Using Networks to explore Narratives

How Network Analysis can help us to study the emergence of Forced Displacement as a National Trauma in Colombia?

Social network data refers to measurements of relations between actors. If we consider narratives as a sequence of actions by which actors are linked to one another, we can suggest that narrative texts are a source of network data. By applying network analysis to newspaper stories about forced displacement in Colombia, I will show the usefulness of this approach by focusing on sociometric notation and graph representation. Those interested in the emergence of forced displacement as a national trauma in Colombia will find this as an useful blog post.

The Networks Perspective

From a network perspective, social phenomena can be expressed as patterns of relations between social units that contribute to the study of society and to define coherent lines of action. The key assumption is that social life is relational and therefore relational information is necessary to understand human behavior. This perspective has been useful to study multiple social phenomena such as multilateral negotiations, processes of conflict resolution, disasters recovery and society composition and coordination.

Bruce Money (1998) used a network perspective to develop a model of international multilateral negotiations in which he examined how coalitions form and roles emerge among negotiators networks positing the effect of social network activity on negotiation outcomes. Susan Allen (2008) applied this perspective in the study of conflict resolution processes and she found a dual impact of social networks since these can serve both, to support meaningful participation in conflict resolution processes in the case of inclusive networks, or to reinforce conflicts, institutionalize exploitation and exacerbate inequalities in the case of exclusive networks.

By focusing in the study of natural disasters, Varda, Forgette, Banks and Contractor (2009) suggest that the reduction in post-disaster social support is associated with the weakening of social networks since victims’ networks are disrupted and the formation of social ties is driven by immediate needs. In such a situation, authors found that individuals embedded in networks with higher-density, gender diversity and higher proportion of men, kin and younger actors, obtained more suppport than individuals embedded in networks lacking these carachteristics. In fact, in difficult situations, ties seem to be decisive for the perception of social support and contribute to the relief of physical or material needs as well as health or emotional needs (Gallicchio et al., 2007).

Another example of the application of this perspective is that of Enemark, McCubbinsb and Wellerc (2013) who examined how network knowledge affects coordination. They found that a more complete view of the network leads to faster coordination and that the magnitude of such effect depends on network structure. Therefore, they suggest that changing what actors know about a network can improve outcomes without having to add connections that may impede overall performance. Other interesting application of network perspective is that of Stephen Vaisey and Omar Lizardo (2010) who found that networks composition is affected by internalized cultural worldviews and suggest that the dynamic link between culture and social structure needs to be reconsidered.

What these studies have in common is that their authors collected network data by observing, interviewing or questioning actors about their ties to other actors. However, social network researchers do not always have direct contact with the actors involved in the network of their interest or they require relational data from so many actors that it would not be possible to take measurements on all of them. In these cases, network researchers have measured ties by examining interactions recorded in journal articles (Burt, 1987, 1979; Doreian & Fararo, 1985; White & McCann, 1988), newspapers (Burt, 1975) and biographical records (Rosenthal, Fingrutd, Either, Karant & McDonald; 1985). This approach has been recently applied to the analysis of narrative texts (Tilly, 1999; DiMaggio & Mullen, 1993; Mohr & Guerra-Pearson; 1998; Franzosi, De Fazio & Vicari, 2012), taking narrative as a sequence of actions and events arranged in chronological order.

Using Networks to Explore Narratives

By applying network analysis to narrative texts, we can identify not only actors and the sequences of actions, but also the structural properties of narratives that can be useful to explain how the actors make sense about social reality. An application of network analysis to newspaper articles will illustrate the usefulness of this approach. For this I have selected 10 articles published in El Tiempo about forced displacement in Colombia, one of the main causes of massive suffering in this country.

From each article I identified the triplets of actors, actions and the objects of actions constituting the narrative, where each action is considered as a transaction taken as a relational tie. Then, to construct the network graphs, I organized two types of triplets: one recording the actions against the victims and the other recording the actions against the victimizers.  For each triplet I constructed a n x n square weighted matrix D in which the relational ties between D_i and D_j measure the frequency of the directed relations (with values form 0 to N). Finally, I imported the matrices into the software UCINET and produced two two-mode network graphs with Netdraw. 

Figure 1: Network of violence against the victims 

The first graph is given in figure 1 and corresponds to the structure of the narrative about the actions against the victims. It is composed by 7 nodes (where victims are labeled with blue squares and victimizers with red circles) and 5 directed ties between the pairs of nodes (from victimizers to victims). The actions against victims are related with sexual violence, land expropriation, murder, extortion, stealing, kidnapping and recruitment. These victims are individuals living in rural (n=14) and urban (n=5) areas as well as indigenous women (n=5). The victimizers are members of criminal bands (n=11), paramilitary groups (n=4), palm oil producer agencies (n=2) and the Army (n=1).

According to the data, the criminal bands have caused suffering on people living in rural and urban areas while paramilitary groups and palm oil producers have concentrated the violent activities against individuals living in rural areas. There is also a dyad apart from the main structure, which indicates the acts of violence inflicted by the army to the indigenous women. In addition, the strength of the tie between the criminal bands and their victims indicates that these bands are considered the main responsible of suffering. 

Figure 2: Network of actions against victimizers  

The second graph is given in figure 2 and corresponds to the narrative about the actions against victimizers. It is composed by 8 nodes (where victimizers are labeled with blue squares and actors who act against them with red circles) and 7 directed ties between the pairs of nodes. The actions against victimizers, which are related with accusation, arrest and sentencing, are made by government entities (n=12), victims (n=1) and the Indigenous National Organization (n=4).

These are directed to paramilitary groups (n=7), criminal bands (n=5), palm oil producer agencies (n=2), the army (n=1) and the FARC thought this guerrilla group does not appear as victimizers in the network about the actions of violence against the victims.  The strength of the ties shows that the government has focused his attention on paramilitary groups and criminal bands. Notice that victims have just acted against paramilitary group.

Measurements of centrality and power, key-players and clusters would provide deep information about the structural properties of these narratives. However, the purpose of this blog post was to show the applicability of network approach to the analysis of narratives by using the sociometric notation and suggests a deep exploration of the structural properties of narrative texts.