When cooperation turns ugly: exploring the dark side of social capital

Purpose – This article explores theoretical assumptions regarding negative consequences of social capital in the empirical case of a failed cooperation project, and how these consequences are related to processes involving people, structures and environments. Design/methodology/approach – The article is based on a case study of a cooperation project within municipal labor market services. The methodology followed a theorizing process, where data were collected through ethnographical methods and analyzed in relation to existing concepts from theories describing negative effects of social capital and shadow organizing. Findings – The results highlight how the development of negative social capital in the project can be understood through three relational processes, namely the social dynamics of insulation, homogenization and escalating commitment. The authors conclude that the quality of social capital is conditional upon complex interactions within social structures. Moreover, the results highlight the importance of studying organizing practices outside explicit structures, in order to identify the development of non-canonical practices and their consequences. Practical implications – Organizing cooperation projects that aim to bridge professional competencies or organizational boundaries have to be attentive toward informal organizing practices which if remaining unrecognized may grow and threaten the original intentions. Originality/value –The studymakes a theoretical contribution by combining a shadow organizing approach with literature on social capital. This combination proves especially useful for analyzing how organizational dynamics can influence the development of social capital into producing negative effects.


Introduction
To achieve purposeful cooperation between professionals from different organizations, it is crucial to secure trusting relationships. The importance of trust and adequate structural prerequisites has been emphasized in multiple theoretical and empirical studies of interorganizational cooperation. (de Rijk et al., 2007;St ahl, 2010;St ahl et al., 2017) Theories on cooperation often point out the complexities in attaining such conditions, and some authors have even recommended against engaging in cooperative initiatives unless deemed unavoidable (Huxham, 2003).
In the research on trust and interorganizational relations, the extent of social capital within the relationships has been identified as a key aspect in explaining how well cooperation goals are achieved (Nahapiet and Ghoshal, 1998;Tsai and Ghoshal, 1998). The literature on social capital is vast, and it is commonly described as a positive resource in various interpersonal and community-based relationships (Szreter and Woolcock, 2004;Woolcock, 2010). However, a "dark side" of social capital has also been identified, i.e. how close relationships may also function to be excluding and destructive toward those not included in the group (Gargiulo and Benassi, 1999;Pillai et al., 2017). Such negative effects have received less focus in empirical research.
The purpose of this article is twofold: (1) to contribute to the understanding of negative consequences of social capital in interorganizational cooperation between public organizations by exploring theoretical assumptions in the empirical case of a failed cooperation project, and (2) how these consequences are related to processes involving people, structures and environments.
Exploring the dark side of social capital Social capital has been discussed for decades with some variations in definitions and applications. The historical roots of the concept can be traced back to fundamental sociological notions of social cohesion, socialization and stratification. One of the first modern theorists that used social capital as a term was Bourdieu, who described it as a resource based on a person's membership in social networks and groups and the gains this could bring in terms of recognition and social benefits (Bourdieu, 1986). Since then, social capital has become an increasingly popular concept and applied across a variety of disciplines and research fields (Rostila, 2010;Woolcock, 2010;Kwon and Adler, 2014). Most commonly, these modern adaptations use it as a positive characteristic of social groups, e.g. for studying the amount of support or cohesion, in many studies related to various forms of outcomes, such as health (Szreter and Woolcock, 2004;Oksanen et al., 2008). Empirical studies using the term in this way generally conclude that social capital is positively related to health outcomes, which is at least partly a consequence of how it is conceptualized and operationalized, leaving little room for exploring potentially negative effects (Szreter and Woolcock, 2004). For instance, in a widely used measure of social capital in workplace settings, the concept is operationalized in exclusively positive terms (Kouvonen et al., 2006). Portes (1998) argues that we tend to have a bias toward seeing good things coming out of sociability, but that we need to emphasize the negative sides in order not to promote networks, social control and collective sanctions as unequivocal blessings and in order to provide serious sociological analyses rather than moralizing statements. In this article, we theorize social capital as primarily related to groups and as a neutral concept which may have positive or negative effects. In this vein, we argue that for scholars taking a sociological perspective, the notion of a "dark side" to social capital should be quite uncontroversial, as this perspective proposes that the positive or negative consequences of social bonds vary according to the situation (Portes, 1998). Acknowledging potential negative effects of social capital are, hence, a way of returning the concept to its roots, rather than considering it as a genuinely positive resource and as a remedy for various problems.
Social capital may be argued to have both individual and collective dimensions (Harpham et al., 2002;Rostila, 2010;Kwon and Adler, 2014). Social capital fundamentally consists of formal or informal social networks (that may be open or closed), social trust (that may be strong or weak depending on the strength of the networks) and social resources, which are derived from the previous two aspects (Rostila, 2010). Networks represent the structural QROM dimension of social capital, while trust represents the cognitive dimension (Nahapiet and Ghoshal, 1998). Social resources, which are products of networks and trust and hence the actual "capital", may be collective or individual and be expected to lead to different returns, either instrumental (such as economic, political or social gains), or expressive (such as physical or mental health benefits, based on the emotional support social capital may bring). Other scholars have suggested different typologies of social capital, where an influential distinction was made by Gitell and Vidal (1998) between bonding social capital (horizontal relations of trust and reciprocity between individuals with similar social identities at the same hierarchical level) and bridging social capital (horizontal relations between persons at the same hierarchical level from different backgrounds). A third variety, linking social capital (vertical relationships between people interacting across gradients in society) was introduced later (Szreter and Woolcock, 2004). The potential negative effects of social capital seem to be more commonly found in closed or isolated networks, which may exclude outsiders, lead to unequal distribution of social resources (Rostila, 2010) and limit connections to outside influences which may result in intolerance and discrimination (van Deth and Zmerli, 2009).
The "dark side" of social capital has been explored in empirical studies across a variety of settings in which researchers examine, for example, adolescent drinking (Martins et al., 2017), sports (Whittaker and Holland-Smith, 2016;Brown, 2018), finance (Carrera et al., 2017), or mental health problems and alienation in workplaces (Zhang et al., 2016;Sakuraya et al., 2017). Portes identifies at least four negative consequences of social capital: "exclusion of outsiders, excess claims on group members, restrictions on individual freedoms, and downward leveling norms" (Portes, 1998, p. 15) and gives examples of social contexts with high levels of social capital which have obvious negative consequences, such as Mafia families, prostitution rings and youth gangs, which are strongly embedded in social structures. Social capital is, hence, indifferent to the underlying values of a group (van Deth and Zmerli, 2009) and functions as a "sociological superglue" (Putnam, 2000, p. 23).
For the topic of the current article, we found a review of empirical evidence of negative effects of social capital in organizations and management (Pillai et al., 2017) to be particularly instructive. The review focuses primarily on the context of firms, and how social capital relates to several organizational outcomes, such as innovation, organizational cultures, organizational inertia, performance, decision effectiveness, and the acquisition, creation and transfer of knowledge. Pillai et al. identify six types of negative effects of social capital, which are all based on dysfunctional social identification processes: (1) dilution of the dialectical process; (2) inhibition of individual learning; (3) groupthink; (4) postponement of structural adjustments; (5) non-rational escalation of commitment; and (6) blurring of firms' boundaries (10, p. 98).
In their review and the following theoretical development of the concept, Pillai et al. argue that social identification processes are divided into structural, relational and cognitive aspects, where a strong social capital within a group will effectively block external impulses through shared sense-making and confirmation bias (Pillai et al., 2017). Such groupthink processes will isolate a group to become a closed information-processing unit that fuels uniformity of beliefs and exaggerates the group's internal bonds, which may cause over-commitment and justification of previous decisions, as well as polarization in relation to other groups. This is also a reason, Pillai et al. argue, that managers tend to not shut down projects which they have invested in. Introducing new decision-makers at points where the future of a project is to be decided is hence an effective method to reduce commitment to failing projects (Pillai et al., 2017). Negative effects of social capital may appear only late in a social process, where it is possible to have a positive development which morphs into more intense and closed networks, thus shifting from the "bright" to the "dark" side. It is difficult to break strong social ties which may lead to a relational inertia where the same contacts are repeatedly used by habit, serving as a filter that isolates practices from new possibilities (Gargiulo and Benassi, 1999). The possible gains of social networks hence come with a risk of negative effects. This focus of social capital as a changing quality of social contexts also raises the question of the mechanisms involved where a situation characterized by mutual trust and consensus turns into its opposite. In such negative circumstances, particular mechanisms determine how conflicts occur and how these interact with social identification processes. While the review of Pillai et al. mainly focused on private business contexts, the types of social identifications and group processes described could be expected to translate well to other types of organizations, although it is likely that any kind of organizational context will have its own influence over how such processes occur and develop. Our study focuses particularly on how negative social capital develops in cooperative projects between public organizations.
To add to the theories of negative effects of social capital, we have also identified a complementary theoretical metaphor, namely that of shadow organizing. This metaphor enables exploring organizing as intra-relating processes and to focus on what happens in the interstices between intentional and unintentional organizing actions (Gherardi et al., 2017). The "shadow" is a symbol of what is obscured, and which can only be seen by looking past the obvious and focusing on what is hidden. In this article, we have used this metaphor as an organizing tool to guide the discussion of the results, which are related to three concepts from this theory: productivity, liminality and secrecy. Productivity is illustrated "through the image of the forest and its sheltered spaces in penumbra"; liminality is seen "through the image of a liminal, grey zone between canonical and non-canonical practices"; secrecy captures "the organizational secrecy kept in-between the sayable and the unsayable" (Gherardi et al., 2017, p. 7).

Methods
The methodological process for this article was inspired by Hammond's (2018) and Swedberg's (2012) perspectives on theorizing. Theorizing is described as a process driven by identification of interesting problems and a motivation to solve them (Hammond, 2018). The analysis builds on a rich ethnographic material, which provides an authentic contextualization of the theoretical assumptions.
Empirical context: a Swedish cooperation project The empirical context for our analysis is a cooperation project within municipal labor market services in a medium-sized Swedish city. Sweden is internationally renowned for its consensus-driven policies, with processes for resolving conflicts in the labor market as a prime example. In Swedish studies of cooperation practices between authority actors, consensus has been identified as a starting point for actors when engaging in joint projects, although such preconceptions may be put to the test by interorganizational dynamics or systems not designed for joint action (St ahl et al., 2011). Sweden is an interesting empirical context for exploring the "dark side" of social capital, since there may be an expectation for actors to share a basic level of social capital in the drive for consensus, and that the people involved in projects organized by authorities could be expected to adhere to project guidelines.
The project studied here was organized through a coordination association, which is an interorganizational structure that is co-owned by municipalities, county councils, the Swedish Social Insurance Agency and Public Employment Services (St ahl, 2010). The project aimed to support teenagers and young adults to participate in the labor market, and the original intention was for the project to function as an arena where the involved authorities could direct them to appropriate support. The project was staffed with personnel from each of the authorities who were meant to work part time in their regular organization and part time in the project to facilitate interorganizational contacts. It was started in 2015 and closed before its allotted time in 2017. The project was followed using an interactive research QROM approach, which included a learning evaluation (Svensson et al., 2009) to inform the development of the project.
In a previous analysis (Strindlund et al., 2020) we outlined four organizational conditions that influenced the development of the project, namely communication, structure, trust and steering. Our previous results identified how flaws in these conditions shaped two different perceptions of the project, where a coordination rationality (i.e. that the project should aim for increased interorganizational contacts and a steady flow of participants through the project) clashed with an empowerment rationality (i.e. that the project should focus on working closely with fewer participants). In this article, we focus on how the organizing in the project, viewed from a processual and relational perspective involving people, structures and environments, can be theorized (Czarniawska, 2008). For this analysis, we make use of theories from the literature on social capital and place this within a shadow organizing framework (Gherardi et al., 2017) in order to illuminate that which takes place underneath the explicit project organization.

Data collection
The first basic step of theorizing suggested by Swedberg is to observe and choose something interesting. In order to theorize a phenomenon, one needs to observe the phenomenon and see it for oneself, preferable through field studies (Swedberg, 2012). The data consisted of ethnographical material collected by the first author from spring 2017 to November 2018 through observations, focus groups, interviews and document studies in line with recommendations for ethnographic studies (Hammersley, 2018). In total, more than 70 observations, nine focus group interviews and 34 interviews were carried out by the first author. Data from observations and informal interviews were documented in field notes and the interviews were recorded and transcribed verbatim. Table 1 presents a compilation of the data collection, comprising methods, interviewees, focus and structure, documentation strategy, timing and numbers. In summary, the data collection met the recommendations of theorizing, namely, to observe and stay in the context of discovery and develop tacit knowledge of the phenomenon by collecting strong and nuanced descriptions.

Data analysis
The next step of theorizing (Swedberg, 2012) is to name and formulate the central concept. The three authors gathered in several analysis sessions all contributing with different perspectives and pre-knowledge of the project and the theories. The first author had firsthand experience of collecting the data, the second author contributed with the shadow organizing perspective and the third author with social capital theory.
The analysis started from the six types of negative effects of social capital identified by Pillai et al. (2017): dilution of the dialectical process, inhibition of individual learning, groupthink, postponement of structural adjustments, non-rational escalation of commitment, and blurring of firms' boundaries. These served as an entrance-point for analyzing the material. Thereafter, the data, consisting of short notes based on the total ethnographic material such as quotations from interviews and reflections formulated in field notes, were mapped onto the effects identified above by the first author using an Excel spreadsheet. The data were analyzed by all three authors together searching for patterns and common denominators. As a further analytical step, the material was analyzed and rearranged in relation to and cross-tabulated with the concepts of bonding, bridging and linking social capital.
The general goal of theorizing is to come up with new ideas. This aspect, abduction, is defined as "observing a fact and then professing to say what idea it was that gave rise to that fact" (Swedberg, 2012). Abduction has both a logical and an innovative character and the quality lies in its "means-of-inferencing". It is a sensible and scientific form of interference intended to help researchers to be able to make new discoveries in a logically and methodologically ordered way (Reichertz, 2007). It is a reflexive and iterative process, where the theoretical concepts and the empirical study stimulate the next step in the theorizing process, namely, to build out the theory. This step aims to give form to the central concepts by outlining the structure of the phenomenon (Swedberg, 2012). In our study, this was done by analyzing how the negative effects of social capital unfolded, which we describe through three distinct social dynamics, namely: insulation, homogenization and escalating commitment. These dynamics comprise our main analytical contribution, which were identified inductively from the empirical material, where the labels were inspired by the descriptions of negative effects of social capital identified by Pillai et al. (2017) Overview of the empirical data QROM was hence done through an iterative process where theoretical concepts were adapted to suit the specific empirical context. The negative effects identified in the social capital literature where interleaved with these social dynamics; specific effects do not map into specific phases but come into all three dynamics in different ways, where these effects are described in the results as they appear in the social dynamics described. The metaphor of shadow organizing, finally, added a terminology for the organizational context and for describing the milieu in which negative effects of social capital could develop (Buch, 2020;Gherardi et al., 2017). An overview of the theoretical concepts used in the analytical process is presented in Figure 1.
The purpose of the analysis was to explore how the theoretical concepts could make sense of the data from the empirical case and what the case could contribute to the existing theories. This is the last step in the theorizing process: complete the tentative theory, including the explanation (Swedberg, 2012) which is summarized in the conclusions, where we outline the development of negative effects as a highly dynamic and non-linear process. For this step, the theoretical metaphor shadow organizing was useful for contextualizing the analysis.

Results
To understand the results and how the social dynamics developed, we will start with presenting the founder's and the financier's view of the project, as conveyed in project documentation and in interviews. In their view, illustrated in Figure 2, the project was intended to enhance collaboration and contribute to optimal use of the welfare system by creating an arena for coordinating services. The cooperative arena, i.e. the project, is here When cooperation turns ugly illustrated as a circle in which representatives from each of the participating actors each represent their respective organization. The representatives would divide their work hours between the project and the home organizations (i.e. the organization that the person originally worked in), thereby having a central, linking function, acting as knowledge brokers. The idea was that this project structure would facilitate project staff to gain access to each actor's resources and internal activities and also contribute to increase the knowledge in the home organization concerning the other actors' legal frameworks and cultures. The project staff was to represent their home organizations and contribute with their specific professional competence. However, this original idea was not clearly communicated to the project team or to cooperative actors (i.e. the organizations that were to cooperate with the project). Instead, the project group developed an independent organization with the aim to empower individuals. This development became a hotbed for negative effects of social capital, where our analysis presents how this unfolded through three different intra-relational processes, namely the social dynamics of insulation, homogenization and escalating commitment.
The social dynamics of insulation Early in the project, the project group began to distance themselves from their organizational environment and to focus on developing internal cohesion rather than fostering links with the surrounding organizations. This process served to insulate the group from external influence and strengthened the project group's steering of the project in a different direction than that of the original idea. This development was driven by strong individuals in the project group wanting to build a project with an empowerment-inspired approach.
This insulation process pushed the focus of the group inward regarding the project group's identification process, where this involved distancing themselves from their home organizations. This is illustrated in the first phase in Figure 3, insulation, as creating a bubble around the project group to protect and insulate themselves. The one-directional arrows illustrate the lack of information from the project to the surrounding actors, and the thick line illustrates how this barrier prevents external information or influences from reaching the project. Representatives from external organizations experienced how communication attempts failed, both concerning getting information from the project as well as reaching the project.
No, it has been very quiet, there has been very little information (. . .) And so we have been standing here knocking on the door, failing to reach them. (Representative Municipal Labor Market Unit) Figure 3. The three social dynamics identified in the study QROM When describing the insulation process retrospectively, actors in the steering group used the metaphor of a "bubble". This corresponds with the picture of the project as an isolated environment. However, unlike a bubble, it is neither transparent nor easy to break. Instead, the bubble membrane grew thicker to become an insulation barrier against the cooperative actors.
This bubble that this project group went into, it was like, it shut others out (...) it would have been easier to punch the bubble if it had been transparent and you could see how things were in day-to-day practice. (Steering group member) The insulation process was enforced by a strong identification process where the project group focused on branding the project as special and identified themselves through this uniqueness. This involved spending excessive amounts of time and money on creating a logo, producing marketing materials such as brochures, pencils and sweaters and requesting project-specific e-mail addresses instead of their regular organizational addresses. All these activities served to highlight the uniqueness of the project.
The bubble creation puts a lot of focus on the internal bonding process in the group and created a strong sense of "us and them". The orientation toward empowerment methods also stressed the use of the group as a tool, rather than promoting coordination with other actors. This inward movement could be seen in the project group's daily work, where they engaged in various group exercises to tighten the ties within the group, which also contributed to creating a distance from the home organizations.
The insulation was strengthened further by structural aspects concerning location and time. The project group was placed in their own office with limited physical access to the home organizations. In this sense, the project group became both isolated from and insulated against their environment. Instead of enabling access points, the project group focused on creating a non-authority-like environment, both for the youths and themselves, further distancing the project from the home organizations. This process of creating a home-like environment was described by a representative from the steering group as a "cozy-cozy" process, interpreted as a strategy to distance the project from the alleged hard and stressful environments in the home organizations. The organization of work hours also contributed to the insulation, as some group members managed to abandon the original idea of sharing their hours between the project and the home organizations and instead spent all their work hours in the project. This limited knowledge exchange further increased the insulation effect.
The social dynamics of homogenization Gradually, the group members began to lose their original identities as representatives for each participating organization and to assimilate into the new project identity. Based on the original idea, the staff members were to contribute each participating actor's unique competence and resources to the project. Instead, the project developed into an independent unit with group members striving for homogeneous views instead of diversity. This process is illustrated in the second phase in Figure 3, homogenization, with the group members (the small circles in the middle) losing their affiliation pattern and assimilating into a new common identity. The communication between the organizations is illustrated as a dotted line to indicate its sporadic character. The process is aptly described by a project group member as the project having created "a world of its own".
Just as in the social dynamic of insulation, the homogenization process was related to social identification. The project group sought homogeneity in the project concerning everything from identification attributes to roles. Work clothes with the project logo and the use of empowerment-oriented expressions strengthened the project members' identification with the project and further enforced the "us and them" feeling. Instead of affirming differences and using the various competencies in the group, in line with the original project idea, they focused on developing work methods where everyone would do everything, regardless of profession and affiliation. This social identification process meant the group members relinquished their affiliation to their home organizations in favor of the project identity, which was strongly criticized by steering group representatives.
And that is a little strange when one calls for different professions, and the business becomes so that one can almost go in and take over each other's work. (Representative Steering Group) Another aspect of the homogenization process was the pursuit of thinking similarly within the group. As the cohesion within the group increased, it became more insulated as an information-processing unit, thus diluting the dialectical process (illustrated as a dotted line in Figure 3). The group used a consultant to train them in a specific empowerment method, which gave them a guideline for how to think and act in the "right" way in order to succeed with the task to empower youths, thus pressuring the group members to uniformity of beliefs. The project leader had no experience of the chosen method or of project management, which led the project members with more knowledge of the method to become informal leaders. These informal leaders were, due to their experience, regarded as experts and as such they gained power and were rarely contradicted. They prioritized meetings to settle group values and thereby directed the members into thinking and acting uniformly.
The pressure to conform to the group's beliefs also led to closed-mindedness. The project group developed a culture where information or ideas that were not in line with the endorsed way of thinking were questioned, ignored and sometimes even ridiculed. This also meant distancing from and minimizing dialog with surrounding organizations. The closedmindedness meant that the project group was not susceptible to counterfactual thinking, and the devotion to the empowerment method limited the project group's openness to new or opposing information. This inhibition of critical thinking convinced the group that their interpretation and choice of method was right and prevented them from seeing or acknowledging alternatives. At times when the project's direction was questioned, the project group reacted strongly and focused on justifying their choice of the empowerment method. This further strengthened the bonding process and made the group members grow even closer by supporting one another, while weakening any cohesion with colleagues from the surrounding organizations or with their managers.
And this, that it was going in the wrong direction, that was a really great thing for us, it became a great thing for us (. . .) and then you feel you've failed, you have worked, you are passionate about this and have worked hard for it and want the best for the project and everyone. (Project group member) The social dynamics of escalating commitment The insulation and homogenization processes enforced groupthink and contributed to an overestimation of the group and a non-rational escalation of commitment. This process is illustrated in the third phase in Figure 3, escalating commitment, where the expanded circle in the middle indicates how the group members experienced a magnified sense of selfimportance. The previous communication paths are now entirely removed to visualize the total breakdown in communication between the parties.
The insulation and the dilution of the dialectical process contributed to the group overestimating themselves. This became evident in the surrounding organizations not sharing their sense of the project's excellence. Instead, representatives from these organizations questioned the project's image of having a unique function and relevance and especially the perception of being the only solution for the target group. This resulted in the group being criticized as having delusions of grandeur. I think that the team felt that they were superior, to the degree that they did not need the other professions in our organizations: they should do everything (. . .) I think they are way off track, because there are many who are experienced in having conversations with and reaching young people. It is not that this is a completely unique skill. I do not think that's good, because then you place yourself above the others and say that the others are not as good. (Representative, Steering Group) The overestimation of the group also meant the project leader's role was diminished. The project leader, according to the original idea, was supposed to have a bridging function between the project members and management in the surrounding organizations. However, the insulation and homogenization processes affected the project leader in gradually beginning to identify herself with the project group rather than as a representative of the management team. As the polarization increased between the groups, she was drawn to the project group where she felt loyalty. The project leader was later dismissed as the steering group did not trust her to be able to fulfill her task. At this point, the project group was convinced that they did not need a project leader, and that the project could be run by the group and its informal leaders.
[The project leader] sided too much with the group, she became a group member (. . .) the group ruled her (...), [there were] strong personalities in the group who were allowed to take over. (Representative, Steering group) The expanding sense of importance was also reflected in a diminished role of the steering group. Due to the insulation and homogenization process, the group had become used to managing themselves and was convinced that they had both the mandate and capacity to do so. When the steering group, through feedback from the learning evaluation, became aware of the project deviation from the original idea, they began to question the project. This strengthened the group's unyielding commitment even further, and they made every effort to question the steering group. The project group spent much time processing and trying to influence the outcome of managerial decisions by reaching out to the home organizations, to the media, to unions and to the young people enrolled in the project. In the end, this commitment only impaired the opportunities of the group to continue the project, as the steering group decided to shut it down.
They had too much room for maneuver, and they got too much to work against and they tried to make us responsible for the closure. . . Of course, you can get angry for a while (. . .), but then I think you have to move on (. . .). Sometimes, it became very disproportionate. (Representative, Steering group) Another aspect of the non-rational escalation of the project group's commitment was the development of downward leveling norms. These behaviors escalated in connection to the decision to close, and group members displayed behaviors that were not in line with how they acted in the past, e.g. rallying over the financier and making threats against the steering group. Apart from the negative effects of these dubious behaviors for the involved actors, interviews with the project group members indicate negative effects also on the ones who performed them.
Yes, it is really bad. I mean, no one wants anything bad for anyone else, but it gets to that point, along the way. Through all the frustration, misconceptions and misunderstandings that arise along the way. And it's really sad, but that s the way it is. (Project Group Member)

Discussion
The aim of this study was to explore theoretical assumptions of negative consequences regarding social capital in the empirical case of a failed cooperation project between public organizations. In so doing, we focused explicitly on social processes within the project, where the organizing of the actual practice was given centrality. By studying such processes, we could identify three distinct phases characterized by three corresponding social dynamics, namely insulation, homogenization and escalating commitment. In this section we will relate these processes and dynamics to theories of shadow organizing and the development and consequences of negative social capital.

Shadow organizing
By studying the three social dynamics in light of the "shadow organizing" metaphor, we can contribute new understandings and perspectives of the negative consequences of social capital (Gherardi et al., 2017). We use this metaphor for interpreting the interacting elements, making it possible to illustrate dynamics of organizing in the interstices of a formal project structure. The different perceptions of the project may be seen as a tree and its shadow. The tree stands for the initially communicated image of the project as a cooperative arena. This view is based on the original idea of the project and illustrates what was expected from the outside. The shadow is a symbol of what is "betwixt and between", i.e. representing the actual development of the project. The metaphor is used to look past the obvious and focus on the hidden, thereby capturing the social dynamics appearing within the project.
First, the social dynamics of insulation can be understood as a process of secrecy. The project group's inward focus and distancing from cooperative actors was described as the "creation of a bubble" and the limiting of transparency into the project relates to the use of the organizational shadow as a secluded environment enabling processes to develop unnoticed. A relatively absent steering group and strong group members created possibilities for the group to act without insight, which made the project develop for a relatively long time before anyone noticed any deviation from the original idea.
Second, the social dynamics of homogenization can be understood through the concept of liminality, which describes the grey zone between canonical and non-canonical practices. There is usually a discrepancy between what is prescribed and what is actually practiced. According to Gherardi et al. (2017), professional discretion is a good example of how organizing and working emerge in a grey zone which leaves room in which professionals can maneuver. The social dynamics of homogenization can be understood as a consequence of this liminality, where the project group was given room to design their own roles and assignments.
Third, the social dynamics of escalating commitment can be understood through the perspective of performativity. Operating under the shadow of group seclusion, the group created a productive and protective space. This penumbra, or the half-lit environment of the project, enabled unintentionally favorable growth for strong social identification processes, which resulted in overestimation of the group and downward leveling norms. Building on this self-perceived strength, the group tried to use its social capital to act against external impulses to make the project change direction, by various actions, such as turning to the media. The penumbra of the project organization hence created a specific "ecology" for the group, where the shady organizational undergrowth bred a specific performativity.

Negative effects of social capital
While shadow organizing is a useful metaphor for describing and illustrating social dynamics, the theories of social capital and its potential negative consequences are valuable in understanding these dynamics. The development within the studied project will in this section be related to the negative effects of social capital as outlined by Pillai et al. (2017) and Portes (1998) and will further be related to the distinction between bonding, bridging and linking social capital (Szreter and Woolcock, 2004).

QROM
The three social dynamics described in the results are highly relational in character and can be understood as unfolding through several phases. Figure 2 illustrates the starting point of the group dynamics, where the original idea of the project was for the project staff to act as representatives for their home organizations and to stimulate cooperation between the project and cooperative actors. This corresponds to the idea of fostering bridging social capital by focusing on establishing horizontal relations between persons at the same hierarchical level from different backgrounds (Szreter and Woolcock, 2004). As shown in the results, this idea was not fulfilled in the studied practice. The project saw a rapid development from heterogeneity to homogeneity. While the project design and staffing were based on the idea of placing different professions in the same space and fostering communication with their home organizations through a part-time staffing structure, the project soon evolved into a singular information-processing unit, where differences between the professionals were gradually erased. The staffing structure changed into promoting full-time project employees rather than sharing work hours between the project and the home organizations.
While bridging social capital is based on the combination of differences, bonding social capital celebrates similarity and focuses horizontal relations of trust and reciprocity between individuals with similar social identities at the same hierarchical level (Szreter and Woolcock, 2004). This was enforced in the project by the social dynamics of insulation and homogenization, as illustrated in Figure 3, where the project group distanced themselves from the surrounding organizations, which enabled them to focus on internal group and communication processes. Pillai et al. (2017) describes such group insulation processes as one aspect of groupthink, which serves to exclude others through promoting introversion and disinterest in communication with actors outside the group. This broken connection to the home organizations induced the group to strive for a common project identity. The following homogenization phase saw a further increased inward focus and limiting of external impulses which resulted in shared sense-making and confirmation bias within the group, with uniformity of beliefs as a consequence. These negative communication effects of social capital can be understood in relation to the negative effects described by Pillai et al. (2017) as dilution of the dialectical process, inhibition of individual learning and postponement of structural adjustments which may be needed to function in relation to the project's environment.
After the social dynamics of insulation and homogenization had fueled the development of bonding social capital in the group, a social dynamic of escalating commitment commenced. The lack of communication and mistrust between the parties solidified a polarization between those inside and outside the project and resulted in a power struggle. Pillai et al. (2017) describe such escalating commitment as non-rational and as a breeding ground for downward leveling norms which escalate negative behavior. As a result, the project experienced a communication collapse and established an alternative hierarchical order. This hampered the potential for developing linking social capital, i.e. vertical relationships between people interacting across gradients in society (Szreter and Woolcock, 2004). Linking social capital did occur, but as the project actors polarized, so did the relationships between organizational levels; linking social capital was evident among the allied in the respective fractions. The project group developed vertical relationships with the project leader during the insulation and homogenization processes, which added to the project group's feeling of independence by having representatives from a higher organizational level on their side. The steering group however, being officially the highest hierarchical level in the project organization, was perceived as a threat to the project.
Bonding social capital is discussed in the literature as the type of social capital that has perhaps the strongest inherent risk of producing negative effects, as it tends to develop the formation of closed networks and groupthink among members (Pillai et al., 2017). This assumption was supported by our data: the bonding social capital identified had strong negative effects that in the end forced external actors to dissolve the group. Pillai et al. (2017) discuss whether there is a threshold effect of social capital where, up to a certain level, it would mainly have positive consequences, but after the threshold is passed it will produce negative effects. Our results indicate that this process is not linear, but affected by dynamic interactions within social structures, where attention should be placed both on the personal relationships and the organizational or structural levels. In the studied practice, the project was organized to promote bridging and linking social capital but was effectively usurped by strong informal leaders who steered it toward a strongly bonded social group which promoted similarity and resistance to the organizational environment.
Although this bonding process unfolded through different phases, this does not imply that what is described here necessarily illustrates a typical trajectory of cooperative projects. The phases in this case were influenced by vague instructions, weak management, strong informal leaders and external influences by a consultant which, when combined, shifted the project from its initial design toward another development and another rationality (Strindlund et al., 2020). The organizing of the actual practice took place in the shadows, making it immune to insight from the steering group. As described in the shadow organizing literature (Gherardi et al., 2017), secrecy and liminality play important parts in allowing for alternative non-canonical practices to develop, where the bonding social capital described here can be explained as a non-canonical practice. In this sheltered space, the project staff could perform their organizing activities and initially even be unaware that they were taking non-sanctioned courses of action. This was revealed only later, after the strong bonds had already been established, causing polarization and communication breakdown and in the end the closure of the project.
What grows in the shadows? By interleaving the theoretical assumptions of negative social capital with that of shadow organizing in the context of an empirical study, we can begin to make out the contours of how these are related. The literature emphasizes how social capital is a complex concept which involves both individual and collective dimensions: from cognitive to interpersonal structures (Rostila, 2010;Pillai et al., 2017). Our study illustrates how organizational conditions, organizing practices, relational structures and social identification processes interacted with cognitive structures, e.g. through the retrospective self-reflection from previous group members on how they got carried away and behaved in ways which previously would have been alien to them. The negative effects of social capital thus had both its causes and its consequences in such dynamic interactions, where it would prove difficult to identify explicit and causal relationships. Rather, the shadow organizing metaphor provides a vocabulary to make sense of the ecology in which social capital may grow into weeds rather than flowers. By focusing more specifically on the non-canonical practices within an organizational structure, we can identify processes that develop in the penumbra and which defies insight from external points of view. In such protective spaces, relationships can be established which result in strong cognitive biases toward specific views, which in turn may prove difficult to reconcile with those views championed in the formal organization.

Methodological considerations
We found the theorizing process described by Swedberg (2012) to be a valuable methodological tool in structuring and guiding the data analysis. In combination with the ethnographic longitudinal approach which allowed a study of the social interactions within the project and in relation to cooperative actors, the theorizing process enabled the identification of the social dynamics which our results are based on.
The research design was interactive in character, meaning that the first author was observing the project and also reported back observations to the project and the QROM management. This can be seen both as a strength and a limitation, where the strength is that this facilitated proximity to the project and insight into the everyday developments. A potential limitation with this strategy is that the researcher becomes involved in this development. Reports from the researchers informed management decisions, including the decision to shut down the project, although this was not a recommendation from the researchers. We have throughout the research process tried to manage a balance of closeness and distance, which included refraining from offering specific recommendations for decisions and to clearly communicate that the researchers had an observing role.
The credibility of the study was strengthened by the broad data collection (comprising observations, interviews, focus groups and documentation) and the inclusion of perspectives from several different actors in the analysis. The transferability of the study results is limited to similar cooperative projects between authorities, which means the results should be interpreted according to the specific context. The dependability of the study was strengthened by the joint analysis process where the authors collaborated and contributed with different perspectives; further, they had varying closeness to the empirical context, where the first author knew the context well, while the other authors could take more of a critical outside view. The trustworthiness of the study was strengthened by the interactive approach, where the results were reported back to the organizations and where the results have been acknowledged and confirmed.

Conclusions
This article has shown how dysfunctional group processes in a cooperation project can be understood from theories addressing the dark side of social capital (Pillai et al., 2017) and the theoretical metaphor shadow organizing (Gherardi et al., 2017). By analyzing a failed cooperation project between Swedish public organizations, we have shown how negative social capital can be a consequence of poor organization and its interaction with different social dynamics. Hampering of communication, transparency and cooperation, in combination with a homogenized group identity, led to polarized relations which effectively blocked any attempts at communicative or cooperative action. These dynamics occurred in the dark and remained unrecognized by management until late in the project. Our study showcases how these processes lead to deleterious consequences, where strong internal cohesion in a project group caused upper management to shut the project down.
The study was carried out in the context of the Swedish public sector, which could, depending on one's preconceptions, be expected to involve relatively consensus-oriented bureaucrats. This was however not the case. Instead, our results point to high levels of conflict between people grouping together to form tight bonds, and the official organization of the project was hi-jacked and used as a vehicle for non-intended ends. We may hence conclude that when it comes to public services, a Swedish setting is no guarantee for consensus or organizational compliance. The social relations described in this study reek of contempt and hostility, and the theories of negative effects of social capital and shadow organizing serve well as tools for understanding the dynamics that created the milieu for such relations to develop.
The contribution of our analysis is how negative effects of social capital are realized in practice through social dynamics. We specifically identify three such dynamics: insulation, homogenization and escalating commitment. These can be used as a terminology identifying dysfunctional identification processes which can characterize interorganizational projects. Here, the boundaries of the project became increasingly non-permeable due to the tight relationships within the group, which made influence from the external milieu nearly impossible. Knowledge of negative effects of social capital and the risks of letting projects become both isolated from and insulated against their environment can increase the understanding of involved actors' different perspectives and driving forces in interorganizational cooperation. It also emphasizes the importance of prioritizing communication, structure and steering, and building trustful relations between involved actors when organizing new interorganizational projects in order to avoid cooperation to turn ugly.
A theoretical conclusion from the study is that the effects of social capital are conditional upon complex interactions within social structures. In this study, we have focused particularly on conditions leading to negative effects, which do not imply that we consider social capital as such to have negative connotationsrather, the concept is theoretically neutral and may be used for analyzing both positive and negative consequences of social organization. As for the negative effects, one interesting finding from our study was that social capital did not develop in a linear fashion where a threshold was reached when its consequences turned from positive to negative. Rather, the effects of social capital were more unpredictable and dependent on a dynamic interplay between people, structures and environments, which influenced the trajectory of the organizing processes. The case described here is illustrative of this point, and we can reasonably assume it is not unique in this respect.
We also combined social capital theory with the metaphor of shadow organizing, which proved to be a useful tool to identify unrecognized organizational practices in which social capital may develop. The combination of these two perspectives can offer explanatory value for determining the organizational and interpersonal dynamics which may lead to negative effects of social capital. The metaphor of shadow organizing lends itself well to the context of interorganizational cooperation between public organizations, as there are strong canonical practices often prescribed in legislation. The metaphor allows for an analysis of what happens when such regulated organizations jointly creates a space between them, where the regulations have less reach. This space, which using the metaphor can be described as a penumbra, is only partly lit and hence obscured from insight. The social dynamics described in our analysis serve to explain how this process resulted in a group strong enough to try to resist the dominant organizational logics of their surroundings.
A practical conclusion from the study is that anyone organizing cooperation projects that aim to bridge professional competencies or organizational boundaries have to be attentive toward the processes within an organization that takes place beyond the canonical and illuminated practices and designs. In this project, protective spaces were created in the penumbra which became a breeding ground for unexpected practices. If such practices are allowed to grow without insight, they may end up threatening the original intentions and organizational goals. By remaining in the shadows, the people engaged such in processes and organizing practices may also be questioned by those standing in the light, since their actions will appear dubious when viewed from the outside. As a result, the performativity of shadow organizing produces an unintended effect of polarization between actors, where we may conclude that such hostility might be prevented if a dialog is established throughout the organizing processes; those lurking in the shadows may, after all, prove to be quite reasonable people.