Every now and then I see arguments about Iraq, Afghanistan, and counterinsurgency flare up in various forums and blogs, now mostly tied to counterinsurgency and preparation for future warfare.
I can’t help but darkly wonder if the immense amount of time and energy spent fighting over the historical memory of Iraq and Afghanistan ( a process that will not end for a while) may not matter as much as it may seem right now.
Let’s pick just Iraq. Assume that Iraq is politically meaningful in some way for a substantial amount of time, and if policymakers can correctly recognize an event as suitably Iraq-like their decision might tilt in one direction or the other about American involvement. Assuming that Iraq-like events could repeat themselves, the other question is whether politicians could correctly recognize an event as “Iraq-like.” I do not think it would so clear.
The memory and time horizon during crisis situations are both very low. And immediate ranges of perceived options specific to the conflict in question — which having its own unique political dynamics would always be different enough to make each crisis seem unique — are considered.
I frankly am coming to doubt that the policymakers faced with the “next Iraq” — assuming there is another conflict that even vaguely fits the general contours of Iraq — will really take into consideration much of what everyone debating Iraq’s history eventually converges on. A shadow cast by the wars may hang, but it is only a script that filters a policymaker’s evaluation of the information they have available during a crisis situation. The script is meaningful, but it can only shade the specific conditions that could lead to conflict. And those conditions and choices are almost always perceived by each administration as unique.
As I’ve gotten deeper into the semester, I’ve often had a thought experiment about what might happen if we were suddenly gifted with the unerring ability to create a set of the Iraq War’s most important variables, actors, decision rules, and world topology. In this thought experiment, the simulation modeler is God, and with omniscient knowledge produces an mini-Iraq that (in another assumption that would never be true) that is isomorphic to the Iraq in the real world. Each model run is as long as the United States stayed in the conflict, from the invasion to the main withdrawal following the Status of Forces Agreement dispute.
Every model run, the model world is generated anew and the previous state is cleared off. The modeler-as-God could generates random starting positions from a specially chosen probability distribution that the all-knowing modeler is able to unerringly tie to the actual range of likely starting positions and values at the beginning of the war.
Because the divine modeler is omniscient, he knows the parameter settings needed to make basins of attraction that would reproduce a group of outcomes similar to the one we all lived through. However, the exact outcome would be different each time. The shape and expression of each outcome would not be exactly the same even if the direction of history is qualitatively clear. Why?
The global state of the model depends on the decisions of each actor, and each actor is assumed to make decisions based on the decisions of others. Hence the combinations of multi-actor behavior — even when the master settings are turned to point in one direction — are probabilistic. Some outcomes, though not terribly likely in the thought experiment, might shock those who followed the war closely.
Some important questions might come up for the modeler. Was the Surge the primary defeat mechanism? Was it the Sunni-Shiia civil war? Was it both? The modeler might investigate this by using the time to settlement, and the nature of the settlement as outputs to be monitored — and varying the important overall settings to experiment on how different conditions might effect the shape of how—or if—a settlement occurs by the time the model is programmed to stop.
For the record, I have no idea. The thought experiment is already unrealistic enough in its use of the narrative device of a infallible Divine Modeler with the ability to fashion an isomorphism from a complex Arab state embedded in a regional and global system to an unerringly crafted simulation model. And I’ve specified no rules, parameter settings, or even loosely thought about the dimension and topology of the mini-Iraq.
If one wanted to make the model even more unrealistic, they could make each model run’s initial conditions influenced by the end conditions of the previous model. Instead of an Iraq that is randomly generated with vaguely similar starting positions at each initialization, they would probably get an Iraq that evolves over a period of time….perhaps into something unrecognizable given the world is created anew after each model run ends.
All computational simulations based on bottom-up principles have interesting and varied out trajectories. From the God’s-eye view it is often possible to see specific loops or cycles with some settings and combinations and fairly random behavior in others. However, the range of qualitative behaviors all tend to be fairly limited, even if the quantitative data they produce is not.
Of course, the real Iraq conflict occurred once and repeatedly generating an rough approximation of Iraq in a form that a policymaker might vaguely recognize (as opposed to regarding the new Iraq-ish conflict as unique) assumes a certain degree of common factors that are held constant during each initialization. And those common factors that must be held constant are a function of other common factors that must be held constant.
Assume that our omniscient modeler creates another simulation, this time of the world itself. This model has two levels. The first level is a surface in which border shifts appear as fluctuations in the terrain and wars as periodic flashes over regions of territory. The second level is an interior world with the resolution of the Iraq model. The duration and frequency of wars is a function of the model’s global settings, specific rules, terrain, and interactions between political entities. And the model could also be started with any number of settings designed to create starting conditions.
However, where the Iraq model reset itself after every “Iraq war” concluded, the modeler has made this larger world simulation continuous. It will keep running until the “stop” button is pressed, and each world state at a given time step is a function of the previous one’s outputs. Because the range of possible behaviors, settings, starting conditions, and other related issues the challenging task of the omniscient modeler will be to reproduce as many Iraqs as possible over the course of the model.
At first the omniscient modeler might try to just aim for surface resemblance. He sets the model in a manner that produces events with some aggregate-level criteria that vaguely resembles Iraq (casualties, number of political factions, war duration, etc).
Because of the abstraction of the model and the fact that the only condition is to make as very generalized Iraq-like events, the actual behavior of the world model could be quite strange. “Iraqs’ could pop up in odd places of the world. The time elapsed between repetitions, though perhaps tunable with model settings, would also be a matter of probability. Under some conditions, Iraq wars could happen frequently. In others, the activity would be burst-like, with a cluster of Iraqs followed by a long pause. Others might involve one Iraq 200 years.
Beyond the fact that it looks like Iraq, though, the rest is unimportant. A policymaker would never know when and where the next Iraq, is right? But at least Iraq-like characteristics are visible. The catch is that the “vision” is that of the God-like omniscient modeler. In order to more closely approximate the way a policymaker might judge the “Iraq-ness” of each Iraq, the modeler needs to look at the interior level as well as the surface.
Peering into the interior level, the modeler would likely see that the interior level rarely resembles the outcomes produced by the detailed Iraq simulation conducted before the construction of the world model. This is not surprising. The political landscape of each Iraq event’s interior level is the generative product of system interactions. A very superficial resemblance can be achieved without making the evolving world model too deterministic, but this loose level of system calibration can only go so far in reproducing an Iraq that a policymaker might be able to identify with the canonical case.
So what would the modeler do? At that point, the omniscient modeler would likely toggle the global model settings, tightening the settings to make the model capable of less complex behavior. Since the specific interior conditions that he wants to reproduce have to be structurally similar to the detailed Iraq-only model, he has to set the model so it will converge towards an interstate configuration capable of producing both a plausible surface and interior Iraq.
As noted before, bottom-up models with very limited ranges of qualitative behavior often repeat themselves. However, this is taken to an extreme when the modeler must aim for a high level of detail when repeatedly reproducing an specific event from a population of interacting entities that can adapt to previous model phases. The model must now only make Iraqs that fits both the superficial exterior criteria and the detailed interior settings.
One could imagine a possible regime of world behaviors in which the model converges to detailed Iraq-producing conditions and stays there, outputting desired Iraqs that fit into a range of similar states. This might be achieved not only by altering various behaviors, starting conditions, etc, but also by giving the system memory. It might iteratively adjust itself until it could reach a state where it would produce detailed Iraqis enough of the time. At that point it would probably not really do much else except try to hug close to the configuration outcomes that make the specified Iraqs — even if it would still likely produce different configuration outcomes.
By now an already grossly unrealistic thought experiment has reached the realm of the grotesque. An international system that iteratively corrects itself until it repeatedly produces surface Iraqs and an interior Iraqs that policymakers might recognize? And then sits in semi-stasis so it could make Iraqs?
The more plausible Iraq-generating configuration still would probably not be an event that could be easily classified as “Iraq-like.” It is difficult to imagine that policymakers would only base a decision on the surface layer resemblance — the actual important elements of the war are specific to the interior level.
Finally, given the assumptions needed to make even a loose thought experiment somewhat coherent one also wonders about the probability of surface layer resemblance. A real international system, not an imaginary divinely-calibrated computer model, produces structural conditions for general classes of events (civil wars, insurgencies, drug wars) with substantial variation in basic to advanced characteristics of each event.
The issue isn’t as much fighting the last war. Rather, it’s that abstracting what is important about Iraq — and likely to continue in various guises in the future — is a political football. And the more the argument about Iraq and Afghanistan devolves into ever more-detailed historical exegesis about granular details of each episode, the more the assumption becomes that we’ll say “never again” to another Iraq or Afghanistan.
That may or may not happen for any number of reasons. We might not do an Iraq or Afghanistan again, and instead invent a new category of unpopular land occupation. That’s probably most likely, as there are any number of potential candidates in the world that could suffice to build the new category should the US become involved. And if we avoided another Iraq or Afghanistan-like event, I doubt it will because of superior wisdom gained from properly appreciating the nuances of either.