Keywords: Network Analysis, Organizational Communication, Terrorism, Dark Organizations

p. 1

Abstract

Although not as common as more traditional organizations like firms, nonprofits, and governments, terrorist organizations also engage in interorganizational communication (IC) relationships. The purpose of this paper is to determine the relationship between IC relationships and terrorism. To do so, we analyzed a stable alliance network of 395 terrorist organizations between 1998-2005. Using Borgatti and Everett’s (2006) four-prong typology of centrality, we assessed how degree, closeness, betweenness, and distance-weighted fragmentation contributed to the identification of key players in the network. The results found that all four sets of key players were more likely to commit more terrorist acts and kill more people from 1998-2005. However, when entered into a full model, key players identified by closeness centrality consistently predicted both indicators of terrorism while the other centrality metrics did not. We discuss the results in light of communication theory, dark organizations, and network science.

Keywords: network analysis, organizational communication, terrorism, dark organizations

Using Interorganizational Communication Networks to Predict Terrorism

“Without communication there can be no terrorism.” (Schmid and de Graaf, 1982, p. 9)

In communication, the networked form of organizing was first explicitly applied to terrorist organizations by Stohl and Stohl (2007), arguing that some of the findings and themes of network theory have significant implications for understanding terrorist structures. Namely, they demonstrated that terrorist networks are multifunctional, multiplex, emergent, boundary specific, and simultaneously local and global. Indeed, General Stanley McChrystal and colleague’s (2015) recent text argues that the US Army needs to complicate itself and match the decentralized network structure of the terrorist organizations they face if governments want any chance addressing modern threats, a sort of requisite variety so to speak.

Although network explanations of terrorist organizing have received much attention, these accountsmostly refer to intraorganizational structures between members and cells of individual terrorist organizations. For example, some of the most well-cited network studies explicitly examine single organizations and the complex webs of connections within those organizations (see Knoke, 2015, for a summary). Namely, what has received less attention amid the growing literature on communication, networks, and terrorism is interorganizational communication (IC) between terrorist organizations (see Karmon, 2005).

To address this limitation, we focus on interorganizational terrorist alliances, defined as “joint or complementary action for the same (intermediate) purpose. This action can constitute activity at the rhetorical, material or operational levels” (Asal et al., 2015, p. 13-14, emphasis in original). According to Shumate and Contractor’s (2013) taxonomy of communication networks, alliances between organizations fall under affinityties because they represent enduringrelationshipsthat focus less on information flow, but more on the negotiation and meaning-making of socially constructed ties.In particular, terrorist interorganizational alliances are important because, although they are rare (LaFree & Dugan, 2007), they are not a new phenomenon. Indeed, Sterling (1981) first revealed a complex web of connections between organized actors from Europe, Latin America, and the Middle East during the Cold War. Moreover, the impact of interorganizational alliances between diverse terrorist organizations, what a recent CNN article labeled “marriages from hell” (Lister, 2015), can be catastrophic because they allow for the exchange of resources that may mutually benefit both parties. Instrumentally they have been responsible for many large-scale attacks (e.g., Bali bombing in 2002 influenced by an alliance between al-Qaeda and Jemaah Islamiyah), and symbolically they contribute to the co-construction of a culture that perpetuates terrorism as a legitimate tactic (Matusitz, 2014). As such, IC between terrorist organizations can “bolster their operational effectiveness, range, and efficiency as well as enhance their legitimacy and stature” (Bacon, 2014, p. 4).

As such, the purpose of this paper is to examine the relationship between IC and terrorism. In other words, are terrorist organizations in alliances more likely to commit deadly and frequent attacks? To answer this question, we focus on terrorist alliances as risky endeavors that can influence the organization’s capacity to mobilize diverse resources needed to carry out attacks, following the basic tenets of network social capital theory (Lin, 1999). Methodologically, we investigated the international terrorist alliance network from 1998-2005 (n = 395). Using the union of two different datasets we conduct a key player analysis (Borgatti, 2006), which selects specific organizations using different network metrics. Our analysis uses four different network centrality metrics: (1) degree, (2) closeness, (3) betweenness, and (4) distance-weighted fragmentation.

This paper makes three main contributions related to communication theory and research. First, though our work is cross-sectional, it begins to demonstrate some of the possible outcomes of IC networks, a trajectory of research that is not as common as the analysis of antecedents of such relationships. Here, we underscore the need for theory investigating the link between IC networks and organizational capacity, expanding network theories of social capital by specifying differences across network positions rather than treating them all the same. Second, we demonstrate how the study of dark organizations challenges our current understanding of IC theory. Previous theory and research have neglected dark organizations, and we posit that current IC theory has challenges in explaining the mechanisms governing dark IC relationships. Finally, we use a valuable tool for network science, key player analysis, to identify prominent terrorist organizations. Our results suggest that while all four network metrics were adequate predictors of terrorism, closeness centrality demonstrates a greater capacity for identifying more dangerous terrorist organizations. Traditional measures of closeness have been criticized in many network circles. The current research challenges this idea and calls for a reexamination of the measure.

Interorganizational Communication and Terrorism

There is a plethora of definitions of terrorism (Crenshaw, 1981) and no grand, uniform agreement on how to use the term. As such, we follow Matusitz’s (2013) definition, which is a synthesis of other widely accepted features from previous definitions. Simply,terrorism is defined as the “use of violence to create fear (i.e., terror, psychic fear) for (1) political, (2) religious, or (3) ideological reasons” (p. 4). Beyond this definition, though, it is important to view terrorism as a communication process (e.g., Fischer et al., 2010; Matusitz, 2013; Schmid & de Graff, 1982; Tuman, 2003), rather than as static behavior or cognition located in a vacuum.In this sense, terrorism involves communicative interactions between two key actors: (1) those committing terrorist acts and (2) an intended audience. As such, to borrow from Griffin and McClish (2014), a communicative view of terrorism can be seen as the relational process of creating and interpreting messages that elicit a response within and between terrorists and an audience (e.g., the wider public, governments, media, corporations, etc.). One of the implications of such a view is that it shifts our thinking to focus on underlying communication structures and systsems that further perpetuate terrorist events and meanings, rather than psycholgocial antecedants behind committing terrorist attacks.

Indeed, one underexplored avenue that is critical to the communicative process of terrorism is interorganizational communication (IC). IC arethe “structures, forms, and processes created by the exchange of messages and the co-creation of meaning among organizations and their stakeholders” (AUTHORS, 2016a, p. 1). Though they are sometimes mandated (Stephens, Fulk, & Monge, 2009), one of the reasons why organizations develop IC relationships is to accomplish something that could not be done in isolation. Things like extracting resources from one another (Monge, Heiss, & Margolin, 2008), constituting a collective identity (Koschmann, 2013), maintaining civil society (Doerfel & Taylor, 2004), responding to a crisis (AUTHORS2016b), and attracting other various forms of social, economic, and political capital (AUTHORS, 2014; Shumate & O’Connor, 2010), are all possible through establishing IC relationships.

One way to analyze IC relationships is through the use of social network analysis (SNA). By using networks to map various IC relationships, researchers can uncover how embedded some organizations are within the network. Structural embeddedness refers to different positions an organization occupies in an IC network. As such, a key assumption of SNA is that different attitudes, identities, and behaviors are due to network position, rather than individual characteristics (Wellman, 1988).

In the current study, we hypothesize that organizations in certain IC network positions are likely to commit more deadly and frequent terrorist attacks than others, partly due to the resources garnered from those alliances. For instance, according to Lin’s (1999) theory of network social capital, positions in a network allow for four advantageous processes that can not be explained merely by looking at individual, organizational capital like human (e.g., membership) or economic (e.g., finances, technology) resources. First, there is access to information flow. For instance, developing alliances may garner specialized knowledge and skill on how to effectively conduct attacks. An example is Miller’s (1985) research, which documents how members of the Japanese Red Army (JRA) connected with the Popular Front for the Liberation of Palestine to learn about various tactics that the JRA could not obtain from other like-minded organizations in the area. Second, positions in networks may allow actors to influence others. In this case, alliances can be used to attract and even convert members to join their group (Scott, 2013, p. 127) and thus, acquire more resources for attacks. An example is Al-Qaeda’s consolidation of Salafist Preaching and Combat (Gerwehr & Daly, 2006). Third, network ties “may be conceived by the organization or its agents as certifications of the individual's social credentials” (Lin, 1999, p. 31, emphasis in original). In this case, resource dependent relationships might develop between organizations, increasing the organization’s reputation as being a desirable group to connect with for certain resources. Social credentialing often occurs in environments with an unequal resource distribution. For instance, Cragin and colleagues (2007) documented how various terrorist organizations in the Mindanao, West Bank, Gaza Strip, and southwest Columbia have formed alliances to exchange various technological resources for increasingthe effectiveness of attacks. Finally, positions in networks reinforce identity and recognition, providing a spirit that can bolster the legitimacy of an organization and hence, increase motivation to continue attacks. For instance, some have claimed the recent alliance between Boko Haram and ISIS is more about image and institutional positioning (see McPhee & Zaug, 2000), rather than extracting tangible resources (Moftah, 2015; Tisdall, 2015). Together, these four elements may “explain why social capital works in instrumental and expressive actions not accounted for by forms of personal capital suchas economic capital or human capital” (Lin, 1999, p. 31).

Following the creation of a terrorist alliances, an interorganizational system begins to develop as is common is most self-organizing systems (Contractor, 1994). In this system, organizations can occupy different positions and spaces in the network. However, little empirical work exists investigating how different IC network positions might enable an organization to carry out more catastrophic attacks than others. In other words, it is not just the quantity of IC relationships that matter, but instead the structure of these relationships. Although network social capital theory provides explanations for why establishing network ties matters, it does not specify which network positions lead to which outcomes. The next section reviews relevant literature and develops hypotheses for four measures of centrality, which refers to the varying structures of organizations’ connections with others. Moreover, alternative hypotheses are also developed based onindividual, organizational characteristics, including other organizational resources and ideologies.

Four Dimensions of Network Centrality

To examine how positions in IC relationships might influence terrorism, we look at centrality, perhaps the most ubiquitous of all network metrics. However, because there are a variety of different centrality metrics, we use Borgatti and Everett’s (2006) unified framework of centrality, which classifies centrality across two dimensions: (1) walk property (length vs. volume) and (2) walk position (radial vs. medial).

We start by explicating several definitions, allowing the reader tounderstand the typology fully. First, there is the concept of a walk. A walk “from node u to node v is a sequence of adjacent nodes that begins with u and ends with v” (Borgatti & Everett, 2006, p. 468). Many researchers are interested in short walks (i.e., geodesic distance). For instance, consider Boka Haram and ISIS. Previously, it might have been necessary for Boka Haram to go through several organizations to reach ISIS. The number of alliance relationships that this organization may have to go through to reach any given organization can be interpreted as the length of the walk: the higher the length, the more degrees of separation they have. However, now that there is a formal alliance between Boka Haram and ISIS, that length is reduced to one because they have a direct connection between one another. When researchers add up the number of alliance relationships (i.e., lengths of one) an organization has produced, they are said to assess the volume of those walks. Volume and length make up the walk property dimension of the typology.

More important however (Borgatti & Everett, 2006, p. 480), is walk position. Radial measures “assess walks that emanate from or terminate with a given node” (p. 473). For instance, when we start with Boka Haram and measure all of its walks to every other terrorist organization, we are looking at how alliances radiate from that organization (i.e., Boka Haram is the center-point). On the other hand, medialmeasures look not at how Boka Haram is not at the start point, but at the middle point, based “on the number walks that pass through a given node” (p. 473-474). Thus, when we look at how Boka Haram lies in a path between other terrorist organizations, we are looking at how Boka Haram mediates alliances between those organizations. The benefit of the typology is that it offers a heuristic guide for researchers to make sense of different centrality measures and can empirically provide a guide to test which measures might be superior in different contexts. In the next section, we hypothesize four measures based on each quadrant of the typology of centrality metrics: (1) degree, (2) closeness, (3) betweeness, and (4) distance-weighted fragmentation.

Radial volume: Degree centrality. The most commonly used network metric is degree centrality. Degree centrality is simply the count of direct connections that each node has. In the current context, it refers to how many alliances each organization has in the network.

Theoretically, degree centrality is important because we associate this metric with power, prestige, popularity, and action (Wasserman & Faust, 1994). The logic is quite simple: the more alliances an organization has, the more likely it will have to chance to obtain various resources it may need to orchestrate effective terrorist attacks:

H1: There is a positive relationship between degree centrality and terrorism

Radial length: Closeness centrality. Sometimes it is not the quantity of short walks, but the length of those walks that may make a difference. Closeness centrality is one measure that takes this into account by summing all of the shortest paths (i.e., geodesic distances) from a focal node to every other node in the network. Nodes with short paths to many other nodes in the network can reach any other node or resource quicker than nodes with just high degrees because degree centrality does not take into account walk length.

In a terrorist alliance network, the main advantage of having short walks is having access to diverse resources, especially if those resources are not distributed equally throughout the network. For instance, if certain technology or weapons were available to only a few organizations in the network, those with short path lengths might have a better chance at obtaining those resources because they can go through a mutual contact that is not far away. Moreover, those organizations that are close in the alliance network may be considered more legitimate because they are central yet known by most of the network, even if only through their indirect connection. As such, we hypothesize:

H2: There is a positive relationship between closeness centrality and terrorism

Medial volume: Betweenness centrality. Medial measures look, not at the walk from A to B, but at how A might lie between B and C. That is, it looks at how A might serve as a broker, bridge, or liaison between any two unconnected nodes. In practice, Burt’s (1992) theory of structural holes argues that there are key resource benefits for nodes with high betweenness centrality because they are more likely to get non-redundant and novel resources (e.g., information about innovations).

Occupying a position between pairs of unconnected organizations in an alliance network can yield resources that might make it easier for organizations to carry out attacks. For instance, when organizations can mediate between others, those indirectly connected organizations might become dependent on the broker for information or other resources because the broker is their source for network capital. Thus, power comes in the form ofcontrol for those with high betweenness in the sense that the broker is the only outlet. Thus, the broker benefits from a lack of connection in the network and would ideally try to fill those structural holes. As such, we hypothesize:

H3:There is a positive relationship between betweenness centrality and terrorism

Medial length: Distance-weighted fragmentation (DF). Borgatti (2006) proposed DF to identify key players in any given network. The measure looks at the average reciprocal distance of a network given the hypotheticalremoval of a node. If the network’s average reciprocal distance increases with a node’s removal, the node is considered influential (i.e., a key player) as its absence makes it more difficult to reach the other nodes in the network.

Although the presence of brokers ensures that structural holes are filled, nodes that would influence DF are better thought of as assemblers, as their presence ensures that an alliance network is not fragmented and un-unified. Without these assemblers, the network would be a large-world, rather than a small-world, requiring organizations to go through a high number of organizations to reach others for resources. Thus, for assemblers, resources may constantly be flowing and passed through them because they occupy positions within the shortest path of any two pairs of organizations. Hence, this rationale is why DF is frequently used to identify ‘key players’ in networks (Borgatti, 2006):