Modeling The Norm Life Cycle

1. Joshua Epstein argues that to have a candidate explanation for a phenomena, one must show that it is possible to grow the phenomena given the assumptions present in the argument. See Epstein 1999.

2. See e.g., Schelling 1978; Axelrod 1997; Arthur 1994a; Cederman 1997; 2003; Epstein and Axtell 1996; Lustick 2000.

3. Epstein and Axtell 1996, 4, emphasis in original.

4. At least those interested in being "scientific." See Wendt 1998.

5. Kollman, Miller, and Page 1997, 465. See also Arthur 1994b.

6. Arthur, Durlauf, and Lane 1997, 5.

9. This is not the only position in this debate—some constructivists are concerned about the overemphasis on structure in constructivist approaches. See Sending 2002, Checkel 1998.

10. Guzzini 2000, 150.

11. Checkel 2001, 561.

12. Guzzini 2000, 164.

13. Guzzini 2000; Checkel 2001.

14. The differences between methodological individualism and an intersubjective ontology may have been overstated. Kenneth Arrow claims that even in economic models (the most strict of the methodological individualist projects), "individual behavior is always mediated by social relations" (1994, 5). One change in this sentence would make it amenable to constructivists—the verb mediated would need to be changed to constituted. Granted, this is an important difference, but perhaps the perspectives are not as far apart as conventionally considered.

15. Checkel 2001, 559; Guzzini 2000, 169.

16. ABM can be employed in numerous ways. Some try and mimic actual social systems in their simulations—the agents would then represent actual states, organizations, or people. Others use the models in a much more abstract way, in the hopes of discerning fundamental features of social systems. The agents thus do not represent actual political actors, but are abstractions like the players in game theoretic models.

17. Moravcsik 1999a; 1999b; Young 1999.

18. Finnemore and Sikkink 1998; Risse, Ropp, and Sikkink 1999; Bianco and Bates 1990; Frohlich, Oppenheimer, and Young 1971.

19. This model was written in C + +, using Microsoft Visual Studio. The computer code for the model is available upon request from the author.

20. The average prediction is a way to aggregate individual actions into population outcomes. There are other ways (voting schemes) and I am currently investigating a number of them.

21. Any of these rules could be a norm. The rule-based behavior is modeled after Arthur's heuristic style of agent-based modeling (1994b).

22. Having a universe of seven rules and endowing the agents with only three of them was done to keep the model very simple. This can be easily altered to explore more complex situations.

23. In the case of a tie for the highest score, the agent randomly chooses one—she flips a coin.

24. I have experimented extensively with this variable and altering it had negligible effects on the results. In the simulations presented below, the interval ranges from ten to twenty rounds.

25. Societal punishment for breaking a norm—sanctions, see Axelrod 1997; Elster 1989.

26. Some coordination studies are moving toward constructivist analysis as the importance of shared "mental models" (intersubjective understandings) for picking among multiple equilibria is being recognized. See Richards 2001.

27. The noise added or subtracted from the average prediction is a random draw from a uniform distribution bounded by 0 and the specified maximum noise level. When the noise is reported as 5 percent, a number between 0 and 5 has been added or subtracted to the average prediction.

28. Keck and Sikkink 1998.

29. As this model was explicitly designed with international norms in mind, there is less emphasis on direct agent-agent interaction.

30. In a set of thirty simulation runs, the agents coalesced around rule 4 in twenty-six runs, they coalesced around rule 5 in two runs, and did not stabilize within one thousand rounds in two runs.

31. There may be synergy here. One reason an idea may be intrinsically attractive is the lack of social complexity surrounding the idea.

32. While any of the rules can rise to normative status, rule 4 (the natural norm) is the most prevalent because in some runs, the system reaches stability before the norm entrepreneur makes a suggestion in round 50. In a set of thirty runs, the agents coalesced around each rule, except for rule 6: rule 1 (four runs), rule 2 (two runs), rule 3 (four runs), rule 4 (eleven runs), rule 5 (seven runs), rule 7 (two runs).

33. Arthur 1994a.

34. David 1985.

35. Arthur 1994b.

36. Ikenberry and Kupchan 1990; Ikenberry 2001.

37. Ruggie 1993; Ikenberry 2001.

38. Finnemore 1996b.

39. Wapner 1996.

40. Risse, Ropp, and Sikkink 1999.

41. Klotz 1995a.

42. Price 1997.

43. Finnemore 1996b.

44. Numerous other sensitivity analyses were performed to assess the parameter sensitivity of the model along a number of dimensions. The reach analysis is the most relevant for the analysis of norm emergence and evolution in the ozone depletion and climate change negotiations.

45. Defining norms as 70 percent same rule-usage was arbitrary. However, I conducted similar analysis with norm thresholds at 80 percent and 90 percent and saw negligible qualitative difference. In cases of norms that last a significant number of rounds (10 + ), generally the entire population of agents adopts the norm.

46. Axelrod 1997.

47. Bernstein 2001; Finnemore and Sikkink 1998.

48. Another sensitivity analysis performed on this model explored the role of varied population sizes and varied number of available rules. The results are akin to those obtained by altering the noise. When the population size increases, while the number of available rules stays constant, the social complexity, in effect, is lowered. It is easier for the population to find the natural norm. When the number of rules is increased as well, the social complexity increases.

49. Pierson 2000.

50. I freely substitute normative political context and political context for structure and international political context.

51. The articles were culled from the Dow-Jones Interactive Database and Lexis-Nexis.

52. Finnemore 1996b, 22.

53. Dessler 1989, 462.

0 0

Post a comment