Philosophy & Simulation: A Review

A review of Manuel DeLanda’s Philosophy & Simulation: The Emergence of Synthetic Reason

This Christmas I received a copy of Manuel DeLanda’s recent book, Philosophy and Simulation: The Emergence of Synthetic Reason. It’s a fascinating book, for two reasons: it puts forth a quite different ontological position from any that I’ve encountered, and it does so by exploring phenomena at all scales, and the patterns and phase spaces that they seem to share.

It’s a demanding read, not because the philosophical or the scientific or the mathematical concepts are particularly difficult, but because there’s an awful lot of them, and if you’re unfamiliar with them as they arise you may be best off ducking off to Wikipedia and looking them up before proceeding, as they will come up again and again. (In particular, make sure you at least have a passing familiarity with finite state automata, neural nets, fitness landscapes and genetic algorithms).

DeLanda begins with some fairly simple definitions; an individual (we’ll come back to that term, it’s philosophically important here) possesses properties, tendencies and capacities. These are very distinct terms. DeLanda uses the example of a knife to illustrate. Sharpness is a property, as it describes a spatiotemporal phenomena, such as the arrangement of metal molecules so as to make an edge, and different arrangements make the edge wider or narrower. It is a tendency of the knife to be solid; given sufficient heat, the tendency of the knife to becomes liquid will become actualized (there’s another philosophical term to watch out for). The tendency for the knife to become a gas is still a virtual tendency, here. And finally, a knife’s capacity to cut implies a relationship to another individual that is capable of being cut. Since this is a relation, it can only be spoken of with regards to another individual, and the phase space for the knife’s capacities is theoretically infinite.

Why the term individual? Well, this is roughly how DeLanda defines entities that the OOO crowd (folks like Kris Coffield, Timothy Morton, Ian Bogost, Levi Bryant and Graham Harman) would call objects, but with some key differences. An individual is shorthand for an individual singularity, a type of assemblage (see DeLanda’s earlier book, A New Philosophy of Society: Assemblage Theory & Social Complexity, for more on that) that is irreducible and decomposible. The irreducible part is how DeLanda gets away with talking about emergence. A convection cell isn’t emergent because it magically adds something to the base phenomena, but it’s emergent because it’s only visible from a certain spatiotemporal scale. If you look at the constituent atoms, you won’t find the convection cell anywhere. Decomposibility means that emergent phenomena are just looking for materials to fit a certain structure. Cells split and die, but I am still an individual. Presidents take and leave office, but the government still exists, new electrons are constantly taking the place of old but the current is the same.

More importantly, an individual isn’t defined by arbitrarily drawing a bounds (Which I still see as a distinctly anthropocentric activity, although Kris Coffield argues otherwise. I still can’t tell if the problem is me or him. My money’s on me.), but by describing the history of its construction. In this sense, DeLanda marries the entity-driven descriptions of object-oriented ontology with a sort of contingent process philosophy.

The properties of individual singularities have material presence in spacetime; they are composed of matter and energy, and to this degree are compatible with general observation, most science, and naïve realism. In addition, however, the individual singularity describes the phase space for its tendencies and capacities; a space that is infinite without possessing all possibilities, and where probability clusters shape the phase space to form attractors.

An explanation for what properties, tendencies and capacities of individual singularities are actualized is necessary, as is an explanation for the evolution of their phase spaces. Universal singularities are just such an explanation, existing as mechanism-independent topologies that individual singularities map to, completely or incompletely. Individual and universal singularities are what DeLanda demands an ontological commitment to, not merely the spacetime matter-energy subset. Indeed, our entire universe may merely be the actualization of a much large singularity that describes a phase space for universes, virtual and actual.

As an aside, while DeLanda never uses the term, this appears to be the exact same ontology described by Amanda Gefter in her answer to this year’s Edge question as Structural Realism.

Throughout the rest of the book, then, DeLanda explores the plausibility of this premise by attempting to map rough estimates of certain phase spaces via simulations and seeing whether the attractors that arise correlate to what we know of the history of the universe, chemistry, life and sociology. The simulations necessarily become more simplistic as the scale and complexity increases, but DeLanda is not looking to prove the thesis as much as show that it is plausible and not grossly contradicted, and in this he succeeds.

However, there are two important things he missed, that I think would actually strengthen his case should he include them. The first is that, in a work so replete with cutting-edge algorithms, the absence of the Price equation is frankly, stunning. Every simulation DeLanda describes operates at one scale, abstracting the scales below it. What the Price equation does is allow for variable-selection across scales. This is not important in a majority of the cases, but the minorities can be game-changers (think of the higher-order effects of one cancer cell, or one person like Napolean or Gandhi). Any discussion of simulations as a way of exploring phase spaces should, in my opinion, at least take a serious look at what the consequences are of blindly abstracting lower-order phenomena.

Secondly, while DeLanda didn’t omit the subject, I don’t think he spent enough time explaining and exploring the relationship of emergent phenomena and available gradients, particularly as they relate to entropy. Gradients are [paraphrasing DeLanda here] “intensive differences that act so as to store and release energy”. These could be the presence or absence of a valence electron, or the fact that all of your relatives signed up for Facebook (something which fits nicely into what Levi Bryant has been exploring as “regimes of attraction”). As John Tooby explains in his answer to 2012’s Edge question:

The world given to us by physics is unrelievedly bleak. It blasts us when it is not burning us or invisibly grinding our cells and macromolecules until we are dead. It wipes out planets, habitats, labors, those we love, ourselves. Gamma ray bursts wipe out entire galactic regions; supernovae, asteroid impacts, supervolcanos, and ice ages devastate ecosystems and end species. Epidemics, strokes, blunt force trauma, oxidative damage, protein cross-linking, thermal noise-scrambled DNA—all are random movements away from the narrowly organized set of states that we value, into increasing disorder or greater entropy. The second law of thermodynamics is the recognition that physical systems tend to move toward more probable states, and in so doing, they tend to move away from less probable states (organization) on their blind toboggan ride toward maximum disorder.

Entropy, then, poses the problem: How are living things at all compatible with a physical world governed by entropy, and, given entropy, how can natural selection lead over the long run to the increasing accumulation of functional organization in living things? Living things stand out as an extraordinary departure from the physically normal (e.g., the earth’s metal core, lunar craters, or the solar wind). What sets all organisms—from blackthorn and alder to egrets and otters—apart from everything else in the universe is that woven though their designs are staggeringly unlikely arrays of highly tuned interrelationships—a high order that is highly functional. Yet as highly ordered physical systems, organisms should tend to slide rapidly back toward a state of maximum disorder or maximum probability. As the physicist Erwin Schrödinger put it, “It is by avoiding the rapid decay into the inert state that an organism appears so enigmatic.”

The quick answer normally palmed off on creationists is true as far as it goes, but it is far from complete: The earth is not a closed system; organisms are not closed systems, so entropy still increases globally (consistent with the second law of thermodynamics) while (sometimes) decreasing locally in organisms. This permits but does not explain the high levels of organization found in life. Natural selection, however, can (correctly) be invoked to explain order in organisms, including the entropy-delaying adaptations that keep us from oxidizing immediately into a puff of ash.

Natural selection is the only known counterweight to the tendency of physical systems to lose rather than grow functional organization—the only natural physical process that pushes populations of organisms uphill (sometimes) into higher degrees of functional order. But how could this work, exactly?

It is here that, along with entropy and natural selection, the third of our trio of truly elegant scientific ideas can be adapted to the problem: Galileo’s brilliant concept of frames of reference, which he used to clarify the physics of motion.

The concept of entropy was originally developed for the study of heat and energy, and if the only kind of real entropy (order/disorder) was the thermodynamic entropy of energy dispersal then we (life) wouldn’t be possible. But with Galileo’s contribution one can consider multiple kinds of order (improbable physical arrangements), each being defined with respect to a distinct frame of reference.

There can be as many kinds of entropy (order/disorder) as there are meaningful frames of reference. Organisms are defined as self-replicating physical systems. This creates a frame of reference that defines its kind of order in terms of causal interrelationships that promote the replication of the system (replicative rather than thermodynamic order). Indeed, organisms must be physically designed to capture undispersed energy, and like hydroelectric dams using waterfalls to drive turbines, they use this thermodynamic entropic flow to fuel their replication, spreading multiple copies of themselves across the landscape.

Entropy sometimes introduces copying errors into replication, but injected disorder in replicative systems is self-correcting. By definition the less well-organized are worse at replicating themselves, and so are removed from the population. In contrast, copying errors that increase functional order (replicative ability) become more common. This inevitable ratchet effect in replicators is natural selection.

Organisms exploit the trick of deploying different entropic frames of reference in many diverse and subtle ways, but the underlying point is that what is naturally increasing disorder (moving toward maximally probable states) for one frame of reference inside one physical domain can be harnessed to decrease disorder with respect to another frame of reference. Natural selection picks out and links different entropic domains (e.g., cells, organs, membranes) that each impose their own proprietary entropic frames of reference locally.

When the right ones are associated with each other, they do replicative work through harnessing various types of increasing entropy to decrease other kinds of entropy in ways that are useful for the organism. For example: oxygen diffusion from the lungs to the blood stream to the cells is the entropy of chemical mixing—falling toward more probable high entropy states, but increasing order from the perspective of replication-promotion.

Entropy makes things fall, but life ingeniously rigs the game so that when they do they often fall into place.

In short, entropy does not preclude phase spaces in which the attractors lead to complex phenomena, particularly when the frame of reference one is using includes a gradient that makes a relatively closed system open.

I very much enjoyed this book, although I grew a little weary of the detailed discussion of simulation limitations in the last two chapters. For a broader exploration of what this means for human social systems, politics and economics, I highly recommend DeLanda’s earlier work, A Thousand Years of Nonlinear History.

I hope these ideas are seriously engaged with by the philosophic and scientific communities.