Introduction: Civil Society, Civil Society Organizations, and Social Networks

The concept of “Civil Society” had achieved greater theoretical importance in the eighties (Cohen and Arato 2000; Kaldor 2003), becoming even ‘fashionable’ in the last decade of the past century and in the beginnings of the present one. As with similar concepts, there is no consensus about its meaning; not because there has not been enough theorization around it but, as Rocut (2009) points out, because no concept is innocuous, even more so when it serves the purpose of defining how human collectives work; paraphrasing Houtart, cited in Rocut (2009): when the World Bank talks about civil society, it refers to a reality that is completely different from the one the same term conveys for Thailand’s Forum of the Poor or the Peasants Without Land movement in Brazil. In this text, we adopt Cohen and Arato’s (2000) approximation where Civil Society corresponds to the whole social subsystem that originates in the differentiation of the Lifeworld from the State and the Market. Without doubt, one of its more active and important expressions refers to Civil Society Organizations (CSOs), organizations that were previously referred to as Non-Governmental Organizations (NGOs) by the United Nations. CSOs emerged in Latin America in the sixties as social actors with a critical position toward government institutions, contesting the role of a centralized State as sole responsible for social problems and only designer of the policies geared toward solving them. But it was mostly in the eighties when they became more visible, most likely because they became a way out of the crisis of the welfare state and a key component of a democratizing and more participative trend within Civil Society. In this sense, paradigmatic examples are the civil society’s response to the 1985 Mexico City earthquake (Monsiváis 1992), the conformation of the Movimento dos Trabalhadores Rurais Sem Terra (Landless Worker’s Movement) in Brazil, the Cocalero Movement (coca cultivating peasant’s movement) in Bolivia, and Human Rights organizations in the Latin America region.

In the last decades of the twentieth century, civil society organizations started grouping into what became known as networks, quickly becoming major actors within Latin American societies. Among the reasons that explain this phenomenon we can consider the following: the process that led to the emergence of social network science with its powerful mathematical foundations (Watts 2006; Barabási 2002) and the appearance of the Internet and some of the social dynamics of globalization (Giménez 2003; Castells 1999). These phenomena are closely connected to the proliferation of interactions that do not follow classical territorial logics (Haesbaert 2011) and that have given rise to a radically new type of social structure that favors diverse expressions. For example, the Neozapatism’s netwar (Ronfeldt et al. 1998), the so-called “new social movements” (Mellucci 1996; Fernández and Riechmann 1994; Dabas et al. 1995), Internet platforms based on social networks like Facebook or Twitter (Boyd and Ellison 2007), social mobilization in several countries (the Arab Spring, the 15-M movement in Spain, the Chilean student movement, Occupy Wall Street, the 15-O movement, #YoSoy132, etcetera), the emergence of netizens (a portmanteau of the English words Internet and citizen), and the cyberactivism of organizations like Moveon.org, Change.org, Causes or Avaaz (entities that work in global networks of online signature campaigns to back public petitions seeking to influence decision-making processes).

Thereby, the civil society organizations networks in the Latin American region are increasingly participating in public policy advocacy. There are many studies that address them, but they do it through more in qualitative methodological approaches but there are few analyzed from a social network analysis approach. We present a case study that analyzes the American Network for Intervention in Situations of Social Suffering (Red Americana de Intervención en Situaciones de Sufrimiento Social, RAISSS), a very significant experience in the creation of a large transnational “network of networks” of civil society organizations from 15 Latin American countries that work with the same meta-model, called ECO2 (Machín 2013). The network develops intervention programs geared toward decreasing social suffering and promotes social inclusion and public policy advocacy.

Context: The American Network for Intervention in Situations of Social Suffering

During the seventies and eighties, three pioneering Mexican civil society organizations emerged to serve to the youth and their communities: Hogar Integral de la Juventud (Integral Home for the Youth), Centro Juvenil Promoción Integral (Integral Promotion Youth Center, Cejuv), and Cultura Joven (Youth Culture). Hogar Integral de la Juventud is a public assistance institution.Footnote 1 that has pioneered treatment for youth with problematic use of psychoactive substances and soon became one of the first Therapeutic Communities for drug abusers in Mexico. Cejuv, for its part, originates in a Mexico City neighborhood in response to the social agitation produced by the chavos banda (gang youths) phenomenon. The organization came up with a program for the promotion of youth communities (creating entities called Centro Juvenil de Barrio, Neighborhood Youth Center), developing strategies that would become a benchmark in Latin America. Cultura Joven formed in the State of Morelos when a wide network of diverse youth programs and experiences—related to human rights issues, environmentalism and cultural and artistic activities—came together in 1987.

Under the advice of Italian experts (Roberto Merlo and Efrem Milanese) and funded by the European Union, the German government, and the German agency Deutscher Caritasverband (DCV), these three organizations undertook, from 1995 to 1998, research to develop a model to prevent, treat, and reduce the impact of drug dependency and socially rehabilitate addicts. The main result was the ECO2 meta-model (Machín 2010). ECO is a play on words closely related to the model’s elements: Epistemology of Complexity (ECO), Ethics and Community (ECO). Using an algebraic analogy, (ECO) × (ECO) = ECO2. The word ECO also refers to the Greek root that means house, conveying the social inclusion processes that it promotes. ECO2 is currently the theoretical and methodological framework that has been used to train more than 3,000 people from more than 300 CSOs in 17 countries in Latin America and the Caribbean. In recent years, experiments have been carried out using the ECO2 model in Bangladesh, Afghanistan, India, Pakistan, and even Eastern Europe. The participation of all these organizations and experiences has enormously enriched the model, widening its field of application. The model has been used beyond the field of drug dependency and has already been applied in the attention of social suffering situations: HIV-positive and/or people with AIDS, juvenile offenders, homelessness, migrant indigenous people, severe social exclusion, stigmatized youth, etc. The CSO networks known as “Red Mexicana de Organizaciones que Intervienen en Situaciones de Sufrimiento Social” (Mexican Network of Organizations that Intervene in Social Suffering Situations, REMOISSS) and “Red Centroamericana de Intervención en el Sufrimiento Social” (Central American Network of Social Suffering Intervention, RECOISS) emerged on the basis of these formative and experimental processes. The former became officially instituted in 2002 and the latter in 2004. In 2005, a network called Red Americana para la Intervención en Situaciones de Sufrimiento Social (American Network for Intervention in Situations of Social Suffering, RAISSS) with organizations in Central America, Chile, Colombia, Brazil, and Mexico was established and its members agree to promote similar networks in countries that lack one. In this way, RAISSS coordinates 142 organizations in nation-wide networks in Argentina, Argentina, Bolivia, Brazil, Colombia, Costa Rica, Chile, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Panama, Paraguay, Peru, and Uruguay.

A Preliminary Approach to Social Network Analysis

The first approximations to networks in mathematics (graph theory, introduced by Leonard Euler in 1736), psychology (Moreno 1953), or anthropology (Barnes 1954) were centered on their analysis as static relational structures, i.e., entities that are invariant across time. Numerous contributions on the analysis of these structures such as indicators like density, incidence, diverse centrality measures, etc. followed Scott 2000. To exemplify this approach, we present a graph of relationships within REMOISSS (Machín et al. 2010) (Fig. 1).

Fig. 1
figure 1

Relation network of REMOISSS. All calculations and figures were developed using NetMiner 4.0.0 Cyram (2011)

Full size image

The network has 23 nodes, with 125 links among them; each link indicates an existing relation between the connected nodes. The structure has an average of six links per node (incidence) and a density of 30 %, which means that out of all possible relations among nodes, approximately only one in three exist.

CSO networks establish relations not only within them but also with other CSOs that are not part of the network and other social agents (universities, government entities, funders, local communities, businesses, other CSO networks, etcetera). The complexity of this type of relational structures grows quickly with the number of nodes in the network. The following figure shows the network of relations among the CSOs that belong to REMOISSS and those established with other actors.

There are 792 nodes in this case, with 1753 links among them. This network has a 0.3 % density and an incidence of 2. In 2011, a similar analysis was carried out for RAISSS-Colombia (Machín 2011). We could carry out the same exercise for every national network and then analyze the whole RAISSS network but we did not have the necessary funds to complete a study of such magnitude. If every network behaves like REMOISSS does, the whole RAISSS network would surpass the 5000 nodes and the links would amount to over 10,000.

From Social Network Analysis to Complex System Dynamics

As we have seen, a network’s graph—something like a photograph of the CSOs in the network—is quite interesting and allows us to analyze several aspects of its structure. However, another interesting question is understanding how the structure in the picture came into being, how a concrete topology originated; to put it this way, we would like to see the whole sequence of photographs, i.e., the whole movie, the evolution of the network over time. We are now dealing with the network’s dynamics and the temporal dimension must be introduced. This is, precisely, what has given rise to the science of networks (Watts 2006).

The first steps in this direction were probably developed by Erdös and Rényi (1959, 1960), who considered that the process of network construction is totally random, which is why their model is known as “random networks” or Erdös–Rényi model. In this model, the starting point is a set of unconnected nodes; next, the first node goes through a random process (as in a coin toss or dice roll) to determine if it will be connected to the rest of the nodes or not; then, the process is repeated for every other node in the network. In this way, connections are equiprobable, i.e., they have the same probability of being established. Obviously, the result of this process will depend on the probability used (e.g., 1/2 in the case of a coin toss or 1/6 if we chose a single number in a die roll) and the resulting graph will thus depend on the number of links that become established. Each possible resulting graph structure is also equiprobable and equal to 1 divided by the total number of possible graph combinations. An interesting feature of this model is that it exhibits an abrupt phase transition (a function of the average number of links) between a state where all nodes are disconnected and a totally connected state. This leap takes place when the average link number exceeds 1. Erdös and Rényi studied the properties of these networks, one of which is particularly interesting since it states that the probability of a node having k links follows the Poisson distribution P(k) = e −λ λk/k!, where λ is a function of the number of links that mark the phase transition.

Any network generated by the Erdös–Renyi model with the same dimensions of REMOISSS, however will show completely different appearing, despite having the same number of nodes and links, this network will have a totally distinct structure than to the one exhibited by REMOISSS: it will look like some kind of chaotic mesh of links randomly distributed, resulting in a totally illegible graph.

The Erdös–Renyi model is a preliminary mathematical approach to network dynamics that is, nevertheless, non-existent in real-world scenarios. Hence, models that more closely resemble the networks of interest have been developed. An alternative model that considers that relations are not established by merely random mechanisms was proposed by Anatol Rapoport (Solomonoff and Rapoport 1951). This model considers that bias factors such as homophilia—the tendency to establish relationships with people with common sociodemographic characteristics (ethnicity, educational level, etcetera)-or transitivity (or triadic closure)—when two unrelated persons both know a third person they will tend to establish a relationship between them—also intervene in the process. This model is thus called random-biased networks model.

A third model uses structures known as small-world networks. The name refers to the familiar situation where one meets a seemingly complete stranger that, after some conversation, turns out to be an acquaintance of someone we know and usually makes us think, “it’s a small world.” Miligram (1967) is credited with the first experiment to verify how far apart two persons in the world are (actually, the relational distance between two persons in the United States); his findings indicated an average distance of only 6° of separation. Watts (2006) and Strogatz proposed a model based in this short-distance interconnection property, seeking equilibrium between a perfectly ordered network and a completely disordered one. The formation dynamics of these networks is the following: the starting point is a perfectly ordered network that, for example, has a circular shape and where every node is connected to the four nearest nodes. A fundamental characteristic of these networks—and the model’s namesake—is the fact that any two nodes in the network are at the same distance of the rest of the nodes.

A model that resembles reality more closely was developed by Barabási and Albert (1999), using the logic involved in Robert Merton’s Mathew effect (1968). This effect refers to a passage in the Gospel of Mathew that reads: “For unto every one that hath shall be given, and he shall have abundance, but from him that hath not shall be taken even that which he hath.” This model incorporates a preferential attachment property: the more connected a node is, the greater the probability it has of increasing its connections. The probability P(k), where k is the degree of the node, follows a power law: P(k) = k − γ, where γ depends on the specific type of network. The result is a network that exhibits a fractal structure where some nodes are highly connected, despite the fact that the degree of connections of almost every node is quite low. These networks are usually called scale-free networks and follow a power law similar to those of Zipf in Linguistics or Pareto in Economics. In contrast to what happens in the models of Erdös–Rényi, Rapoport, and Watts–Strogatz that start-out with a fixed number of nodes that randomly connect (Erdös–Rényi), that randomly connect with a bias (Rapoport) or that reconnect randomly (Watts–Strogatz) without changing the original number of nodes, the Barabási–Albert model is closer to reality because it proposes open networks that add new nodes. In this way, the system does not have a set number of nodes but a growing one and, thus, the network is in constant expansion. This type of network has been useful in modeling, for instance, the World Wide Web, which grows exponentially. Another characteristic of this model that sets it apart from those of Erdös–Rényi and Watts–Strogatz is that the connection between two nodes is not equiprobable but, instead, exhibits a preferential connectivity that more closely resembles real networks and that it is different from the transitivity or homophilia biases (Rapoport): nodes with more connections are more attractive to nodes seeking new connections.

Any network generated by the free-scale model with the same dimensions of REMOISSS also will show completely different appearance, despite having the same number of nodes and links, this network will have a totally distinct structure than to the one exhibited by REMOISSS or any Erdös–Rényi’s network: it will look like some kind of highly (but not totally) centered mesh of links.

Yet another model is that of Jon Kleinberg (2000), who proposed a coefficient to describe connectivity at different scales. The resulting network structures depend on the characteristics of the social structure, particularly one that could be considered a parameter for homophilia (Watts 2006). Lastly, Watts et al. (2002) point to the need to include multiple and independent social dimensions to which people belong as social identity characteristics. This multiple membership is a key to understanding the establishment of new connections: the homophilia bias is multiple because so is social identity.

The Evolution of RAISSS

In what follows, we analyze the evolution of RAISSS in further detail, since its origin is in action research, from 1995 to 1998, to date. How did the three original CSOs become related to each other and to the Italian experts and the funding agencies? How did the network evolve to the structure shown in Fig. 2? When analyzing CSO networks, it is remarkable to find that even when we are using the actual CSOs as nodes, networks are built by relationships among people; Joaquín del Bosque, Hogar Integral de la Juventud founder, knew Father Zubillaga, founder of Cejuv, because they both knew Father Alejandro García Durán, better known as Father Chinchachoma, founder of Hogares Providencia (Providence Homes, a CSO that pioneered work with children in street situations). In this first moment, triadic closure is evident: Father García Durán knew Joaquín and Father Zubilaga, and they ended up getting to know each other. It is also clear that there was a homophilia bias on several dimensions that considerably increased the chances for the three of them to meet: they shared, for example, that they all lived in Mexico City, had a strong relationship with the Catholic Church, and worked in social promotion, especially with youth in marginalized neighborhoods. The author of this paper, founder of Cultura Joven, met Father Zubillaga when the latter was named advisor to the Pastoral Juvenil de Cuernavaca (Cuernavaca Youth Ministry), where the author collaborated; the author met Joaquín del Bosque through Zubillaga. Once again, we find triadic closure and a homophilia bias in several dimensions (for example, a strong connection to the Catholic Church and social promotion with an emphasis on youth in marginalized neighborhoods).

Fig. 2
figure 2

REMOISSS CSO relation network

Full size image

Through Father Zubilaga, Roberto Merlo, a member of Gruppo Abele in Italy, begun providing advisory for Cejuv and Cultura Joven, while Efrem Milanece, of Piccola Comunità, collaborated in the same way with Hogar Integral, both in 1988. In 1994, Cejuv and Cultura Joven met Efrem and the Hogar Integral team met Roberto. Once again, these are instances of triadic closure and homophilia biases. What is most interesting is that, even when Efrem and Roberto are both Italian and work in the field of drug dependency they only met in Mexico in 1994. Homophilia and triadic closure are not, in and of themselves, definite causes for the establishment of a link between two nodes; instead, this process, the various dimensions of homophilia and sharing a common node (a condition for triadic closure) are biases—situations or factors that increase the probability for something to happen in a given way—and constitute random processes that are not, strictly speaking, deterministic.

The action research project that these actors put together was presented to the European Economic Community. When this entity called for a co-funder for 1996, Father Zubilaga—then Director of Cáritas Arquidiócesis de México (Caritas, Mexico Archdiocese)—requested the support of Martín Salm, Director of DCV. They met when Cáritas Arquidiócesis de México was supported by DCV in wake of the 1985 earthquake. The homophilia bias is clear: belonging to the Cáritas network increases the probability for two Cáritas representatives to meet each other.

Several CSOs (Centro de Asesoría y Promocion Juvenil, Fundación “Ama la Vida”, Reintegra) that participated in the formative processes that took place between 1995 and 1998 became mutually acquainted because they all knew Father Manuel, Joaquín or both, and Efrem and/or Roberto in addition to be related to intervention in social suffering situations in Mexico; the homophilia and triadic closure biases were key in the emergence and growth of REMOISSS and later RAISSS. When the action research project concluded, DCV started to connect CSOs in other countries (first Guatemala and El Salvador; Costa Rica, Panamá and Nicaragua later; and finally Honduras) through training processes developed by the Mexican team and, simultaneously, in Colombia and Chile through the processes developed by Efrem Milanese.

Brazil was added to the process through Efrem’s relationship with the Director of Lua Nova. Several CSO directors were, at the same time, members of the Red Iberoamericana de ONG que trabajan en drogodependencias (Iberoamerican Network of NGOs Working with Drug Dependency, RIOD); using these relationships, CSOs from Argentina and Uruguay were integrated (a process enhanced by the Uruguayan government’s request for advisory on the social inclusion model to its Colombian counterpart). DCV encouraged this collaboration process through regional programs that included formation, advisory, and implementation of pilot experiences. In each country, CSOs joined (and also left) the national networks. In Machín et al. (2010), we present an alternative account of the process by which CSOs became part of these experiences, based on the role of the projects developed with DCV’s support.

A very important reason for CSOs to become integrated into RAISSS is related to the Mathew effect: CSOs want to belong to RAISSS because many others are already members. But, why did these CSOs, in turn, choose to join when the network of networks was not too big? Without doubt, it is because belonging to the network allows CSOs to access a complex system of human resources formation and training in the ECO2 model (which has actually become an internationally renowned intervention model). Furthermore, through ECO2, CSOs systematize their practices using a common theoretical and methodological framework that includes standardized information management tools for documentation, administration, synthesis, and analysis (Milanese et al. 2001; Milanese 2009). This was crucial for CASA, CEJUV, Reintegra, and Machincuepa Circo Social (Machincuepa Social Circus), all of which use ECO2, to be the six Mexican CSOs selected among 56 CSOs and corporate community programs as the ones that used the best practices for crime prevention by USAID’s Citizen Coexistence Program and by the Centro Nacional de Prevención del Delito y Participación Ciudadana (National Center for Crime Prevention and Citizen Participation) (CNPDyPC-USAID 2012).

In addition to the formative processes and systematization possibilities offered by RAISSS, membership gives CSOs access to specialized technical advisory and, in some instances, the possibility of getting DCV funding. Another incentive for CSOs is to become collectively organized to enhance their public policy advocacy capabilities: RAISSS has managed to translate practice into science by systematizing different social intervention schemes and influencing evidence-based public policy; the network has ‘bridged the gap between research and practice’ (Wandersman et al. 2008). Some examples are the successful experience of Fundación Procrear and Corporación Viviendo in Colombia, organizations that developed a social inclusion program for psychoactive substance users as part of the country’s nation-wide campaign to reduce psychoactive substance usage (Fergusson and Góngora 2007); the ACIA (Acompañamiento Comunitario Integral de Adolescentes or Integral Teenager Community Support) model, developed by Mexico City’s Dirección Ejecutiva de Ttratamiento a Menores (Executive Unit for the Treatment of Minors) (actually named General Directorate of Treatment for Adolescent in the Federal District (Dirección General de Tratamiento para Adolescentes del Distrito Federal)) and REMOISSS CSOs to provide treatment to children in conflict with the Law (SIJA 2009); or the project implemented by Uruguay’s Junta Nacional de Drogas (National Drug Board) in 2010, where community-based social inclusion methods were put in place to tackle problematic drug use, to mention a few.

In this way, we can see that the evolution of RAISSS has clearly followed the mechanisms described by Barabási–Albert, Kleinberg & Watts and Dodds & Newman: the network does not exhibit a fixed node number (as the models of Erdös–Rényi, Rapoport o Watts–Strogatz), rather, the number of nodes has been increasing through a process that is partly random but is also determined by selective aggregation processes. These processes exhibit high levels of transitivity, triadic closure, and biases related to the prevailing social structure and homophilia on multiple dimensions of the social identities of CSO representatives (for example, being mostly Latin American, belonging to CSOs that work on drug dependency and other social suffering situations, in many cases connected to the Cáritas network and the Catholic Church). This network formation process is fairly similar to the one described by Christakis and Fowler (2010) in the development of couples.

Conclusion

We have used an approach based on several dynamic models of network formation to analyze the evolution of RAISSS. However, we may also study internal network dynamics from other perspectives, particularly those that employ agent-based models in combination with sociologic theories such as Mellucci’s (1996). These models consider systems to be sets of indistinguishable elements that interact with their immediate neighbors and generate global processes and dynamics from this local interaction (Axelrod 2004).

In our analysis of the evolution of RAISSS and some of the most important models that have been developed to describe social network dynamics, we can see that the reticular shape of RAISSS is isomorphic; these preliminary approximations were the obvious choice to describe its dynamics. Using RAISSS as a case study, we can infer that the models of Barabási-Albert, Kleinberg & Watts, and Dodds & Newman provide us with clues to understand, albeit partially, the evolution of some CSO networks. Given the importance that these networks have for our Latin American societies, we are certain that any progress that broadens our knowledge about them will be welcome.