« Chaos, Complexity, Curriculum, and Culture: A Conversation | Main | Adaptive agent Modeling in a Policy Context »

Complexity Theory & Political Science

After searching academic journals for articles combining complexity theory and political science, I propose the following five as randomly representative.

• Geyer, Robert R. “Globalization, Complexity, and the Future of Scandinavian Exceptionalism.? Governance (London), vol. 16, no. 4, pp 559-576, October 2003

• Ma, Shu-Yun. “Political Science at the Edge of Chaos? The Paradigmatic Implications of Historical Institutionalism.? International Political Science Review , vol. 28, no. 1, pp. 57-78, January 2007.

• Hoffmann, Matthew J. and John Riley Jr. “The Science of Political Science: Linearity or Complexity in Designing Social Inquiry.? New Political Science, vol. 24, no. 2, pp. 303-320, June 2002.

• Feder, Stanley A. “Forecasting for Policy Making in the Post-Cold War Period.? Annual Review of Political Science, vol. 5, pp. 111-125, 2002.

• Brunk, Gregory G. “Why Do Societies Collapse? A Theory Based on Self-Organized Criticality.? Journal of Theoretical Politics, vol. 14, no. 2, pp. 195-230, April 2002.

(Note: for a quick read, skip to the “Summary and Conclusions? section at the end).

Each of these articles argues that complexity theory is necessary to political science. This is because traditional linear thinking has proven inadequate to explain complex systems, among which are human political organizations. Typically, linear models are deterministic and reductionist, attempting to order the world according to a single set of universal principles. Such were the prevalent visions of social order in the Cold War, communism and capitalism, and “it was the pursuit of these extreme forms of order that brought about extreme forms of human suffering? (Geyer). Unexpectedly, to “devotees of the linear model,? Scandinavian countries have thrived by incorporating mixed elements of both these supposedly inconsistent models, remaining flexible in their response to the demands of a global economy, and avoiding “neat, orderly, and universalistic conclusions? (Geyer).

Moreover, analysis of past national security and foreign policy decisions of the United States government indicate that they have suffered from the linear nature of single-outcome forecasting and “a prejudice toward continuity of previous trends? (Feder). The desire for certainty in the complex system of international relations has not, in effect, reduced uncertainty; but has “only increased the margins of surprise? (Feder). This has been counterproductive, because the basic value of a forecast in the context of foreign policy is not that it accurately predicts the future, but that it can “keep us from being surprised? (Feder). To do this, it must provide a survey of several possible outcomes, together with leading indicators for each, rather than prediction of a single outcome. Non-linear models show promise in being able to do this. “As the inputs are varied in plausible ways, the models indicate which outcomes are possible and which are impossible? (Feder). It is a case of becoming familiar with the properties of the system rather than focusing on finding the one input which will produce the desired outcome. By using models to “examine ‘what if’ scenarios, one can develop a sense [i.e. intuition] of which changes in the political environment will have a significant effect on a particular issue? (Feder). In the early 1980’s the planners at Royal Dutch/Shell Oil considered the possibility that radical changes in the Soviet Union could cause the price of oil to fall. “They sought evidence that such an event [i.e. societal collapse] was possible and found it. Shell’s insight came from ‘asking the right question. From having to consider more than one scenario.’ (Schwartz)? (Feder).

On a more theoretical level, complexity theory includes several concepts which better explain political behavior than strictly linear models. Two of these concepts are path dependence and the economics of increasing returns. At bottom, they are very similar concepts, for they both posit that political systems are like autocatalytic sets: their “outcomes at critical junctures trigger feedback mechanisms that reinforce the recurrence of a particular pattern into the future? and “once a social process has started, it will produce its own law of inertia . . “ (Ma). They thus explain things like political momentum; why success feeds upon success; why, for example, American presidential candidates are so eager to start off well in the early state primaries. They controvert the linear notion that the outcomes of a system are always proportional to its inputs, for they consider the catalytic properties of the system, rather than just its inputs. Thus, given a critical juncture of a human system, even a trivial cause can have a large effect (the “butterfly? effect). This idea is central to all five articles. Moreover, the idea is problematic for political scientists doing traditional, linear, cause-and-effect analysis. For when all inputs, both small and large, may prove equally efficacious in a complex system, how do you decide which ones are important? Only by mastering all the complex relations which may exist among different nodes of a human network, could you be able to predict outcomes. It is this realization which requires the acceptance of uncertainty and the rejection of reductionist determinism as unrealistic.

Another complexity concept which may prove applicable to political phenomena is self-organizing criticality (SOC). This is introduced to help answer the question of why human societies have collapsed throughout history. Noting that traditional political science has failed to discover a linear law to explain the phenomenon, Brunk imports SOC from the physical sciences (where it has many applications). One appropriate SOC metaphor is Per Bak’s power-law sandpile. As each grain of sand falls on the pile, the pile’s complexity increases, to the point at which it becomes hypersensitive to even the smallest of shocks. At this point, dropping another grain on the pile results in a partial or complete collapse of the pile. Thus do human societies, as they always tend to become more complex by adding nodes and dependency relationships, tend toward the point whereby a small shock can result in collapse of the whole. The size of the reaction is “not caused by the size of outside shocks, but by how shocks are transmitted within a system as complexity cascades? (Brunk). Just such reactions were the First World War (supposedly triggered by the assassination of a single man, Austria’s Archduke Ferdinand), the 1929 stock-market crash, the Great Chicago Fire, etc. It is because societies have become adept at “dampening? their sensitivity to complexity cascades, by such techniques as FDIC insurance of American banks, river levies, cartel price agreements, etc., that civilization has been able to advance; but these efforts are often too weak or ill-designed to hold back the onslaught of chaos. “Wars, like forest fires, are SOC processes. . . . Unless the fundamental rules that govern the behavior of such a system change, it s only a matter of time before a catastrophic war destroys any given nation-state? (Brunk). However, the author bails-out of the darkest fatalism by adding that “there is not enough empirical data on wars to directly examine these patterns.?

Granting the general superiority of non-linear models over linear models in the context of political science, questions remain as to how to advance our understanding of political systems using this new paradigm. “If complexity theory is to be more than a metaphor or a critique of the Newtonian method for political scientists, then it must facilitate the articulation of a research program, as succinct and as accessible as the traditional scientific approach? (Hoffmann). One suggestion is that political scientists begin to gauge the probability of events, “such as the likelihood of war, by systematically looking at the effects of initial conditions and small changes. . . . The analytic ‘trick’ is to identify points in a system whereby disturbances can have exponential effects on the direction of the system? (Hoffmann). In this context, Hoffmann and Riley review the work of two complexity theory pioneers, Robert Jervis and Robert Axelrod. Unfortunately, they find that Jervis “lends little advice on how to incorporate system thinking into [political scientists’] work.? They compare his thinking to a “conceptual jailbreak,? implying its main effect is to separate us from the old worldview, without providing the means of establishing a new program. Their review of Axelrod’s agent-based models is in a similar vein, indicating that they are “artificial and, by design, simple.? While most of them are “exceptionally interesting and powerful illustrations of fundamental processes, they do not analyze real political phenomena.? The authors are far from despair over this situation. They simply suggest that an empirical gap is still to be bridged between computer simulation and real phenomena.

For Feder, all complex system scenarios must be valued, without assigning degrees of probability to each. Degrees of probability tend to focus the mind on the single outcome with the greatest degree of probability; whereas in a complex system, each possible outcome must be considered. In addition, the gap between models and real phenomena is for him reflected in the necessity of asking the right questions: “analytic methods alone will not guarantee that policy makers and academics will not be surprised by political events. Preventing surprise depends on asking the right questions…? (Feder). So, non-linear models, by themselves, are no substitute for real-world experience. They must be combined with that experience by asking the right questions of them.

Brunk’s argument rests mainly on analogy, on the better fit of the SOC model to societal dynamics than linear models; but he also proposes a change in methodology, decrying the narrow specialization of traditional political scientists. Rather, “a holistic approach is sometimes needed, because some processes only emerge at the system level. . . they cannot be discovered by examining individual events, no matter how intently they are studied. . . The generic, but non-deterministic stochastic pattern of a SOC system always repeats in a general way, but never repeats in exactly the same way. In other words, while its general contours can be described, it is not deterministic in its individual events? (Brunk).

Summary & Conclusions

My goal in reading these five articles was to get a sense of how complexity theory has impacted political science. The impact is significant, on levels practical, theoretical and methodological. It seems that political science was ripe for this kind of impact, due to its historical inability to both establish itself as a true science and to resolve some of its fundamental problems. First, its impact can be felt on the practical level, where linear models have proven to have disastrous political consequences. Such have been the rigid, linear models of capitalism and communism, with their failure to predict the success of systems intermediate to these two types (the Scandinavian societies) and to predict such events as the sudden collapse of the Soviet Union. Second, it has had effects at the theoretical level, where ideas such as path dependence, the economics of increasing returns, and self-organizing criticality, each add a level of understanding to the study of complex systems which seems outside the scope of linear thinking. For example, it would seem difficult to understand the dynamics of political movements without recourse to the first two of these ideas; and it is tempting to explain the phenomena of societal collapse in terms of self-organized criticality, although that idea needs more rigorous development to be persuasive. Third, it has had an impact on political science methodology, in the recognition that understanding human organization requires more attention to the properties of system process and less to cause-and-effect analysis. Thus, 1) familiarity with a system’s points of bifurcation, 2) familiarity with the range of its possible outcomes, and 3) identification of its behavioral patterns, rather than the details of any one scenario, are all touted as techniques appropriate to the non-linear approach. In this respect, it seems that a highly inductive attitude is required by the new paradigm, even to the point of favoring unconscious intuition over conscious ratiocination. Whether this new approach satisfies Hoffmann and Riley’s requirement that the new paradigm “facilitate the articulation of a research program, as succinct and as accessible as the traditional scientific approach? is an open question.

An interesting divergence among the five articles is Geyer and Brunk’s differing perspectives on war. Whereas Geyer indicates that non-linear thinking will make for more flexibility of thought among human societies, and thus less conflict, Brunk treats war as a natural consequence (a “complexity cascade?) of self-organizing systems. In his view, it is only by societies’ ability to dampen down their tendency to organize to points of criticality that war may be avoided, and he is noncommittal as to how they can do that. Moreover, he indicates that many such attempts to dampen criticality are ill-designed and ineffective, even leading to an increase of criticality. So his is a decidedly more pessimistic view than Geyer’s.