The Complexity of Conflict Management

Open Access
- Author:
- Pechenkina, Anna O
- Graduate Program:
- Political Science
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- July 29, 2013
- Committee Members:
- David Scott Bennett Jr., Dissertation Advisor/Co-Advisor
Douglas William Lemke, Committee Member
Philip A Schrodt, Committee Member
John Yen, Committee Member - Keywords:
- Mediation
reinforcement
agent-based model
conflict management
genetic algorithms
parameter space
third-party intervention. - Abstract:
- The literatures on conflict management and intervention have developed in isolation, rather than informing each other. This project initiates the bridging of the two lines of research by representing mediation and external support in the same agent-based computational model. Two questions motivate my dissertation. First, I explore the effect of external support on war duration by simulating the underlying bargaining process in the conflictual and peaceful interactions of a multi-player international system. The empirical research on external support in conflict has had inconsistent results. I follow previous research in assuming that third parties choose to reinforce either side when it can make a difference in the outcome, which means support is never given to those belligerents that can win regardless. Given that third parties in my model observe the behavior of supporters for the opposing side, I also derive that support is more likely to be given when one's protege's opponent receives support. This in turn implies that support on average makes wars longer because it is meant to alter the outcome of the conflict. In departure from previous research, my model also accounts for the presence of power mediators who may impose prohibitive costs on the belligerents and may make the provided support irrelevant. Finally, my model also yields that support may alter the calculus of belligerents during the conflict of interests stage, preventing the conflict from escalating to violence in the first place. Therefore, the effect of support on conflict duration is bimodal: preventing the onset and prolonging those armed conflicts that have started. Second, I focus on the effect of power mediation (i.e., a mediation style that exerts pressure on the belligerents) on peace duration. Power mediation is the most effective mediation style at achieving ceasefires. This finding may have important policy implications: the more pressure mediators put on the belligerents the more likely they are to precipitate peace. However, power mediation may also reduce the duration of ceasere that follows, which implies that third parties may sacrifice long-term peace stability for the sake of short-term benefits of an unstable ceasefire. I demonstrate that the actual effect of power mediation on ceasefire survival is more nuanced and should be analyzed in concert with the actions of support-providing third parties. First, if powerful mediators stop inflating the costs of fighting the conflict of interest and armed conflict may indeed recur in my model, as previous scholarship argues. However, this dynamic takes place in those cases that receive no external support. When reinforcement is large enough to eliminate uncertainty over who is going to win the war, my model yields that a peaceful transfer of resources to the dissatisfied belligerent will take place. In contrast, in those cases when reinforcements are only large enough to recreate conflicts of interests but do not remove uncertainty, an initially fragile ceasefire may become even more prone to conflict recurrence than previously argued. Therefore, the effect of power mediation on ceasefire duration depends heavily on the behavior of support providing third-parties. The flexibility of computer programming to accommodate a finite number of heterogeneous actors and my focus on individual-level decisions make agent-based modeling a suitable tool for hypotheses development in this project. I build a benchmark agent-based model of intra-war and post-war bargaining in the world without third parties in chapter 2 and an extended model with mediating and support-providing third parties in chapter 4. One of the disadvantages of computational models is that parameter spaces are often prohibitively large to generate and analyze exhaustively. This results in scholars often ignoring the boundary conditions for a model's parameters and non-systematically derived conclusions. I introduce a systematic way to interpret agent-based models by interfacing a genetic algorithm search program with simulation models, conceptualizing numerical output from a simulation model as a fitness function of the genetic algorithm. Genetic algorithms do not need to create an exhaustive parameter space to find global solutions, and therefore users are able to explore their models more quickly, yet systematically. The software program SimGA is presented in chapter 3 and verified by recreating the logic underlying the benchmark model of chapter 2. I also use this program to analyze the extended model in chapter 4. Finally, I use observational data to test empirically some of the extended model's implications in chapter 5. I test for effect of external support on war duration against a sample of civil wars and find support for my model's implications. With respect to external support shaping peace duration, in the sample of interstate ceasefires, external support precipitates conflict when it recreates uncertainty and stabilizes peace when it eliminates uncertainty.