Just started reading Manuel DeLanda's 'A New Philosophy of Society: Assemblage Theory and Social Complexity'161 - a brave title! But it's got me braincells going on a Monday, so it can't be all bad. He's immediately argued against reductionism at the micro and macro levels and started talking about things being 'more than the sum of their parts'. He proposes to take the reader on a journey through all the nested levels existing between micro and macro -

It is my hope that once the complexity of that forgotten territory between the micro and the macro is grasped at the visceral level, the intellectual habit to privilege one or the other extreme will become easier to break.

Digs are had at structuralists (macro-reductionists) and economists / social scientists who build theory on the individual, and aggregates thereof (micro-reductionists.) Oh, and Anthony Giddens (a 'meso-reductionist', apparently!)

Here's some Monday musings its caused, using Icosystem's Game and the genes of ants to bounce off.

1. The Game: have a read of the page via the link above. Icosystems claim this game demonstrates that -

by using a simple agent-based simulation in which each person is modeled as an autonomous agent following the rules, one can actually predict the emergent collective behavior. Also, by using the simulation as test bed, one can explore the design of the rules to produce a desired outcome.

They do this to convince clients that their deep understanding of complexity will allow them to predict real-world complex systems, or at least offer an informed steer. The Game is a pristine little example of how a model can capture something about the real world. (Perhaps too pristine: maybe it's an agent-based version of the kind of nonlinear equations taught at school - that is, the ones that were nicely solveable.) You can play the Game with flesh and blood hoomans and the result will look the same.

The question I want to ask of DeLanda at the start of this book (and maybe he'll answer it) is - if you're going to put such a weight on building up micro-to-macro emergence, and shouting about the nesting of layers, where does the Game fit? The agent-based model has evertying the system needs to work. In the meatworld, it also works - but it does it by skipping several layers of nested physical systems, and relying on other social ones like language, to produce its dynamic result. It doesn't rely on any other feature of the physical or human world, and can get around the make-up of our actual bodies rather well. It skips these nested systems, it uses others, and it doesn't need to rely on the participants knowing where the rules they're acting out will lead. (In fact, it couldn't rely on that - it's just not that type of dynamic.) What other systems might work like that? Money? Another system that has social-symbolic rules that people act out, that skips layers of nesting, and that appears to have its own Platonic dynamic? (Has someone already done work on dynamic forms being Platonic in the same sense that static, geometric ones are? I guess they wouldn't be, since they're time-bound.)

Next, ants. In the stupidly fecund tropical rainforest in Queensland last summer I spent a good number of hours watching green tree ants do their thing. They were moving between our balcony and a climbing plant with leaves slightly larger than a human hand. It was utterly fascinating. Biologists are still trying to nail exactly how eusocial insects evolved into colonies where all but one has ceded their individual claim to pass genes on. They may be no different to any cell in a human body - carrying genes, working selflessly for the entire organism - but the biological question is, how did they get from one to the other - from individual to subsumed cell? E.O. Wilson, it seems, is shifting back towards thinking that selection happens at the whole-group level from Hamiltonian evolution. (Not that I know much about that, but the Hamiltonian rule seems to beg the question somewhat.)

There are two features of this problem I find mind-boggling. First, how does selection for emergent behaviour occur? Simple answer: same way it does for any other phenotypic thing. Evolution doesn't care if its a wing or an ability to collectively fold a leaf and build a nest in it or a penchant for cleaning the insides of predator's mouths. Survival's the thing. Is it really harder to imagine an evolutionary path to cleaning predators than it is for evolving emergent behaviour?

Maybe not, but watching ants really draws attention to it. Not only did the green leaf ants work in vast teams to fold leaves and make nests, they guarded key points of transit (in pairs, perfectly still) and bought material to the nest. (Eventually - with much pulling back and forth, the best destination for their foraging eventually emerged.)

Start from a different place, Flake says of a cellular automata version of virtual ants:

For any of these ants, we know their Theory of Everything, in that all the 'physical' laws that govern the ant's universe are simple and known to us. We also know the initial configuration of the ant's universe. Yet we are helpless to answer a simple question: does the ant ever build a highway? Putting all this in perspective, if physicists ever uncover a Theory of Everything for our universe, and even if we deduce the initial state of the universe, we may still be helpless to deduce the long-term behaviour of our own universe.160

A green-tree ant's genes can be considered its laws and (notwithstanding some notions about networks) those genes encode every aspect of a whole colony. It's perhaps easier to imagine how this could happen for an ant colony that may select as a whole unit, but what about birds and flocking? They are definitely genetically distinct and yet each contain genes that, when the bird has matured and is interacting with other birds, mean it can flock. What might that mean? Well, for ants and birds at least, evolution has a more powerful vision than any analysis or algorithmic computation could, and it links the micro and the macro because it doesn't care about the difference. As I mentioned regarding the evolving boids, evolution doesn't really care what particular reality its working in as long as there's time and selection to work with.

Sorry, tangent there. Ants and birds are simpler than social systems, but apparently we still can't model them. However, 'the Game' demonstrates that maybe that's thinking about it entirely the wrong way. An emergent dynamic - leaf-building, flocking, the Game - may exist in something many more times complex than that dynamic, and may interact over many different nested layers.

I have a mental picture that's the complexity version of planets aligning: dynamics might pop into existence for a moment as a system allows signals to connect in a certain way, before the connections drift away again. Such dynamics could appear and disappear at all scales: the point is, might they exist long enough to push a whole system in a particular direction / change its trajectory in some way? Might they learn ways to persist? In the case of ants, pheremone exchange is key. Indeed, watching the green tree ants, you can see that every single one has a little 'conversation' with every other one - without exception - that it bumps into. This involves a brief antenna-contact-waggle that may last for a split second or up to two seconds. There are stable dynamics at the pheremone level that the ants know nothing about, yet are integral in broadcasting. Trying to imagine how that might work when every ant's pheremones may be affected by every contact they have gives me much cause for squinting.

To conclude: complexity may be a distraction from the simple. Also, learning is like composting. I find that I have to feed myself really obvious things over and over again in different and increasingly complicated ways until, one day, they've sunk into the mulch, gone manky and... mm, metaphor breakdown, I'll stop.