Monday, February 8, 2016

Modeling the Role of Networks and Individual Differences in Inter-Group Violence

There is significant heterogeneity within and between populations in their propensity to engage in conflict. Most research has neglected the role of within-group effects in social networks in contributing to between-group violence and focused instead on the precursors and consequences of violence, or on the role of between-group ties. Here, we explore the role of individual variation and of network structure within a population in promoting and inhibiting group violence towards other populations. Motivated by ethnographic observations of collective behavior in a small-scale society, we describe a model with differentiated roles for individuals embedded within friendship networks. Using a simple model based on voting-like dynamics, we explore several strategies for influencing group-level behavior. When we consider changing population level attitude changes and introducing control nodes separately, we find that a particularly effective control strategy relies on exploiting network degree. We also suggest refinements to our model such as tracking fine-grained information spread dynamics that can lead to further enrichment in using evolutionary game theory models for sociological phenomena.

…The process of attitudes spreading across society bears some resemblance to voter models [], though the rules are more complicated. In our multi-scale model, leaders recruit locally, followed by small friendship-based bursts for collective violence. The overall social attitude toward participation is shaped by the success or non-success of these local events. A particularly interesting aspect of the model is that this recruiting structure appears to be amenable to society-wide interventions that affect risk-seeking attitudes as well as fine-grained control []. While this framework is motivated by observations of raid recruitment in small-scale populations [], it would be interesting to relax the assumption of local recruitment and consider how different information flow rules lead to coordination—e.g. when control nodes (saints/devils) can effectively be “bypassed”. We suggest that this model framework and the control architectures discussed in this context may be further investigated in the context of different spreading processes, such as epidemiological models [].

Our results may demonstrate the mechanism by which cultures transition from intergroup violence to peace. Two well-documented small-scale societies illustrate this process. Until recently, the Enga of New Guinea and Waorani of Ecuador both had intense warfare. After several decades of increasingly intense in inter-clan warfare, the Enga have recently transitioned from a period of chronic warfare to comparatively peaceful relations due to several factors that parallel our model. First, the death of many of the most violent individuals known locally as “Rambos” occurred [] and the population of youth who functioned as warriors aged out []. Second, and more importantly, communities’ attitudes towards violence have changed towards to become more pacific, in part because of outside influences and recognition of the decreasing or marginal benefits from warfare in this case []. Among the Waorani, whose intense cycles of violent revenge threatened the group with extinction, the process has been similar. Lethal conflicts have been nearly extinguished through the adoption of cultural values of peace [] and prominent individuals forgoing previous norms that would have called for them to engage in violent revenge []. This is similar to the processes in our model where decreasing α (tuning population-wide attitudes) and having prominent (high-degree) individuals become “saints” provide a powerful net effect towards the adoption of less violent action. Our model is illustrative of the dynamics by which these processes occur, which, at their core, exploit inter-personal influence and social learning, and applies to decentralized contexts including those of sectarian, ethnic, or religious conflicts in which there are no state actors imposing top-down controls. In reality, however, collective non-violence is not simply the lack of coordinated violence—the path to peace may itself be driven by its own leaders and follow its own dynamics, which suggests an interesting extension of adding a competing “pacification process”. And, insofar as the physical removal (death) of particularly violent individuals (and hence also their connections) was observed to change the social values towards non-violence, a further enhancement to the model could include time- or process-dependent effects such as death. Our framework may also apply to state-level conflicts that depend on the mobilization of participant support, and can be integrated into existing frameworks that explore the interaction of participants and states [].

Of course, the model is only an approximation to reality. One aspect that it leaves out is preferential learning, such as one based more closely on social ties. This may be expected to be particularly relevant in larger societies where each person only sees a small minority of all possible people. Such a scenario can give rise to two scales of learning: a general socially acceptable baseline and much smaller local clusters of more or less aggressive behavior. Relatedly, this model does not address the social network evolving in time and can thus be viewed as a relatively short time-scale model. Expanding the model to more realistically treat large-scale societies or groups as they evolve over time is a fruitful area of further research…

Below:  Network of connections and designated raid leaders

Full article at:

Tobias Preis, Editor
1Department of Physics, Harvard University, Cambridge, Massachusetts, United States of America
2Department of Mathematics, Yale University, New Haven, Connecticut, United States of America
3Department of Human Evolutionary Biology, Harvard University, Cambridge, Massachusetts, United States of America
4Program for Evolutionary Dynamics, Harvard University, Cambridge, Massachusetts, United States of America
5Department of Sociology, Yale University, New Haven, Connecticut, United States of America
6Department of Ecology and Evolutionary Biology, Yale University, New Haven, Connecticut, United States of America
7Yale Institute for Network Science, Yale University, New Haven, Connecticut, United States of America
University of Warwick, UNITED KINGDOM
#Contributed equally.
Competing Interests: The authors have declared that no competing interests exist.
Conceived and designed the experiments: AI AH LG NAC. Performed the experiments: AH. Analyzed the data: AI AH LG NAC. Wrote the paper: AI AH LG NAC.

No comments:

Post a Comment