actors. The mechanism of how each factor triggers, enhances, facilitates or inhibits power exercise is explained below.
Estimated practical benefit. If some actors expect practical benefit for themselves from implementing new knowledge, then such benefit will trigger and enhance their power exercise for advancing the knowledge. Then, if the knowledge contributes to a practical benefit for an entire organization, then it will achieve internal pragmatic legitimacy, facilitating advocates' power exercise. Meanwhile, if the benefit only applies to one group but not others, then conflict of interest arises (Field, 2011), which will be discussed later.
Practical benefit takes the form of economic benefit in for-profit organizations (Gupta & Govindarajan, 2000). Furthermore, by learning, organizations obtain new knowledge that enables the acquisition or possession of scarce resources. Hence, they become more independent of other organizations (Simonin, 2004), which constitutes another practical benefit.
If an organization recognizes the benefit of the knowledge, then such recognition will be transferred to its original holders and enhance their power, leading to altering power  Politics-based knowledge legitimation model relations. For example, Hisoka Maejima, who imported the practices of postal service to Japan in the 19th century, gained significant power in the government (Westney, 1987). Thus, practical benefit of knowledge impacts politics of learning. Estimated cost. Organizations also estimate the cost of learning. Actors are not willing to advance the knowledge if the cost to learn is expected to be higher than the expected benefit, which in turn inhibits power exercise.
Moreover, the estimated carnivalization cost within an existing business (Akgün et al., 2007) influences politics. If implementing the new technology contributes to the profit of one business unit but it has a negative impact on another unit, then the affected unit will trigger and enhance their power exercise to hinder the implementation. As a result, the technology may not be adopted by a company. Thus, carnivalization cost typically triggers political process, which may result in suboptimal learning for an entire organization.
Reliability, validity and urgency of knowledge. The attributes of the knowledge also impact politics of organizational learning. The reliability and validity (March, Sproull & Tamuz, 1991) of knowledge facilitate learning, as its advocate's intentions are more likely to be accepted by other members. Specifically, these factors facilitate the advocate's power exercise.
The urgency of knowledge, though not an attribute of knowledge itself, prompts learning (Desai, 2010) by altering the priority among different ideas (Sine & David, 2003). For example, a fatal maritime disaster in the 18th century drastically sped up the diffusion of a new navigation method by enhancing its advocate's power (Cattani, Ferriani & Lanza, 2017). However, a too rapid implementation causes conflict (Newman, 2000), as the urgency of the knowledge and its compatibility with existing templates conflict with each other, as we discuss next.
Compatibility with existing templates. Knowledge that fits existing templates is likely to be preferred by an organization (Coopey & Burgoyne, 2000;Newman, 2000), where advocates' power exercise is facilitated. In particular, compatibility with existing institutions is important (Hargadon & Douglas, 2001;Heusinkveld & Reijers, 2009). Existing institutions already meet the audience's interest and norms and are taken for granted, having already been conferred with pragmatic, moral or cognitive legitimacy. Thus, new knowledge compatible with existing institutions is likely to gain internal legitimacy (Sanders & Tuschke, 2007). Such knowledge tends to satisfy audiences' expectations, minimizing the tension between new and existing knowledge, which mitigates conflict among actors. For example, Hargadon and Douglas (2001) assert that Thomas Edison was able to succeed in his electric light business by maintaining compatibility with existing gas light systems in addition to emphasizing newness. Hence, compatibility with existing institutions facilitates advocates' power exercise.
Existing templates regarded as the frames of reference for learning are embedded in various aspects of organizations. For example, dominant logic, by constructing leaders' belief, influences knowledge filtering (Bettis & Prahalad, 1995). Similarly, organizational members take actions to protect their organizational identity (Brown & Starkey, 2000). Thus, organizational identity constrains the way of organizational learning (Christianson et al., 2009).
These templates are formed through organizations' histories (Levitt & March, 1988), reflecting their successes and failures (Sanders & Tuschke, 2007), which shapes path dependence (Sydow, Schreyögg & Koch, 2009). Meanwhile, if the templates become obsolete, then subsequent learning may become suboptimal. Knowledge legitimation with reference to existing templates can occur even after the templates have become obsolete or even a ritual, with its original practical benefit lost (Dougherty & Heller, 1994). Such persistence with obsolete templates may lead to the failure of unlearning (Hedberg, 1981).
Belief and interests of influential actors. The determinants of politics in interpretation are not limited to properties of knowledge or an organization. Such factors exist also on the actors' side. Importantly, actors aim to advance knowledge that meets their interests as Willem and Scarbrough (2006) observed in the case of a finance service company and an energy company. To this end, they attempt to manipulate others' cognition and preference to conform to their own belief and interests (Hardy, 1996). Thus, organizational learning reflects the belief and interests of influential actors (Field, 2011).
Organizational leadership is one of the most influential factors in orienting learning direction (Christianson et al., 2009;Valentine, 2018). Hence, a leader's belief (basic assumptions of what values and knowledge are to be emphasized) influences learning strategy (Nag & Gioia, 2012). It also shapes members' cognition (Hardy, 1996) in determining the direction of interpretation. The interest of knowledge holders, boundary spanners (Willem & Scarbrough, 2006) and actors occupying network hubs in a social network (Schilling & Fang, 2014) may affect knowledge transfer and filtering, even if they are not at the top of the hierarchy. If these actors recognize that some knowledge meets their interests or that they can use it to influence others or display their capability or knowledgeability, then they appropriate their positions so that the knowledge is learned by the organization (Berends & Lammers, 2010). Contrarily, actors occupying network hubs may pretend to forget, lie or hinder information transfer for their interests (Schilling & Fang, 2014).
Thus, actors' belief and interests trigger and enhance their power exercise. They even form interest groups, which may create conflict with other groups and occasionally result in suboptimal learning for an entire organization (Field, 2011).
External legitimacy. Organizations adapt to external pressure to gain legitimacy (Meyer & Rowan, 1977). Thus, organizational learning is oriented in such a way that external legitimacy is achieved (Burgoyne & Jackson, 1997;Dekker & Hans en, 2004), influencing associated politics.
Notably, actors in an organization are not only constrained by but also proactively use external pressure to promote organizational learning. In Fields' (2011) case study, for example, managers appropriated consultants' legitimacy, where the consultants catalyzed organizational learning. Mulgan's (1997) case illustrates that pressure from external legitimacy is also used by groups in a government. In the 1990s, the Government of Japan (GOJ) accepted agricultural trade liberalization policy under the pressure from US Government. However, GOJ did not merely bowed to the pressure. A school within GOJ who supported the policy used the pressure to silence opponents. That is, the school mobilized the external legitimacy of US Government to control the political process. These examples (Field, 2011;Mulgan, 1997) imply that external legitimacy facilitates advocates' power exercise.
Each type of external legitimacypragmatic, moral and cognitive (Suchman, 1995) facilitates learning in the following manners. Knowledge with moral legitimacy, achieved by adhering to professional norms, is likely to be supported and adopted by those who share the norms. For example, ISO 9000's QMS diffuses faster in countries where it is believed that an emphasis on quality aligns with engineers' professional norms (Guler et al., 2002). Laws and regulations orient organizational learning (Dougherty & Heller, 1994;Madsen & Desai, 2018) so as to meet legal requirements in pursuit of moral legitimacy. Expectations from customers also act as pressure for learning. For example, Taiwanese OEM manufacturers implemented CRM and SCM to respond to pressure from their customers (Cheng, 2010). In this case, the manufacturers gained pragmatic legitimacy by aligning with their customer's Politics-based knowledge legitimation model interests. If a given practice is widely recognized as successful, then it is taken for granted and gains cognitive legitimacy (Chandler & Hwang, 2015). In the case of the Taiwanese manufacturers (Cheng, 2010), the successful adoption of CRM by competitors urged the focal firm to adopt CRM. However, as Drori and Honig (2013) stress, external legitimacy may conflict with internal legitimacy. External legitimacy does not contribute to internal legitimacy if other factors prevail. For example, the Israeli National Research Institute, examined by Sapir (2020), originally valued the logic of science in basic research, which served as its primary criterion for internal legitimacy. In 1959, the institute established a subsidiary company for managing for-profit matters and pursued external legitimacy by meeting the logic of the state of newly formed Israel (contributing to national industry and economics) and the logic of the market (gaining profit). However, these criteria for external legitimacy heavily conflicted with the existing criterion for internal legitimacy (logic of science). As a result, with a lack of internal legitimacy, the subsidiary company lost stakeholders' support, and its status was drastically phased out. This case illustrates that knowledge must gain internal legitimacy to be learned, even if it has already achieved external legitimacy.
In addition, as Meyer and Rowan (1977) claim, excessive conformity to external pressure may cause unanticipated consequences such as degraded efficiency. That is, excessive emphasis on external legitimacy may lead to suboptimal learning. Thus, to ensure efficiency while apparently conforming to external pressure, organizations "decouple" their inherent practices from formal practices visible to audiences (Meyer & Rowan, 1977).

Effects of power on internal knowledge legitimation
As has been discussed, internal legitimacy is the key for new knowledge to be learned (MacLean & Behnam, 2010;Tregaskis, 2003). For internally legitimating new knowledge, actors persuade others to identify, discuss and constructively question the legitimacy of the knowledge (Burgoyne & Jackson, 1997). In this process, actors exert their power to alter the cognition and behavior of others (Hardy, 1996). For example, through discipline (Lawrence et al., 2005), actors intervene in others' interpretation on knowledge and in their belief and behavior. Thus, power shapes and manipulates the legitimacy of knowledge. Certainly, such legitimation process is not straightforward, as an actor's power triggers another's, leading to complicated interactions. Therefore, in this subsection, as an initial step to explore the effect of power on knowledge legitimation, we examine the general tendency of the way how knowledge advocates' power leads to internal legitimacy of the knowledge, by considering the tactics used in each type of power.
According to Lawrence et al. (2005), each type of power is exerted through specific tactics. Thus, different types of power are likely to lead to specific types of legitimacy. In general, on one hand, subject-oriented powerinfluence and disciplineleads to all typespragmatic, moral and cognitiveof legitimacy because its associated tactics aim to shape members' cognition on desirability, taken-for-grantees and interests. On the other hand, subject-oriented powerforce and dominationcontributes only to pragmatic legitimacy, as these powers aim to alter the benefit-cost structure embedded in an organization's practices without directly shaping members' cognition. The paragraphs that follow elaborate on how each type of power leads to a specific type (Suchman, 1995) of legitimacy by meeting its evaluation criteria, which were explained in the subsection "What is legitimacy?" Among subject-oriented power, through influence, actors can bestow moral legitimacy on knowledge through "moral suasion" (Lawrence et al., 2005). Actors may also attempt to manipulate others' sense of "taken-for-grantedness," leading to cognitive legitimacy. If the actor succeeds in persuading that the knowledge contributes to others' interest, then the knowledge will gain pragmatic legitimacy. For such mechanism, Willem and Scarbrough (2006) emphasize the consummatory aspect of learning. That is, interactions between power exercisers and exercisees contribute to sharing values, norms, trust, mutual understandings, organizational commitment and goal congruence, which lead to moral, cognitive and pragmatic legitimacy.
Actors attempt to discipline others by socialization, compensation and training (Lawrence et al., 2005). As a result of these tactics, organization members regard new knowledge as desirable and take it for granted, leading to moral and cognitive legitimacy. Valentine (2018) illustrates the effect of discipline for cognitive legitimacy; in a project aimed at implementing a new practice in a hospital, the top management improved the cognitive legitimacy of the practice through collaboration with members. Thus, leaders' way of thinking is reflected in that of members (Nag & Gioia, 2012).
On the other hand, considering its inherent nature, object-oriented powerforce and dominationis not likely to contribute to moral or cognitive legitimacy, as they do not intend to directly alter others' cognition. Therefore, members will not voluntarily regard new knowledge as desirable or take it for granted. For example, in the case of an Israeli research institute (Sapir, 2020), domination through the establishment of a subsidiary company could not change stakeholders' sense of moral or cognitive legitimacy on the company. Instead, these types of power restrict options in members' behavior and increase the cost for opposition. For example, by forcing through agenda setting (Lawrence et al., 2005), leaders set formal tasks, thus restricting options. In an extreme case, opponents may be removed (Lawrence et al., 2005), leading to a considerably high cost for the opposition. Through domination, powerful actors increase the cost of opposition by manipulating the physical layout of the workplace or information systems to conduct members into a predetermined path (Lawrence et al., 2005). Thus, as a force or domination exercised, with an increased cost for the opposition, the relative benefit for accepting new knowledge increases, leading to pragmatic legitimacy, with moral or cognitive legitimacy relatively unchanged.
Power exercise for interfering with learning Thus, power advances organizational learning by promoting legitimation of new knowledge. Contrarily, actors also exert their power to interfere with learning, hindering legitimation of knowledge so that the organization would not learn it. For example, actors may interfere with learning when it does not fit their interests (Akgün et al., 2007;Örtenblad, 2002). Knowledge holders may be reluctant to disclose their knowledge (H eliot & Riley, 2010) if they do not recognize incentives in disclosure (Willem & Scarbrough, 2006). If actors at network hubs (Schilling & Fang, 2014) exert their force to disturb the knowledge flow, then other members will not have an opportunity to evaluate whether the knowledge fits their interests, desirability or taken-for-grantedness. In these cases, with restricted access (Berends & Lammers, 2010), the knowledge cannot achieve internal legitimacy.
Thus, organizational internal politics may inhibit learning, and associated learning processes may become irrational (Vince, 2001). For example, in NASA, deficiency in learning caused by internal politics resulted in the accident of the space shuttle Challenger (Seo, 2003). In this case, for the fear that knowledge generated by one group may threaten the political background of another group, the knowledge was ignored, leading to the accident.
Interplay between power and knowledge So far, we elaborated on the mechanism of when and how actors exert their power in knowledge legitimation. Moreover, knowledge can be the source of power as it creates a Politics-based knowledge legitimation model dependency among actors (Tregaskis, 2003). Hence, actors use knowledge for political purposes (Karatas -Özkan & Murphy, 2010). Holders of beneficial knowledge, as well as those who can control the flow of such knowledge, can enhance their power in organizational politics (Willem & Scarbrough, 2006). This is witnessed in the case of a person who gained power as a result of importing postal services to Japan (Westney, 1987). Thus, power and knowledge are mutually related; actors' power bestows internal legitimacy on knowledge, and knowledge enhances the power of its holder and other related actors.

Adoption of internally legitimated knowledge
Thus, specific knowledge achieves internal legitimacy through political process. Knowledge that has gained either of the three types of legitimacy (Suchman, 1995) is likely to be advanced in the learning process as follows. First, knowledge with internal pragmatic legitimacy, recognized as it meets the organization's interest, will be supported by more members, and they will mobilize more power for implementing it. Second, knowledge with internal moral legitimacy is likely to be supported by those who share the common norms. Hence, organization members will mobilize more power so that an entire organization adopts the knowledge. For example, ISO 9000's QMS was supported by engineers who share the norms of QMS, and thus, its implementation was promoted (Guler et al., 2002). Third, as in its definition (Suchman, 1995), organizational members will take knowledge with internal cognitive legitimacy for granted and, thus, be likely to exert their power to promote learning process without scrutiny. Furthermore, all types of internal legitimacy will reduce the required power for promoting learning as resistance by opponents will be weakened. Thus, internally legitimated knowledge is adopted and implemented by an organization (Burgoyne & Jackson, 1997;MacLean & Behnam, 2010).

Feedback loop of organizational learning
As discussed so far, new knowledge is interpreted and learned through politics. Then, the learning result is fed back to the politics and associated interpretation in the next learning cycle as follows.
First, learning results will alter the determinants of politics. For example, actors' belief and interest, knowledge reliability and benefit-to-cost structure may change as a result of the learning. In addition, if an actor obtains stronger power, then this will alter the power balance within an organization. For example, changes in the interests of knowledge holders (Willem & Scarbrough, 2006) or the position of the network hub (Schilling & Fang, 2014) will influence the balance and actors' power exercise.
Second, the learned knowledge is sent back to "re-interpretation" and is interpreted in a different way reflecting the result of the action taken. Thus, the interpretation is revised, and the interim one will converge into a permanent one.
Third, in a long run, the politics that reflect actors' belief and interests, as well as their power relations, change over time as a result of recurrent organizational learning cycles. Consequently, path dependence emerges as learned practices become efficient and further preferred by members (Sydow et al., 2009). Finally, the path dependence leads to selfreinforcement, constraining the scope and direction of future organizational learning. Occasionally, suboptimal learning for an entire organization, or even failure of unlearning (Hedberg, 1981), may occur if learning resulted from path dependence becomes obsolete.

Discussion and conclusions
Features of the politics-based knowledge legitimation model By integrating the assertions of organizational learning literature, a politics-based knowledge legitimation model (Figure 1) was developed, which has the following features. First, internal legitimacy is the key element in the interpretation of knowledge (Burgoyne & Jackson, 1997;MacLean & Behnam, 2010;Sapir, 2020). Internal legitimacy is the frame of reference for interpretation. No matter how beneficial new knowledge is, it cannot take hold in an organization until it achieves internal legitimacy as Sapir (2020) illustrates.
Second, internal legitimacy is bestowed on knowledge as a result of power politics (Burgoyne & Jackson, 1997). Knowledge that fits an actor's interest is supported by them. Contrarily, knowledge that does not fit the actors' interests will not be supported even if it is beneficial for an entire organization. As a result, such knowledge is screened out without achieving internal legitimacy. At times, learning results may be suboptimal for an entire organization. This can happen if a powerful leader endorses knowledge that is not optimal for the entire organization (Christianson et al., 2009;Örtenblad, 2002;Valentine, 2018). Knowledge with external legitimacy is also subject to this political process. Although external legitimacy often contributes to internal legitimacy, external legitimacy is not the sole frame of reference for internal legitimacy as in Sapir's (2020) case of an Israeli research institute. Thus, knowledge with external legitimacy cannot be learned unless it gains internal legitimacy through political processes.
Third, specific types of power (Lawrence et al., 2005) lead to specific types of legitimacy (pragmatic, moral and cognitive; Suchman, 1995). Notably, while subject-oriented power (influence and discipline) leads to all three types of legitimacy, object-oriented power (force and dominance) leads only to pragmatic legitimacy. This is because the latter only addresses formal practices such as reward system, without directly shaping members' sense of desirability or taken-for-grantedness.
Forth, various factors shape the politics. Such factors include practical benefit; a relatively low estimated cost; the reliability, validity and urgency of the knowledge; compatibility with existing templates; the belief and interests of influential actors; and external legitimacy. These factors trigger, enhance, facilitate or inhibit actors' power exercise.
Last but not least, power and knowledge affect each other. While power relations influence the learning process (Berends & Lammers, 2010;Field, 2011;Örtenblad, 2002), knowledge can enhance the power of related actors (Tregaskis, 2003;Willem & Scarbrough, 2006). Thus, organizational learning and power politics are mutually related.

Theoretical contributions and practical implications
This paper aimed to develop a model of knowledge legitimation in organizational learning that incorporates power politics and its determinants (Figure 1). Our theoretical contributions are as follows. First, we identified which type of power (Lawrence et al., 2005) leads to which type of legitimacy (Suchman, 1995). While Burgoyne and Jackson (1997) and Tregaskis (2003) assert that organizational learning is a process of knowledge legitimation, its underlying mechanism has not been sufficiently examined; in this study, we complemented their discussions.
Second, we revealed the factors that shape actors' power exercise in politics of organizational learning. We also identified the mechanism through which these factors Politics-based knowledge legitimation model shape actors' power exercise: triggering, enhancing, facilitating and inhibiting. These findings refine the discussion by Lawrence et al. (2005) on the exercise of types of power by elaborating on its underlying mechanism of power exercise. Third, we identified the interplay between power and knowledge. In the literature, the influence of power on knowledge (Berends & Lammers, 2010;Dekker & Hans en, 2004) and that of knowledge on power (Tregaskis, 2003;Willem & Scarbrough, 2006) have been discussed separately. By integrating these assertions, we elaborated on the dynamic nature of organizational learning; power and knowledge influence each other.
As this study's practical implication, it is suggested that organizations' management should recognize the complexity of organizational learning. Organizational learning is not a straightforward process of importing and storing knowledge. It is facilitated or even inhibited by politics. To facilitate organizational learning, management should monitor and control actors' power exercise. In particular, if management seeks to shape members' cognition for moral or cognitive legitimacy of knowledge, then they should not exert force or domination but influence or discipline. Furthermore, as this study identified, various factors influence the politics. These factors should be considered when managing the learning process.

Limitations and future directions
As our model has a tentative nature, further examination is needed. Possible directions for future research are suggested as follows. First, the priority among the determinants of politics should be explored. Specifically, we need to examine which factor prevails when multiple factors conflict. The answer to this question will depend on the situation. Therefore, the second question should explore contingencies. For example, we should examine the influence of environmental uncertainty and ambiguity. Under such conditions, organizations tend to imitate other organizations (DiMaggio & Powell, 1983), which indicates that they are likely to prioritize external legitimacy. Similarly, internal contingencies also need to be examined. For example, active and passive organizations may tend to have different priorities (Daft & Weick, 1984). Public and private organizations (Perry & Rainey, 1988) will also have different priorities. Third, priority among factors depends on types of learning. For example, it is expected that in exploration and exploitation (March, 1991), individual actors and an entire organization will have different priorities. Such difference among learning types should be examined. Fourth, we should explore interactions among multiple actors' power. While we examined how power leads to the internal legitimacy of knowledge, such interactions may have different consequences. Again, our model is tentative. While we identified the determinants of politics and the subsequent knowledge legitimation process, such mechanism should be validated empirically.