// My Complexity thread in SA is starting to pick up some discussion. Here’s an essay I wrote for the discussion:

McDowell wrote:

Adam Curtis’ “All Watched Over By Machines of Loving Grace” deals with the history of Systems Theory, Ecology, and the political implications – primarily in part 2

I’ll repeat that this is a terrible documentary. Systems thinking and cybernetics should definitely not be conflated with individualism or Randian-style libertarianism, yet the documentary takes a critique of the latter as sufficient for damning the former. The move is not without precedent; as I mentioned earlier, Hayek famously argued (as you and Curtis seem to be endorsing) that the complexity of natural systems (especially human social and economic systems) makes them impossible to model and predict, and therefore the project of governing and planning for such systems is a hopeless waste of time, causing more problems than it solves. Hayek concludes the obvious free market libertarian positions; Curtis is a little more reserved and simply critiques the hype over computers as a stabilizing and organizing force.

While it is true that computers aren’t necessarily a stabilizing force (anyone who has lived for the last 20 years has plenty of empirical evidence to the contrary), it is just as true that computer modeling is a successful way of generating reliable predictions in some domains, and that the predictive success of a model depends a lot on the nature of the model and the complexity of the system being modeled.

Perhaps this is a place to talk a little more about complexity. One of the defining characteristics of a complex system is that there are many perspectives to take on the system, not all of which will be consistent, but each of which might nevertheless be useful for making predictive inferences about the future behavior of the system. Modeling complex systems face precisely the challenge of integrating and unifying multiple perspectives (including a variety of methodological approaches) into a single predictive framework.

Aside for Guy DeBorgore: For this reason, I’m interested to hear more about the critique from the Marxists and Dialectical Materialists, because as far as I can tell systems theory requires the same sensitivity to perspectival pluralism and science as a pragmatically grounded social project as the Marxists would endorse. I see thinkers like Quine, Rorty, and other neo-pragmatist thinkers, who identify coherence, solidarity, and consensus as primary epistemological or methodological virtues, to be the early progenitors of systems-thinking in the philosophy of science. While neither Quine or Rorty are Maxists, the fight they have with Marxists seems to be a brotherly fight between more or less like-minded positions, and I radically unlike the fight Maxists have with capitalists or libertarians, or that the neopragmatists have with essentialists or foundationalits.

I suppose I should just read that book. /aside

In any case, systems thinking on my view requires a kind of methodological pluralism that I don’t think is being appreciated by McDowell (or Adam Curtis). Take the incredibly important issue of modeling the global climate, which is a paradigmatic complex system. Lots of competing processes contribute to climate dynamics, and we have dozens of climate models that highlight certain features of the climate. Consider, for instance, the Milankovich cycles, which model the Earth’s climate relative to exposure to sunlight.

These models generate all kinds of predictions and interesting correlations, but it is only one of an enormous number of factors that contribute to climate dynamics. These cycles, for instance, don’t begin to consider the various feedback dynamics that contribute to the global climate, including the ice albedo effect and the rate of atmospheric carbon. Each of these feedback systems contribute something to the dynamics of the climate, but it is often extremely difficult to know how to integrate these factors into a single model.

And that’s just the problem of the foundations of the model; there is an entirely different problem of incorporating data sets of evidence from a varity of different sources, each using different techniques and assumptions, but all of which contribute to the model. Atmospheric carbon estimates taken from ice cores occasionally yield different results than samples taken from C. davisiana estimates in deep sea cores, without any obvious or easy eay to reconcile discrepancies.

This is why you often see temperature estimates that look like this:

which just overlays many different and otherwise incompatible models to spot global trends that seem to be predicted by all models. So in the above graph (which is rather out of date), we see a clear consensus among all the models of an increasing temperature over time, though each model predicts a different rate of temperature increase, and thus might yield different policy decisions on that basis.

These models and data are all relevant to the climate; a philosopher would say they are all about the climate. Yet each offers a different (and possibly incompatible) description of the situation, yields (possibly incommensurate) predictions, and each might suggest different (and contradictory) courses of action going forward. None of these models are sufficient for modeling the climate on their own, but reconciling the models is an incredible challenge. The IPCC, which is the UN council on climate change, simply uses the arithmetic average of the temperature rise predictions in these models to generate their policy suggestions.

Hopefully it is clear that the plurality of models here makes the average used by the IPCC a completely meaningless number, and it is no surprise that their conservative predictions and policy suggestions tend to miss the mark rather dramatically. Nevertheless, there are obvious and pressing pragmatic reasons to settle on a consensus model in order to enact successful policies for dealing with a disaster of this scale. This is the fundamental social challenge presented in dealing with the complexity of these systems.

Quote:

I’m skeptical that anything and everything can be quantified, fed into a computer, and modeled. If the predictions/suggestions of a model are politically inconvenient they will be fought/ignored (climate change) – and forcing people to conform to model behavior creates problems as well.

For what it is worth, nothing about the challenge described above suggests that planning and managing a complex system (even the climate) is impossible, as Hayek (and Curtis) imply. Hard, but not impossible. Consider the much easier task of, say, cultivating a garden, which is also an unbelievably complex system that can be approached from a variety of perspectives, but no one would deny that we can manage and tend a garden to yield fruitful results. It is a challenge, but it is not impossible.

Notice also that nothing I’ve said requires that we “quantify everything” into a single model. Dealing with complex systems will always require dealing with a plurality of models and learning how to negotiate them all depending on the tasks at hand.

Consider the garden again: sometimes cultivating the garden requires tending to the soil; other times it requires pest control, and still others require consideration of plant varieties and seasons for planting. Each of these domains require a more or less specialized knowledge of that domain, and though each has intimate relations to the others (they are all about garden cultivation), the models in one area are not likely to be very useful when moves to another domain. Being a skilled gardener requires some familiarity with all these domains, and the understanding to know how to move fluidly back and forth between them as and when required.

As we continue to fail ourselves in our completely amateur attempt at dealing with global climate change, it also becomes increasingly clear that it isn’t the complexity of the environment that poses the major challenge. Rather, the tools we have for meeting that challenge (especially State Capitalism) are showing themselves to simply not be up to the task. Luckily, the same tools that are exposing these systemic weaknesses are also allowing us to organize ourselves into more skilled and capable configurations.