The assumption expressed in this classic proposition has been tested to quite precise values by modern science. It is implicit in the conservation laws of physics. It's what keeps the cosmic books balanced. All of modern physics and chemistry are erected on this reliable foundation (except for that one tiny exception, the Big Bang, which created the whole of the visible universe but that's another story). It's also just common sense. Magic is trickery. Things don't simply vanish or appear from nowhere. It takes something to make something. And so on.
There is a corollary to this as well, when it comes to processes that produce new structures: the more rare, complicated, or 'well-designed' something appears, the less likely that it could have 'just happened'. It tends to take more effort and care to construct something that doesn't tend to form on its own, especially if it's composed of many complicated parts. And even 'found' objects often require effort, when 'good fit' is required. In general, when things work well despite the many ways they could potentially fail, when they exhibit sophisticated functional matching to their context, especially where this matching is highly contingent, then such 'fittedness' tends to be a product of both extensive effort and intelligent planning. So the type of thing that 'just happens' (or so bumper-stickers will tell you) is generally unattractive, undesirable, and inappropriate. Good luck is rare, bad luck is the norm, and most problems left unattended don't improve on their own.
This tendency of things to fall effortlessly into messiness the inevitable increase in entropy is the essence of the second law of thermodynamics. All other things being equal, and without outside interference (or, more specifically, in a hypothetically closed physical system in which energy neither enters nor leaves), entropy will inevitably tend to increase. There is a simple reason for this, as Gregory Bateson once explained to his daughter.[2] There are so many more possible arrangements of things that are messy (i.e. aren't regular) than those that are ordered usually vastly more. Nature, being unbiased, tends to shuffle through all the possibilities with respect to their relative probabilities of occurrence, and so the very miniscule domain of arrangements of things that are highly regular (or that we judge to be so) is often never sampled spontaneously and tends to become progressively more improbable over time.
So,when spontaneous processes like the complex adaptive functions of living bodies tend to produce increasing orderliness, complex interdependencies, and designs that are precisely correlated and matched to one another and the world, we can be excused for being just a little mystified. And when introspection confronts us with the everyday experience of living in a world of representations, anticipations, and efforts to mobilize energy to alter future conditions, we can perhaps be forgiven for treating this as magic, imported from another non-material realm. There is the dead, uncaring world and its rules, and the living feeling world and its rules, and the two seem to be quite contradictory.
On their surface, the first and second laws of thermodynamics appear to exclude the possibility of true teleological processes, such as functional design, representation, and intentional initiation of action. The notion that something absent, like a represented object or a possible future state, could be a cause of physical change appears a bit like something coming from nothing, and the possibility that functional design could arise other than by preserved accident seems to violate the very logic of physical causality. It has become common for contemporary science to treat all teleological phenomena as purposive in name only teleonomic[3] and to assume that true teleology is illusory and that the supposed role of representation and the experience of intentionality even in human actions must ultimately be epiphenomenal.
There is something unsatisfying about this denial, however. We are aware, for example, of some quite stark reversals of causal logic as certain transitions are crossed. The same atoms that constitute your body now once comprised merely inanimate bits of matter when they were scattered about the galaxy some billions of years ago. They will doubtless resume this passive existence again as they move on to comprise dust or air. Nevertheless, together in your living body they now share in a mode of existence that is quite distinctive and discontinuous from their separated inanimate existences. Together they are alive; apart they are not. And when they are together in this form, a curious and atypical inversion of thermodynamic tendencies seems to characterize the whole collection. What besides being-in-proximity makes the difference? Even if such dichotomies are illusory and there is unbroken causal continuity across the threshold from non-life to life, machine to mind, we nevertheless require an explanation for why causal architecture changes so abruptly at these transitions and why it is so difficult to follow the logic linking human teleological experience with its physical basis.
The most sophisticated early recognition of a corresponding distinction among modes of causality comes from Aristotle. In fact, he considered the problem to be even more complicated than this. He distinguished four modes of causality: material, efficient, formal, and final. If we use the example of carpentry, material cause is what determines the structural stability of a house, efficient cause is the carpenter's modifications of materials to create this structure, formal cause is the plan followed in this construction process, and final cause is the aim of the process, that is, producing a space protected from the elements. A final cause is that 'for the sake of which' something is done. For Aristotle these were different and complementary ways of understanding how and why change occurs. There has been an erosion of this plural understanding of causality since Aristotle that, although an important contribution to the unity of knowledge, may in part contribute to our present intellectual (and indeed spiritual) dilemma.
By the Renaissance, final causality was in question. Seminal thinkers like Bacon, Descartes, Spinoza, and others progressively chipped away at any role final causality might play in physical processes. Bacon argued that teleological explanations were effectively redundant and thus superfluous additions to physical explanations of things. Descartes considered animate processes in animal bodies to be completely understandable in mechanical (i.e. efficient) terms, while mental processes comprised a separate extensionless domain. Spinoza questioned the coherence of the literal sense of final causality, since it was nonsensical to think of future states producing present states. Appealing to intentions as physical causes accomplishes little more than pointing to an unopened black box. Even positing purposes for actions still requires a physical account of their implementation. And inside that black box? Well, a further appeal to purposive agency only leads to vicious regress. Accordingly, it was held, purpose and intention are intrinsically incomplete notions. They require replacement with something more substantial. A purpose, conceived as the 'pull' of some future possibility, must be illusory, lacking the materiality to affect anything. As exemplified by the early explanations of the power of vacuums and buoyancy, only 'pushes' seem allowable as determinants of the efficacy and direction of physical changes.
This heritage of modern science has guided a relentless effort to replace the black boxes and their end-directed explanations of function, design, or purposive action with mechanistic accounts. This effort has yielded astounding success. Perhaps the greatest triumph of this enterprise came with the elucidation of the mechanism of natural selection. It showed, in principle, that through accidentally produced variations, competition for resources, and selective reproduction or preservation of lineages, living mechanisms with apparent purposiveness and fittedness to local conditions could have evolved. Where the natural theology of Paley had concluded that observing functional organization, complexity, and perfection of design in an object (e.g. a watch) implied the operation of prior intelligence to fashion it, Darwin and subsequent researchers could suggest ways that the organization could be the result of preserved chance variation. The assumption that end-directedness was needed to explain the origin of these features was unnecessary. For Darwin, organisms were mechanisms like watches, but their adaptive organization could have arisen serendipitously by matching accidentally formed mechanisms with conditions favouring their persistence.
The metaphor of the world as an immense machine full of smaller machines is however deeply infected with the special assumptions of human artefact design. Hence when Richard Dawkins caricatures evolution as a 'blind watchmaker' he still characterizes organisms as machines, and machines are assembled to do something for some end. Though we typically think of organisms as analogous to engineered artefacts performing some designed task, the analogy can provide quite misleading expectations. Design is a function of imposed order that derives from outside. The integration of parts in a machine is the result of the careful selection of materials, shaping of parts, and systematic assembly, all of which occurs with respect to an anticipated set of physical behaviours and ends to be achieved. Although living processes are at least as precisely integrated with parts which are as interdependent in function as in any machine yet conceived, there is little else that makes them like anything that could have been engineered. Whole organisms are not assembled by bringing together disparate parts but by having their parts differentiate from one another. Organisms are not built or assembled. Although they grow by the multiplication of cells, these divide and differentiate from prior, less differentiated precursors. Both in development and in phylogeny, wholes precede parts, integration is not imposed, and design is a post hoc attribute. The machine metaphor is too limited. Indeed, many embryological processes and regulatory mechanisms resemble micro natural selection processes (e.g. between cell lineages) more than pre-programmed construction. Because of this engineering preconception, however, caricatures of natural selection often fall into the trap of replacing the absent designer with amazing accidents producing lucky coincidences and 'hopeful monsters' that are preserved like serendipitous inventions because of their novel usefulness.
This tacit importation of a human artefact view of the world, with its implicit design logic, into a materialist metaphysics that restricts the introduction of anything like final causal relationships, creates the logical necessity of a telos ex machina universe. In such a world we appear as accidental robots blindly running randomly generated programs. But there is an implicit contradiction in this conception, though it is not due to the exclusion of telos as much as to the limitations of the machine metaphor. Machines are intentional simplifications of the causal world. Abstractly conceived, a machine is finite and all its features and future states are fully describable. They are essentially closed off from all physical variations except those that are consistent with a given externally determined function. Thus the whole notion of machine causality is predicated on the very logic of causality that is excluded from consideration. Perhaps it would make more sense to expand beyond the watch metaphor altogether than to argue over whether it is necessary to include or exclude a watchmaker. Paying attention to the broader range of processes that share some but not all features with the logic of ends-determining-means is a good place to start the search for a middle ground.
The processes I have in mind are those that appear on the surface to violate the spirit if not the law of increasing entropy: processes in which orderliness appears to increase spontaneously rather than decrease as one might normally expect.
In the moment-to-moment processes of cellular metabolism and in the grand sweep of evolution there are myriad examples of spontaneous processes that produce an increase in order. They are harnessed with such subtle biasing as to seem almost inevitable until we consider how curious and improbable they are in the context of physical processes in general. The molecular processes that constitute metabolism in a living cell manage to produce chemical reactions that defy the odds by many orders of magnitude. The production of selected molecules can be billions of times more likely to occur in a cell than could take place in a test tube, even if one adds all the right ingredients in the right proportion in the proper sequence. This makes it appear as though cells can micromanage individual chemical reactions in ways that would exceed the wildest Machiavellian dreams of any CEO. Yet, as molecular biologists have looked into these processes more closely, it has become increasingly clear that the outcome is not accomplished by precise control of every detail, as if in some chemical processing factory where everything is carefully measured and mixed. Rather it is accomplished by a remarkably prescient system of mediating molecular relationships (catalytic relationships) which bias and constrain other molecular interactions so that they occur with vastly greater or lesser probability than if unmediated. At the molecular level, evolution appears to have made the nearly impossible all but inevitable.
Unlike the logic of machine design, however, in which things must be forced to occur, pushed into place, and restricted in their deviant tendencies, the logic of organism 'design' instead depends on recruiting the spontaneous intrinsic tendencies of molecular substrates and structural geometries. A superficial interpretation of the evolutionary process might suggest that the introduction of new ordering principles in organism design derives solely from lucky accidents achieving serendipitous functions. Yet an analysis of intracellular molecular processes and of the morphogenetic mechanisms of development suggests that a significant fraction of the order-generating processes of life are instead due to self-organizing dynamics, which are intrinsic to molecular geometries and cellcell communications. Indeed, the majority of macromolecular structures in cells tend to self-assemble, and the majority of critical chemical reactions occur within self-reinforcing cycles of reactions. Machines need to be built with extrinsic means, but organisms must develop themselves. Where there is no external means for the generation of order, order must arise from tendencies already present.
So, even if the mechanistic gambit seems on the verge of succeeding, there is some reason to be cautious about its eventual completeness, at least as modelled on the image of human artefacts. There clearly is a sense in which our descriptions of living functions and mental representations are nothing but glosses of incompletely described mechanisms. But the machine metaphor is implicitly incomplete itself, resting as it does on the assumption that the design logic of real machines can be bracketed out of consideration without changing the very meaning of mechanism. To the extent that organisms don't resemble machines in the spontaneous logic of their component processes, we should question the core assumptions of the eliminative enterprise, according to which all 'pulls' are merely illusory. Is there some way to identify a real and substantial sense of the 'pull' of future possibilities in terms of 'pushes' from the past?
Spontaneous order-production is deeply counterintuitive, like magic or something coming out of nothing, and so it demands a special explanation. The general tendency of things to degrade according to the second law of thermodynamics makes these apparently contrary phenomena stand out and appear enigmatic and intriguing. Orderly arrangements should spontaneously degrade, and they shouldn't gain complexity with time. When highly regular and complicated patterning appears spontaneously, or when future states of organization appear to be the drivers of antecedent processes, they demand a special explanation, because they give the impression of time running backwards.
This superficial appearance of time-reversal is the original motive for describing functional and purposive processes in terms of final causality. Of course, even Aristotle was clear that this could not be a literal ends-causing-the-means process, but rather something about the organization of living systems that made it appear as though this were the case. This time-reversed appearance is a common attribute of living processes, albeit in slightly different forms, at all levels of function. The highly ordered interactions of cellular chemistry succeed in maintaining living systems in far-from equilibrium states, thus resisting the increase of internal entropy, as though thermodynamic time were stopped. The increasing complexity of organism structures and processes that characterizes the grand sweep of evolution thus seems to be running counter to the trend of increasing mess and decreasing intercorrelations, as though thermodynamic time were running in reverse. Furthermore, mental processes recruit energy and organize their implementation with regard to the potential of achieving some future state that does not yet exist, thus seeming to use even thermodynamics against itself. Viewed introspectively, intentional action seems time-reversed in a particularly convoluted way. One detects both the generation of ordered behaviour without a more ordered antecedent state causing it and a currently nonexistent future state that initiates and regulates this time-reversed decrease in entropy. Of course, despite superficial appearances, time is not stopped nor running backwards in any of these processes. Moreover, thermodynamic processes are proceeding uninterrupted. Future states of things are not directly causing present events to occur. So what is responsible for these appearances? To answer this we must look beyond life and mind.
Though the epitome of this reversal of causal logic is found in living and thinking beings, the roots of the time-reversed character of these processes can be traced to inanimate processes. Less enigmatic apparent deviations from thermodynamic expectation are found in many non-biological phenomena, though they are fewer and fleeting in comparison with those in life. They are not processes that suggest any final causal logic either, because there is no end or function implied. Final causal organization, whatever it entails, is more complex. Yet many physical processes share in common with their biological and mental counterparts at least one aspect of this time-reversal character: order developing from disorder. Understanding the dynamics of this intermediate inversion of the logic of the second law offers hints that can be carried forward into our explorations of the causality behind life and mind.
In these processes we glimpse a backdoor in the second law of thermodynamics that allows even promotes the spontaneous increase of order, correlated regularities, and functional complexity under certain conditions. Curiously, these conditions inevitably include a reliable and relentless increase of entropy. In many nonliving processes especially when subject to a steady influx of energy or materials self-organizing features may become manifest. These spontaneous ordering features are dependent on the flow of energy provided by increasing entropy because they are not so much regularities of structure as they are regularities in the dynamics of a process, though it may also leave a structural trace. Among these processes are simple dynamical regularities like eddies and convection cells, coherence-amplifying dynamics like the conversion of incoherent white light into monochromatic coherent light within a laser, structural pattern-generation processes like snow crystal formation, and complex chemical dynamics like autocatalysis (all of which will be discussed in more detail below). Even computational toy versions of this logic, such as are found in cellular automata and a variety of recursive nonlinear computational processes, exemplify the need for constant throughput of change and energy as well as a recycling of constraints and biases.
What is common to all these seeming reversals of causality-as-usual is that on the surface they seem to exhibit something like a violation of the ex nihilo nihil fit dictum. Since time isn't running backwards in these cases, it appears as though the appropriate antecedent conditions aren't actually present to produce these consequences. Order should not spontaneously appear within a previously chaotic system. When it does, it gives the equally counterintuitive impression that these things must have happened due to uncaused causes. Indeed, that is the impression that our own introspection provides us when it comes to our intentional actions. We experience ourselves as originating points for action, not as mere consequences of previous states. It is as though the cause of my thoughts and behaviours, while influenced by my past states, is not determined by them and, to the extent that it is separable from these states, this intervening source me becomes something akin to a self-caused cause. Creating something from nothing cannot be what it appears, of course. Nor can the appearance of time-reversed causal sequences really reflect a correspondingly reversed physical causal architecture. But what convoluted causal relationships could make it appear as though something order is appearing out of nothing, that is, out of the absence of prior order?
Perhaps we are thinking about something and 'nothing' in the wrong way. Let me contrast the ex nihilo perspective with a quite different view presented in one of the oldest written texts in history, the Tao Te Ching, produced by the 'ancient sage', Lao-tzu:
Thirty spokes converge at the wheel's hub to an empty space that makes it useful. Clay is shaped into a vessel, to take advantage of the emptiness it surrounds. Doors and windows are cut into walls of a room so that it can serve some function. Though we must work with what is there, use comes from what is not there.[5]
This is a very different sense of nothing from that offered by medieval Western scholars as quoted at the top of this chapter a specific absence rather than, well, just nothing. I think it is more applicable to the present problem than one might otherwise have imagined. That which is empty or unfilled in these examples creates possibilities. Each case involves a highly selective absence that leaves space for something else. What was there is taken out of the way, so to speak, to make room for that something else. It is not so much the absence itself that is critical, but how it affects what is left and how this may relate to other things. The Western mind sees causality primarily in the presence of something, in the pushes and resistance that things offer. Here we are confronted with a different sense of causality, in the form of an 'affordance': a specifcally constrained range of possibilities, a potential that is created by virtue of something missing.[6]
What I want to show, using a number of examples, is that the processes of self-organizing dynamics all involve taking advantage of an affordance logic, in the sense just defined. Consider a whirlpool, stably spinning behind a boulder in a stream. As moving water enters this location it is compensated for by a corresponding outflow. The presence of an obstruction imparts a lateral momentum to the molecules in the flow. The previous momentum is replaced by introducing a reverse momentum imparted to the water as it flows past the obstruction and rushes to fill the comparatively vacated region behind the rock. So not only must excess water move out of the local vicinity at a constant rate; these vectors of perturbed momentum must also be dissipated locally so that energy and water doesn't build up. The spontaneous instabilities that result when an obstruction is introduced will effectively induce irregular patterns of build-up and dissipation of flow that 'explore' new possibilities, and the resulting dynamics tends toward the minimization of the constantly building instabilities. This 'exploration' is essentially the result of chaotic dynamics that are constantly self-undermining. To the extent that characteristics of component interactions or boundary conditions allow any degree of regularity to develop (e.g. circulation within a trailing eddy), these will come to dominate, because there are only a few causal architectures that are not self-undermining. This is also the case for semi-regular patterns (e.g. patterns of eddies that repeatedly form and disappear over time), which are just less self-undermining than other configurations. In the jargon of complexity theory, such patterns are called 'attractors', as though they exerted a 'pull' toward this form. This term captures a non-mechanical sense that is implicit in this causal logic. The flow is not forced to form into a whirlpool. This dynamical geometry is not 'pushed' into existence, so to speak, by specially designed barriers and guides to the flow. Rather the system as a whole will tend to spend more time in this semi-regular behaviour because the dynamical geometry of the whirlpool affords one of the few ways that the constant instabilities can most consistently compensate for one another.
The term that is most often used by scientists to describe the spontaneous appearance of unprecedented orderliness in nature is 'emergence'. This special use of the term has been around for over a century.[7] It still owes much of its relevance to the fact that it is applied to the same troublesome explanatory 'gaps' as it was over a century ago: the unprecedented nature of life and of mind with respect to other physical processes. The term 'emergence' connotes the image of something coming out of hiding, coming into view for the first time something without precedent and perhaps a bit surprising. Emergence used in this context is intended to convey the something-from-nothing impression that is produced when unprecedented properties are produced spontaneously without the intervention of external modifications of a system.
Additionally, most uses of the emergence concept implicitly assume an effect that is manifested at ascending levels of scale. Natural phenomena that are described as emergent tend to be mostly compositional in some sense. An early precursor to this idea can be found in discussions of the unprecedented properties of chemical compounds in comparison with those of the elements that comprise them. Thus, John Stuart Mill found the poisonous character of pure sodium and pure chlorine surprising in comparison with the dietary necessity of their compound, table salt.[8] Though most contemporary scientists prefer to reserve the term 'emergence' for describing more systemic phenomena, this sense of discontinuity due to compositional effects remains a persistent refrain. Scale is of special importance to the problem of emergence because an increase in numbers of components increases iterative interaction possibilities. With every iterated interaction, relational properties are multiplied with respect to each other, so an increase in numbers of elements and chances for interactions increases the relative importance of interaction parameters and related contextual variables. Consequently, a somewhat more extensive definition of emergence might be something like:
unprecedented global regularity generated within a composite system by virtue of the higher-order consequences of the interactions of composite parts.
Over the past few decades, this compositional usage has become more and more prominent as scientists in different fields have encountered similar transitional patterns in systems as diverse as liquid convection patterns and the appearance of unprecedented social dynamics. In non-technical discussions the phrase 'the whole is more than the sum of the parts' is often quoted to convey this sense of novelty generated via ascent in scale. This phrase originates with Aristotle and captures two aspects of the emergence concept: the distinction between a merely quantitative difference and a qualitative one, and effects involving the combination of elements whose patterns of interaction contribute to global properties that are not evident in the components themselves. There is something a bit misleading about this way of phrasing the relationship that harkens back to a something-from-nothing conception. Exactly what 'more' is being appealed to, if not the parts and their relationships, is seldom made explicit.
This additive conception has often led to the expectation that new classes of physical laws come into existence with increases in scale and the interaction effects that result. This conception of emergence is often described as 'strong emergence' because it implies a dissociation from the physics relevant to the parts and their relationships. It is contrasted with 'weak emergence' that does not entail introduction of any new physical principles. The latter is often seen merely as a redescriptive variant of standard reductionistic causality, and thus as emergence only with respect to human observers and their limited analytic tools. In this essay I will argue that we can still understand the emergence of novel forms of causality without attributing it to the introduction of unprecedented physical laws. Indeed, I will argue that only to the extent that an unbroken chain of causal principles links such higher-order phenomena as consciousness to more basic physical processes will we have an adequate theory of emergence.
In the last decades of the twentieth century the concept of emergence has taken on a merely descriptive function in many fields. It is applied to any case of the spontaneous production of complex dynamical patterns from uncorrelated interactions of component parts. This shift from a largely philosophical to this more descriptive usage of the term emergence has been strongly influenced by the increasing use of computational simulations to study complex systems. Some of the more elaborate examples of these phenomena have been the topics of so-called chaos and complexity theories, and have become commonplace in computational models of dynamical systems, cellular automata, and simulations of non-equilibrium thermodynamic processes. This more general conception of emergence has been adopted by many other fields where complex interaction effects may be relevant, such as in the social sciences. Evolutionary and mental processes are also treated as producing emergent effects, though the complexity of evolution, not to mention cognition compared with dynamical systems, suggests that more subtle distinction between kinds of emergence may be necessary (see below). Because of this terminological promiscuity there is likely to be no common underlying causal principle that ties all these uses together. Nevertheless, I think that with care a technical usage tied to a well-characterized class of empirical exemplars can be articulated for which a clear theory of emergent processes can be formulated.
The exemplars of emergent phenomena that serve as guides for this analysis occupy a middle position in the taxonomy of different emergent dynamics that I describe below. They represent a well-understood set of physical and computational systems that all share a form-amplifying, form-propagating, form-replicating feature. This feature is exhibited irrespective of whether they are physical or computational phenomena. These phenomena are often called self-organizing, because their regularities are not externally imposed but generated by iterative interaction processes occurring in the media that comprise them. They serve as a useful starting point because they allow us to extrapolate both upward to more complex living phenomena and downward to simpler, merely mechanistic phenomena.
I decry using emergence as an anti-reductionistic code word in holistic criticisms of standard explanations. In this use, the concept of emergence is a place holder, indicating points where standard reductionistic accounts seem to be incomplete in explaining apparent discontinuities. In this negative usage, emergence serves only as a philosophically motivated promissory note for a missing explanation that, critics argue, is needed to fll in a gap. In contrast, the purpose of the present essay is to outline a technical sense of emergence that explicitly describes a specifc class of causal topologies (i.e. self-constituting causal structures) and then attempts to show how this may help to explain many of the attributes that have motivated the emergence concept. This approach avoids engaging the pointless semantic debates about the completeness of reductionism or dealing with metaphysical questions about the ontological status of emergence. The term will only be applied to well-understood empirical processes, and yet I will argue that it does indeed mark the transition to unprecedented and indecomposable causal architectures.
It may be wondered, then, what more besides a taxonomic exercise is provided by identifying the emergent architectural features of known physical processes? By providing an explicit account of how apparent reversals of causal logic come about, how variant forms of these processes are related to one another, and what aspects of their dynamic organization are most critical to the development of these attributes, we can gain critical perspective on the apparent discontinuities between simple mechanistic and teleological models of causality.
The image of a snake biting its own tail (ouroboros) is an ancient sign for the mysterious. Circularity is also the key to unlocking the mystery of the apparent time-reversed causality of self-organizing and teleological processes. The principal hypothesis of this essay is that emergent phenomena grow out of an amplification dynamic that can spontaneously develop in very large ensembles of interacting elements by virtue of the continuing circulation of interaction constraints and biases, which become expressed as system-wide characteristics. In other words, these emergent forms of causality are due to a curious type of circular connectivity of causal dynamics, not a special form of causality. This circularity enables certain distributional and configurational regularities of constituents to reinforce one another iteratively throughout an entire system.
The relative autonomy of higher-order 'holistic' properties of complex systems is largely a function of this recycling of constraints and biases. It is also the means by which apparent 'top-down' effects from global system attributes may come to influence the properties and dynamics of constituents. By virtue of an amplification dynamic, higher-order causal properties can be generated that effectively 'drag along' component constituent dynamics, even though these higher-order regularities are constituted by lower-order interactions. By means of these circles, nature tangles its causal chains into complex knots in such a way that the global effects can come to resemble a reversal of time. Discerning the major topological variants of these 'knots' of causal organization and identifying the conditions under which they form is the primary aim of this theory of emergence.
Wherever it occurs in nature, amplification is accomplished by a repetitive superimposition of similar forms. It can be achieved by mathematical recursion in a computation, by the recycling of a signal that reinforces itself and cancels the uncorrelated background noise in an electronic circuit, or by repetitively sampling from the same biased set of phenomena in a statistical analysis. In each case, a reciprocal relationship between interaction (or sampling) regularities and form regularities serves as the basis for amplification. Amplification depends on redundancy of form and on a process that enables a repeated reinforcement of these redundancies while damping non-redundant variations. In this way, certain minor or even incidental aspects of a process can come to be the source of its dominant features.
Coupling these two factors a stochastic amplification logic and reciprocally reinforcing patterns of bias and interaction constraint serves as the basis for the present account of emergent dynamics. Additionally, by distinguishing progressively higher-order nested forms of this circularity we will be able to differentiate between mere order-from-chaos forms of emergence (e.g. self-organization) and teleological processes.[9] Historically, theoretical discussions of complexity and emergence have regularly cited examples with this causal architecture whether in terms of non-linear dynamics or computational recursion but to date I know of no effort to formalize this intuition or to use it as a general analytic tool.
Perhaps the simplest and best known example of 'circular causality' is embodied in a thermostatic control system. By connecting a heating device to a temperature-sensitive switch located in the space being heated, the coupled devices can be configured to change one another's states reciprocally. This creates a self-undermining pattern of cause and so-called negative feedback which tends to produce behavioural oscillation around some set-point. If this causal linkage is reversed, so that deviation away from the set-point activates mechanisms to cause the environmental temperature to deviate yet further, a very different and unstable behaviour results so-called positive feedback. This latter runaway is checked only by outside constraints. Even simple deterministic engineering devices where a number of such feedback control devices are coupled together can produce highly complex quasi-periodic behaviours or even deterministic chaos as the time-lags in effects interact.
Though emergent effects arise from an analogous logic of nonlinear interactions, and in part derive their causal indirectness from it, emergent dynamics differ from simple feedback dynamics by virtue of the contribution of massively stochastic features. Recursive causal interactions that develop up-scale in large stochastic systems exhibit progressive amplification of feedback-like effects between different dynamical levels. As distributional and configurational features of components and their interactions become differentially damped and amplified by virtue of their circulating influences, their global characteristics can further bias these component interaction patterns. Both runaway and self-regulating effects can in this way be manifested at a higher-order system level. . In more colloquial terms, one might describe it as 'compound interest' of form across adjacent levels of scale such that global attributes alter component attributes alter global attributes, and so on.
Three general categories of emergent dynamics can be derived in this way, and they are distinguished from one another by the way recurrent causal architectures can be embedded in one another across levels of scale. This embedded relationship can be described as non-recurrent, simple recurrent, and hyper-recurrent causal architectures (in the latter, simple recurrent causal architectures are embedded in a yet higher-order recurrent architecture). These produce phenomena that I will correspondingly call first-, second-, and third-order emergence, respectively. In the discussion that follows I will argue that many thermodynamic effects correspond to first-order emergent relationships; that self-organizing phenomena (the prototypical exemplars of emergence in most current discussions) correspond to second-order emergent relationships (a mode of causality I will call morphodynamics); and that life, evolution, and mind all correspond to third-order emergent relationships (a mode of causality I will call teleodynamics).
The most basic class of emergent phenomena, exhibiting what I have called first-order emergence, are higher-order thermodynamic phenomena. This sense of the term emergence is often applied to descriptively 'simple' higher-order properties of stochastic systems. Some commonly cited examples are liquid properties. Laminar flow, surface tension, viscosity, and so forth are all first-order emergent properties in this sense. Statistical dynamics and quantum theory have provided a remarkably complete theory of how the properties of molecules can produce liquid properties under appropriate conditions. Thus in one sense they are considered to be fully reducible to relational molecular properties. But such relational properties, as opposed to intrinsic molecular properties (e.g. mass, charge, configuration of electron shells, etc.), are not symmetric across levels of description. Precisely because they are relational, these higher-order properties are not applicable to descriptions of, for example, water molecules in isolation.
More importantly, interaction relationships between molecules are what become amplified and summed to produce aggregate behaviours that emerge as liquid properties with ascent in scale. This is why a highly diverse class of molecular species are capable of exhibiting similar liquid behaviours in appropriate conditions. Philosophers of science often refer to the dependence of higher-order properties on lower-order properties as 'supervenience'.
Liquid properties supervene on these lower-order properties, including their interaction effects, and are therefore entirely determined by them. And yet we require a separate explanation for the fact that these properties are also to some extent independently converged upon despite a diversity of substrates. Liquid properties reveal an independence from the details of matter and energy with ascent in scale, even though these details contribute to the particular values of liquid behaviour parameters (e.g. viscosity). This fact suggests an interesting reducibility issue that has been periodically noticed. Knowing all the details of the liquid parameters does not allow us to predict such component details as the molecular structure of components or the many microscopic peculiarities of their interaction features, except in a statistical sense, because the stochastic features of interactions reflect combinatorial relationships between the various parameters. So there are many possible ways that different micro-details of structure and interaction can converge to produce the same higher-order properties. A given higher-order liquid property 'supervenes' on specifc lower-order interactions to the extent that the former always entails the latter, but the vast iterative dynamics of these interactions also has a variety-cancelling that converges to similar results across a wide range of substrates and modes of interaction.
Before continuing, it is worth reflecting on a parallel that will become more relevant later in this essay. This many-to-one mapping is analogous to a related mapping issue in the philosophy of mind, which has been cited extensively in comparisons of mental processes to computing. This asymmetric many-to-one relationship between substrate and higher-order properties is analogous to the core assumption of a paradigm called 'functionalism'. Functionalism is basically the view that it is the form of a process, not its substance or its energetics, that determines its intentional (read: mentalistic) properties, and that the same form embodied in different media (read: same algorithm on different computing platforms) is functionally the same. Forgetting the mentalistic implications for a moment, notice that calling a multitude of molecular systems 'liquids' effectively exemplifies this logic, that is, that a collection of entities is expressible as a single functional state. Of course, the fact that this hierarchic re-description is a quite generic feature of compositional entities (such as liquid water) suggests that it is not likely, by itself, to provide key insights into the emergent features of mind, any more than it changes what we think of water.
Liquid water properties 'supervene' on the properties of water molecules primarily because of relational features. In repeated microscopic interactions the specifc unique features of individual molecules (e.g. their charge, geometry, orientation, momentum, internal vibration, etc.) distribute in such a way as to cancel one another in aggregate, thus leading to a higher ordered state.
These astronomically many details cancel out, except for the average effect expressed globally via the relative linearity of the summed stochastic processes.
The net result is a reduction in complexity and increase in regularity that correlates with ascent in scale, and in the whole system with continuation in time (e.g. an increase in entropy). Only attributes that are additive and non-cancelling are relevant. This selective cancellation and amplification of interaction parameters is, I suggest, the key to emergence even if in simple thermodynamic systems the result is a direct extrapolation from micro to macro. In the real world, with vastly many parameters that can interact in any process, it is almost inevitable that their non-correlations and non-coherence will result in a cancelling dynamic.
So why consider these higher-order relationships emergent? The answer is that they are what I would call stochastic dispositional properties (for want of a more compact description). These properties and the trends they exhibit (i.e. expressions of the second law of thermodynamics) are not merely the results of Newton's laws. Their aggregate character requires a statistical account because of the additional critical role played by distributional features. In a very general way, as will shortly be made more evident, the lesson of emergence is that 'shape' matters. By shape, I mean something quite general including ultimately the geometry, topology, form, and so on of components, their distributional characteristics, their interaction possibilities, and their boundary conditions in general. Shape matters because it introduces dimensional biases, and these can sometimes uniquely amplify instead of cancelling with scale. Typical thermodynamic conditions are, all other things being equal, cases where shape-specifcity can be ignored.
The cancelling dynamic of a simple thermodynamic system dominates because there are no special features caused by the influence of shape at the component level that could constitute a process of reciprocal amplification via iteration of interactions. For this reason there is also no way for large-scale regularities (e.g. macroscopic interactions, distribution asymmetries, etc.) to reinforce or amplify complementary biases in microscopic interactions. Hence large-scale patterns of distribution and interaction ultimately dominate, making this a standard case of simple supervenience. The result is a causal transparency from micro to macro of the sort that makes reductive analysis possible.
Phase changes represent a special case, though ultimately they are an exception that proves the rule. Change of phase is a higher-order property, one that can have a macro-to-micro biasing. In super cooled water, for example, the seeding of crystallization can produce a rapid chain reaction that is accelerated the more molecules become bound to the growing lattice. The seed for crystallization is an external factor; by virtue of shape effects, it then initiates a process of microscopic interaction that rapidly propagates throughout the system. In such cases, the propagated shape- alignment into the crystal lattice also has a cancelling that is evened out throughout the system. Crystallization is intrinsically a shape-determined dynamic. There are conditions where this dynamic can produce amplified biases that affect macroscopic properties which in turn can reinforce these microscopic biases even further. Such a runaway amplification is exhibited by snow crystal growth, which is discussed below as an exemplar of a second-order emergent dynamic.
Philosophical and scientific discussions of the mindbrain mystery often invoke some notion of supervenience to model the presumed relationship between higher-order mental phenomena and the lower-order cellular-molecular processes on which they depend. But it is clear that the functionalist analogy with simple thermodynamic properties isn't nearly adequate. One major factor glossed over in such direct comparisons is development across time. In thermodynamically simple systems the features of each individual component (say, a molecule) are uncorrelated with those of others. Consequently, the properties of the constituents do not cause any consistent biasing effects over large numbers of interactions; the temporal development increasingly tends to cancel deviations from the normal distribution. Another way to think about the inexorable trend to increasing entropy in simple thermodynamic systems is that the simple, non-interactive state is the unbiased condition unbiased by external perturbations and unbiased by internal form relationships. By contrast, as we will see, very different trends can develop if biased conditions prevail, whether they are due to extrinsic or intrinsic influences. In these cases, the very same interaction dynamics that normally 'cancel out' perturbations can, in fact, come to amplify them.
The thermodynamic simplification processes that I have so far described including the second law must be understood as something more than mere mechanism. Classical thermodynamics assumed Newtonian mechanics but required something more as well: an account of parameters affecting the pattern of the average interaction (i.e. shape factors). Thus thermodynamic properties might better be understood as physical dispositions of material systems, because they depend critically on formal, distributional, and configurational contributions to change.
Simple thermodynamic dispositions (e.g. the tendency for entropy of a system to increase) can be characterized as processes of system change that are unbiased by any structural or temporal regularities. These are dispositions in which configuration variables and regularities of system perturbation are uncorrelated, and so cannot introduce reinforcible biases because they reciprocally cancel. In any system of components with even a modestly high dimensionality of potentially variable properties non-correlation is the overwhelmingly likely condition. But there are conditions where this is not the case conditions that produce (more or less) chaotic dynamics or self-organized behaviours. Under chaotic conditions, for example, certain higher-order regularities become unstable, resulting in unpredictable global dynamics. This unpredictability of chaotic systems derives from the fact that the interaction dynamics at lower levels become strongly affected by regularities emerging at higher levels of organization. This can produce a deviation-amplifying dynamic that propagates throughout the system. If perturbations of this type are incessant, bias comes to dominate over distributive tendencies.
A classic simple example is the formation of Benard cells in a heated liquid. These are regularly spaced hexagonal convection cells of hot-rising and cool-descending liquid that form spontaneously if there is relatively uniform depth and an even heating from below (see Fig. 5.2). This phenomenon depends on thermodynamic tendencies settling into higher-order stable states; it is also the product of the constant perturbation of these regularities by continuous heating. As heat is conveyed out of the liquid by moving molecules, others must take their place in such a way that there is no persistent local accumulation of heat due to uneven convection. The hexagonal regularity forms because hexagonal close-packing is geometrically the most even and dense distribution of regions of constant size on a surface. In Benard cells, the precise regularity of dynamical organization emerges out of a more or less disorganized state as the various unstable patterns of convection mutually cancel. The system eventually settles into this hexagonal tessellation of columns of rapidly circulating liquid because this close-packing pattern most uniformly distributes dissipation of heat by moving liquid.
If the geometric constraint on close packing is to express itself, so to speak, a variety of factors must be present, including uniformity in depth, uniformity of heating, a large surface to volume ratio, and limited asymmetry of the sides of the container. Their effects are contributed by virtue of what they don't do: they don't introduce countervailing biases or asymmetries. So the role played by symmetry constraints in 'attracting' the dynamics of convection to converge to a regular hexagonal pattern is generic. In other cases, however, the symmetry constraints may come from the components. Consider for example the amplification of form constraints in growth processes, where constant instability is introduced by continually adding similar components, as is the case in snow crystal formation.[10] The structure of an individual snow crystal reflects the interaction of three factors: (1) the hexagonal micro-structural biases of ice crystal lattice growth, inherited from water molecule symmetry, (2) the radial symmetry of heat dissipation, and (3) the unique history of changing temperature, pressure, and humidity regimes as a developing crystal falls through the air. Snow crystal growth occurs across time in diverse regions in a variable atmosphere, the history of temperature and humidity differences it encounters is captured and expressed in the variants of crystal structure at successive diameters. In this way, the crystal is effectively a record of the conditions of its development. But snow crystals are more than merely a historical record of these conditions because of a 'compound interest' in which prior stages of crystal growth progressively constrain subsequent stages. So even identical conditions of pressure, temperature, and humidity, which otherwise produce identical lattice growth, can produce different patterns depending on the prior growth history of the crystal. The global configuration of this tiny developing system plays a critical causal role in its microscopic dynamics; it excludes the vast majority of possible molecular accretions and growth points and strongly predisposes accretion and growth at certain other sites (see Figs. 5.3 and 5.4).
Snow crystals are self-organizing. Reciprocally-reinforcing biases of molecular configuration and the contingencies of crystal growth together determine their macroscopic patterning. Contingent events in the growth history of the crystal also play an important role in determining the final configuration. For example, as partially formed crystals or water droplets randomly collide and freeze onto the growing crystal lattice they unbalance its temperature and bias subsequent growth, as the temperature asymmetry propagates throughout the developing crystal to influence the probability of subsequent accretions. As growth continues, the increasingly complex crystalline form leads to progressive constraint on the potential growth options.[11] In this sense, snow crystal growth also includes the unpredictable influence of these random accretions and incorporates them into the complex symmetry of the crystal. This includes even the effects of melting and refreezing, resulting in symmetric semi-regular shapes as well. This is what contributes to the proverbial individuality of each crystal.
Laser physics provides another example of a shape-mediated amplification that is manifested in temporal regularity. Lasers produce intense beams of monochromatic light where all the waves are in precise phase alignment. Light with these precisely correlated features is called coherent light; it is generated from white light, which contains mixed wavelengths aligned in every possible phase. The conversion of white light to coherent light is accomplished by virtue of the recurrent emission and reabsorption of light by atoms whose emission features correlate with their excitation features. When the energy of out-of-phase polychromatic light is absorbed into the electron shells of the atoms of the laser material, it is incorporated into a system with very specifc energetic regularities. When the polychromatic light energy is re-emitted as the atom reverts to a lower energy state, the light it emits carries a discrete amount of energy and is thus in a specifc frequency. If the lasing material is uniform, the excitation results in uniform colour output. Amplification of the features of this light is achieved by causing the emitted light to re-enter the laser by virtue of partially silvered mirrors. Thanks to the character of the light-absorbing-and-emitting atoms and the frequencies intrinsic to their structure, light at this emission frequency is most likely to induce an energized atom to emit its excess energy as light, and in a phase that is precisely correlated with the exciting light. Repeated charging with white light and recycling of emitted light thus amplifies this pattern by many orders of magnitude. To return to my previous analogy, this recurrent emission and reabsorption results in a 'compound interest' of both frequency and phase.
Consider one final example of a second-order emergent phenomenon, albeit one that is more indirectly determined by shape: autocatalytic reactions. In snow crystal dynamics the micro-configuration of each molecule is the same, producing symmetric interactions and strongly constrained structural consequences. When a system is composed of different types of components it can also exhibit a more distributed interactional reflexivity. In autocatalysis the interaction of a set of different molecules is constrained both by the configurational properties of the whole collection, as above, and by the configuration symmetries and asymmetries that exist between the micro-configurations of the different classes of its components.
For example, molecules that interact in a highly allosteric fashion that is, they weakly bond selectively with some but not other types of molecules can constitute interaction sets with more elaborate self-organizing features. Both the configurations of the different classes of individual interactions and the configuration of the whole set of possible interactions become critical organizing influences. This can occur in a chemical 'soup' that contains enough different types of molecules. Among all these types there is a subset in which each type of molecule can catalyze the formation of some other member in the set, thus constituting a closed loop of catalyzed syntheses. So long as suficient energy and other raw materials are available to keep reactions going (i.e. it must be an open system), the set will continue to be 'autocatalytic'. A functioning autocatalytic set will play an inordinate role in determining both what chemical reactions can take place and how the soup as a whole will be constituted. It is this higher-order distributed ordering and reordering of the interactions of the different classes of constituents that matters. Such a system can generate far more complex micro-and macro-dynamics than if the interactions were symmetric. Chemical reactions with these features were well-described by Prigogine and colleagues a generation ago;[12] they have since become the basis for extensive research with both real and simulated chemical systems. Ultimately, the metabolic dynamics that constitute living cells depends on numerous fully and partially autocatalytic sets of molecules. Together, these sets constitute a system dynamics that is 'autopoietic' (literally, 'self-making').
What do these examples of second-order emergent phenomena have in common? In each case we find a tangled hierarchy of causality, where microconfigurational particularities can be amplified to determine macroconfigurational regularities and where these in turn further constrain and/or amplify subsequent micro-configurational regularities. In such cases, it is more appropriate to call the aggregate a 'system' rather than a mere collection, since the specifc reflexive regularities and the recurrent causal architecture are paramount. Although these systems must be open to the flow of energy and components which is what enables their growth and/or development they additionally include a closure as well. These material flows carry structural constraints inherited from past states of the system which constrain the future behaviours of its components. As material and energy flows in, through, and out again, form also recirculates and becomes amplified. In one sense this form is nothing more than a set of restrictions upon and biases toward possible future material and energetic events; in another sense, it is what defines and bounds the higher-order unity that we identify as the system. This centrality of form-begetting-form is what justifies calling these processes morphodynamic.
We find a further difference, however, between merely chaotic or self-organizing emergent phenomena, like snow crystal growth, and evolving emergent phenomena, such as living organisms. The latter, in addition to the effects mediated by second-order processes discussed above, also involve some form of information or memory (as represented in nucleic acids,for example) that is not seen in second-order systems. The result is that specifc historical moments either of higher-order regularity or of unique micro-causal configurations can additionally exert a cumulative influence over the entire causal future of the system. In other words, thanks to memory, constraints derived from specifc past higher-order states can get repeatedly re-entered into the lower-order dynamics which lead to future states. This is what makes the evolution of life both chaotically unpredictable on the one hand, and yet on the other hand also historically organized, with an unfolding quasi-directionality.
For these kinds of phenomena we must introduce a third order one that recognizes an additional loop of recursive causality that transcends and encloses the second-order recursive causality of self-organized systems. Third-order emergence inevitably exhibits a developmental and/or evolutionary character. It occurs where there is not only an amplification of the global influences on parts, but also a redundant 'sampling' of these influences which reintroduces them into different realizations of the system over time. The result is, in, a higher-order stochastic process extending across time that like the limited stochastic processes of thermodynamics and morphodynamics is capable of both cancelling and amplifying biases. Under these conditions, there can be extensive amplification of lower-order relationships (both supervenient and self-organizing relationships) due to the fact that historical residues of such processes repeatedly get re-entered into the system.
Third-order emergence is to morphodynamic processes as second-order emergence is to thermodynamic processes. It occurs when there is a recursive stochastic amplification of complementary morphodynamic relationships. In other words, it is a function of one non-equilibrium process that tends to converge to a stable pattern which is then reciprocally reinforced by another. But whereas second-order emergent phenomena involve amplifications among energetically coupled thermodynamic regularities, third-order emergent phenomena involve the amplification of reciprocally reinforcing morpho-dynamic relationships despite vast spatial, temporal, and energetic separation.
Third-order emergence is the basis for the selection logic of evolution. Its reciprocity of form and production creates a dynamic that can be called self-similarity maintenance. This is the basis for the existence of discrete individuals and lineages that are linked by unbroken continuity of changing structures and relationships. It is also the basis of what we mean by memory. This maintenance of a discrete unit through the correlation of form and dynamics is a necessary condition for evolution. We might thus describe natural selection as a stochastic 'exploration' of variant morphodynamic relationships of reciprocity with respect to environmental regularities. For such exploration to take place, morphodynamic processes must be reliably reproducible.
Third-order emergent phenomena can thus be considered as a form of the self-organization of self-organizing dynamics. As morphodynamic processes become increasingly synergistically interdependent over the course of evolution, the amplification of complexity and of self-organizational dynamics can become enormously complex. This is because memory allows every prior morphodynamic relationship itself to become a potentially amplifiable initial condition contributing to any later relationship. Through the amplification of size and time constraints alone, the stochastic amplification capacity is vastly greater than what can occur within a morphodynamic process. Second-order energent, that is, morphodynamic, processes depend on energetic continuity, but third-order processes require morphodynamic continuities, and these, as we have seen, are to some extent substrate independent. This linkage by form, rather than by shared specifc material or energetic substrate, allows for a much vaster domain of amplification. Distant separation in time and the disruption of energetic continuity are not barriers. Moreover, because there is a remembered trace of each prior 'self ' state contributing to the dynamics of future states, such systems develop not merely with respect to the immediately prior state of the whole, but also with respect to their own remembered past states. This contributes to the characteristic differentiation and divergence from, and the convergence back toward, some 'reference' state, which organisms standardly exhibit.
In order to describe the relationship of representation that is implicit in third order emergent phenomena, we need to employ a combination of multi-scale, historical, and semiotic analyses (analyses based on the relationships between signs). This is why living and cognitive processes require us to introduce concepts such as representation, adaptation, information, and function in order to capture the logic of the most salient emergent phenomena. It is what makes the study of living forms qualitatively different from other physical sciences. It makes no sense to ask about the function of granite, or the purpose of a galaxy. Though the atoms composing a neurotransmitter molecule or a heart-muscle fibre have no function in themselves, the particular configurations of the neurotransmitter molecule or the heart and its cell types do additionally beg for some type of teleological assessment, some function. They do something for something. Organisms evolve and regulate the production of multiple second-order emergent phenomena with respect to some third-order phenomenon. Only a third-order emergent process has such an intrinsic identity.
So life, even in its simplest forms, can't be fully understood apart from either history or representational relationships. Indeed, it may be that any third-order emergent system must be considered 'alive' in some sense. This suggests that third-order emergence may be something like a definition of life. If this is so, then the origins of life on earth must also be the initial emergence of third-order emergent phenomena on the earth. More generally, emergence of this type constitutes the origination of information, semiosis, and teleology in the world. Its embedded circular architecture of circular architectures definitely marks the boundary of a unit of causal self-reference that is extended both in space and time. It is the creation of an 'epistemic cut', to use Howard Pattee's felicitous reuse of a classic phrase: the point where physical causality acquires (or rather constitutes) significance.
Any of the components of an organism say, a haemoglobin molecule can be given an arbitrarily complete and precise description in the language of atomic physics or chemistry, and yet this description will miss something that is nevertheless materially relevant to its structure and its very existence. Specifcally, it will provide no hint of why this highly improbable molecular configuration is so prevalent, as compared with the astronomical number of molecular forms that are not present. Haemoglobin, and indeed any complex structure within an organism, has the structure and properties it does because it is embedded in a vast elaborate evolutionary web. This evolutionary disposition is the third-order analogue to the increase in entropy.
Comparing haemoglobin to a molecular form of even vastly less complexity for instance, a diamond reveals the comparative incompleteness of describing haemoglobin merely in terms of its structure and physical properties. Knowing the atomic structure of the carbon atom gives us a considerable ability to predict the probability that a diamond crystal will form. In contrast, knowing the atomic structure of haemoglobin provides almost no information about its probability of formation, its prevalence in certain environments, why it is found in context with certain other molecules, and why a normal distribution of related molecular forms is nowhere to be found.
Every atom in a haemoglobin molecule has a determinate physical history. Specifc converging tributaries of 'pushes' from one molecular event to another over vast stretches of time helped to determine how each of the thousands of atoms (created in perhaps dozens of distant supernovae) came together to form a given haemoglobin molecule. But this is almost irrelevant. The more important causal story is told by what is not around, what did not end up as part of the molecule. Haemoglobin's existence must be seen against a backdrop of vastly more numerous molecular forms that were eliminated via natural selection, leaving haemoglobin as the one representative of the set. And this is so for every complex biomolecule as well as for their dynamical relationships to one another within each organism.
Almost every feature that biologists find interesting about haemoglobin has to do with how it fits with other things in the living context in which it has long been embedded. This fit was not created by the 'push' of specifc antecedent molecular events, but by the evolutionary cancelling dynamic of natural selection that pushed alternative forms out of existence. Haemoglobin occupies the space of possibilities that was left. The 'function' that haemoglobin provides is thus vaguely analogous to an 'attractor' in a self-organizing dynamic. Haemoglobin is what is left after the self-cancelling consequences of not-fitting-well have cleared away potentially competing similar configurations. So in physical terms what haemoglobin is, is a result of what it is not.
Molecules like haemoglobin exist, then, because of a process that historically 'captures' and interlinks many self-organizing processes. In a functional sense, the dynamical forms of these self-organizing processes are the equivalents of the spokes and rim of the wheel in the Taoist verse. They determine a constitutive absence with respect to the conditions they require in order to persist. The historical process that stabilizes these dynamical forms with respect to each other and the available resources is of course evolution. Looked at in this way, however, we can see that evolution must involve the self-organization of self-organizing processes, and so must be a higher-order relationship emergent from morphodynamic relationships. But how can self-organizing processes self-organize? The answer is that certain relationships among morphodynamic processes must be capable of producing the necessary and suficient conditions for evolution.
Elsewhere I have described in some detail a simple example of a system of self-organizing processes that together spontaneously bring an evolutionary (i.e. teleodynamic) process into existence from morphodynamic precursors (Deacon, 2006). It is a simple molecular system called an autocell; its basic logic is depicted in Fig. 5.7. It is comprised of two interlocking self-organizing (i.e. morphodynamic) processes: an autocatalytic process and a self-assembly process. Autocatalysis occurs when the catalyst that aids the formation of one molecule is itself (either directly, or indirectly by the intermediary of other catalysts) a catalyst that aids the formation of the first. This produces a circle of catalytic reactions that becomes self-amplifying. Self-assembly is essentially a form of crystallization in which duplicate molecules tend to accrete into larger aggregates with specifc geometric forms. In life these are typically tubes, sheets, and polygons. Autocellularity occurs when one catalyst in an autocatalytic set is also able to self-assemble into a structure that can contain other catalysts. Thus, autocatalysis will generate molecules that tend to enclose regions of space that are likely to include the catalysts of the very set that creates such enclosures. This makes autocells self-repairing if they are broken open; moreover, they are potentially self-reproducing if brokenopen in the vicinity of suficient raw materials to support many additional cycles of autocatalysis. The transition from self-reproduction to selection dynamics occurs as an autocell lineage happens also to enclose one or more additional molecules that get caught up in the autocatalysis and increase, in some manner, the reproductive capacity (e.g. by increasing rate, reliability, or matching to more plentiful substrates in the environment). In this way autocells can spontaneously evolve, even though they are not in any typical sense alive.
The point of introducing this example is that autocells embody a definite potential as well as the tendency to achieve this potential. They manifest a definite self-other relationship; their parts can be said to have functions with respect to this potential; and their evolutionary 'adaptations' can be seen as embodying information about the environment. These are teleodynamic features not evident in simpler systems or in any isolated components. Finally, one does not have to postulate in advance any particular assumptions about information or take as given the existence of information-bearing molecules like DNA in order to understand these teleodynamic features. The autocell's teleodynamic features are emergent: they are embodied and instantiated in the dynamical topology that is constituted by the interdependency of morphodynamic processes.
This self-reproducing 'potential' may be viewed as a higher-order constitutive absence. It is what defines the autocell as distinct from a mere colocation of self-organizing processes. One can interpret the further emergent potential to evolve, and thus to generate additional new 'aboutness' relationships as in, a capacity to generate constitutive absence. The locus of this capacity is physical and material; and yet with each replication the thread that ties this potential together is only its complex causal topology passed down through the generations and even this can become further augmented and differentiated over time. Such could be the precursor of life.
In the case of autocells, the embodied potential is also a tendency to achieve that potential. Specifcally, it is the tendency to reconstitute the morphodynamic and thermodynamic resources that are required. As described, it is a self-realizing potential or, to put the point differently, a constitutive absence that tends to fill itself.
In this way, the adaptations of organisms are like the wheel or the vessel in the Taoist verse quoted above. Organism adaptations, and the processes they include, are materially bounded structures and processes; and yet, in a curious way, they are defined by a fundamental incompleteness. Third-order emergent dynamics are thus intrinsically organized around specifc absences. This physical disposition to develop toward some target state of order merely by persisting and replicating better than neighbouring alternatives is what justifies calling this class of physical processes teleodynamic, even if it is not directly and literally a 'pull' from the future.
All three of the dynamics we have discussed have one general logic in common: they can all be described as processes in which the most salient feature is a 'least-discordant-remainder'. In other words, it's not so much what was determined to happen that is most relevant for future states of the system, but rather what was not cancelled or eliminated. It is the negative aspect that becomes most prominent. This is the most general sense of constitutive absence: something that is produced by virtue of determinate processes that eliminate most or all of the alternative forms. It is this, more than anything else, which accounts for the curious 'time-reversed' appearance of such phenomena. This logic itself becomes self-reinforcing in teleodynamic processes, because they are the result of a least-discordant-remainder dynamic operating on a least-discordant-remainder substrate higher order constitutive absences based on lower order constitutive absences.
This is, of course, also the essence of representation, or intentionality: something whose existence is conditional upon something it is not. It is this feature of mental phenomena that has most mystified scholars for millennia: their 'aboutness'. The implication of the present analysis is that the 'constitutive absences' characteristic of both life and mind are the sources of this apparent 'pull of yet unrealized possibility' that constitutes function in biology and purposive action in psychology. The point is that absent form can indeed be eficacious, in the very real sense that it can serve as an organizer of thermodynamic processes. We are now in a position to explain more precisely how the specifc absence of something can itself can do work, that is, how a possibility can constitute a locus of thermodynamic 'push'.
What this three-level analysis suggests is that a constitutive absence derives its efficacy by virtue of a series of thermodynamic and evolutionary reversals (a combined 'double reversal') which each results in a least-discordant-remainder dynamics. These reversals are the consequence of distributed dynamical interactions that stochastically cancel each other out, leaving serendipitously non-discordant tendencies in their stead. Through the progressive layering of what are essentially negative determination processes, the organizing capacity of these constitutive absences is amplified until, at the level of human mental causation, it appears that a very large fraction of all material and energetic processes in the body are entrained by what is no more concrete than the 'conceivably possible'. Let's try to break these double-reversals down into their component steps.
The first reversal occurs via morphodynamics: self-organizing processes that generate regularity by virtue of the spontaneous reciprocal cancelling of non-reinforcing forms of dynamical interactions. Morphodynamics can be caricatured as a process of falling toward regularity through the mutual cancellation of pushes occurring in most alternative configurations. The dynamic form that 'survives' and persists is in this sense 'left over' after others have taken themselves out of the way. Although the stable forms that arise and are eventually amplified in morphodynamic processes are perhaps best viewed as reliable side-effects of the underlying thermodynamics, this thermodynamic basis remains a necessary condition. Ultimately, this necessary coupling carries over into the higher-order relationships among the morphodynamic processes that constitute life and evolution.
The second reversal occurs via teleodynamics: the amplification of morphodynamic synergies due to the differential preservation of more contextually fitted variants. Evolution is the paradigm exemplar of this second reversal. As in the previous case, Darwinian selection processes can also be caricatured as a 'falling toward' or 'backing into' regularity. In this case, however, the relationships between morphodynamic processes are what are pitted against one another; they are the units that use up resources for self-replication. But when morphodynamic processes themselves fall into reciprocally reinforcing relationships as they do in life they do so only to the extent that they maintain mutually reinforcing thermodynamic conditions at the same time. This interdependence of form and dynamics constitutes the condition for selection, because it allows alternative formform relationships to be 'sampled' by virtue of their thermodynamic correlates (i.e. the relative 'cost' of the morphodynamics that produced them). The formform relationships that tend to persist and propagate in evolution are those generated by morphodynamic linkages that minimize chaotic dissipation by 'falling into' dynamical short-cuts between dynamical forms. This is the essence of 'fittedness' in a biological sense. It is both a formal and an energetic relationship, one that is continually reconstituted and updated by virtue of mutually reinforcing least-discordant-remainder processes at work across levels of scale.
Darwinian selection processes, like morphodynamic processes, are the expression of indirect and mutually cancelling 'pushes'. But what is doing the 'pushing' if the morphodynamic processes that constitute organisms are themselves a reflection of the space of least resistance in a context where all other pushes cancel? Since the mutually reinforcing morphodynamic processes that define 'organism' are essentially a linked set of convergent tendencies due to non-resistance, it might seem that they would be unable to provide any source of resistance themselves. Yet, it turns out, the self-similarity maintenance that results from a series of morphodynamic processes can itself determine a locus of resistance. The component morphodynamic dispositions of an organism reflect an underlying non-equilibrium thermodynamics; hence the reciprocal relationships between the various morphodynamic processes that allow organisms to remain self-similar over time must be organized so that they maintain the self-similarity of these non-equilibrium dynamics as well. The competition on which natural selection is based arises from these thermodynamic 'requirements', culling morphodynamic relationships with respect to their relative thermodynamic consequences. Again, the work is done by thermodynamic processes. But this work is harnessed to create formal and thermodynamic conditions that are not immediately present, driven by the tendency to resist deviation from a target form that incessantly reconstitutes itself. This is what ultimately licenses functional terminology in biology: component processes and structures are indeed organized 'for the sake of' achieving future target states of least discord or best fit.
Like the reciprocal form-reinforcing relationships that constitute morphodynamic processes and produce a higher-order appearance of thermodynamic time-reversal, teleodynamic processes produce a yet higher-order appearance of morphodynamic time-reversal. In morphodynamic processes, order can spontaneously arise without a similarly ordered antecedent state. In teleodynamic processes, specifc order can arise spontaneously because of its specifc absence in an antecedent state. In this way, absent order can in effect bring itself into being.
This is only the beginning of an analysis of teleological causality. It offers no more than a demonstration of plausibility, albeit one that has long been needed and missing. There are many embeddings of these processes that must be considered just to get to biological systems as they are currently understood, and many many more above the level of biological functionality before we can approach anything like what is involved in mental processes. Within this analysis I think we can nevertheless discern a modus operandi for ascending the hierarchy of processes. The key additional ingredient can be found by noticing that the three-step hierarchic embedding required to achieve this simplest level of teleology function is itself susceptible to recursive embedding. Embryological processes, neural development, and (I predict) neural signal processing itself represent progressively embedded teleodynamic processes within teleodynamic processes. With each progressive embedding, the achieved adaptations and functions of lower domains become the ground for subordinate higher-order teleodynamic processes; one might think of it as evolutionary dynamics 'for the sake of ' evolutionary possibilities. In this way, by backing into possibility level upon level via least-discordant-remainder dynamics, teleological causality has grown. The thermodynamic constraint on which forms of constitutive absences can come to affect which forms of physical processes has, correspondingly, been radically reduced.
In this essay I have defined three subcategories of emergent phenomena that can be arranged into a hierarchy of increasing topological complexity, each growing out of and dependent on emergent processes at the level below. Thus third-order emergent processes (teleodynamics) require self-amplifying second-order emergent processes (morphodynamics) to create their necessary conditions, which in turn require self-amplifying (non-equilibrium) first-order emergent processes (thermodynamics) to create their necessary conditions. Conversely, teleodynamics is a special limiting case of morpho-dynamics, which is a special limiting case of thermodynamics. The three categories are distinguished by their causal topology in the sense that the circularity of their dynamics creates a certain 'closure'. This closure helps to explain both the discontinuities that they evidence with respect to one another and the reversal in dispositions as one ascends from one to the others.
As a result, while it is technically correct to say that life and mind supervene on chemical processes, it is misleading to say that they are 'merely' or 'nothing but' chemical processes. Moreover, because higher-order emergent phenomena are dependent on and constituted by lower-order emergent phenomena, their probability of formation is substantially lower. There is a vastly higher probability of the spontaneous formation of simple thermodynamic phenomena than morphodynamic phenomena, and a vastly higher probability of the spontaneous formation of morphodynamic phenomena than teleodynamic phenomena. But whereas it is almost astronomically improbable that teleodynamic systems might form spontaneously, whenever they do their self-similarity-maintaining dynamic results in a powerful disposition to further reinforce their persistence, which we call evolution. As such a system evolves, it becomes able to expand vastly the self-reinforcing interconnections of this organizational pattern in which underlying morphodynamic and thermodynamic relationships are (or are made to be) mutually complementary. Spontaneously generated morphodynamic phenomena are transient and unstable, however, so the vast majority of morphodynamic processes in the world occur within organisms. This tail-wagging-the-dog reflects the higher-order disposition of teleodynamic relationships to self-replicate, thereby replicating their constitutive morphodynamic features.
We are now in the position to give a more precise formulation to the insistent criticisms that systems theorists have made both of genetic reductionism in evolutionary theory and of computational reductionism in cognitive theory. Life and mind cannot be adequately described in terms that treat them as merely supervenient because this collapses the complex levels of emergent relationships that stand in between. More critically, supervenience analyses entirely overlook the defining dispositional reversals that occur within these higher-order transitions. As a result, these analogies miss the most salient and descriptively important dynamics of these phenomena, which are precisely what make them emergent in the sense discussed above.
In many ways, I see this analysis of causal topologies as a modern reafirmation of the original Aristotelian insight about categories of causality. Whereas Aristotle simply treated his four modes of causality as categorically independent, however, I have tried to demonstrate how at least three of them efficient (thermodynamic), formal (morphodynamic), and final (teleodynamic) causality are hierarchically and internally related to one another by virtue of their nested topological forms. Of course there is so much else to distinguish this analysis from that of Aristotle (including ignoring his material causes) that the reader would be justified in seeing this as little more than a loose analogy. The similarities are nonetheless striking, especially considering that it was not the intention to revive Aristotelian physics.
There is a sense in which all is 'reducible' to thermodynamics (efficient causality), though only to the limited extent that the higher-order forms assume lower-order forms in their constitution. But the topological closure created by the circular relationships at each succeeding level makes each of the two higher-order dispositional dynamics irreducible to mere combinations of the lower-order forms. Ignoring these topological transitions, as reductionistic analyses do, also obscures the source of the higher-order reversals of disposition, which is what distinguishes the formal and the final causal levels from simple efficient mechanisms.
So what are the implications for the efficacy of human desires, reasons, and intentions? Of course, the elaboration of this dynamic in neurological processes, which produces that peculiarly convoluted version that we call thought, does not yield to any simple solution. Our brains are constituted of hundreds of billions of densely interacting cells, each of which is itself a vast third-order emergent dynamo. In a sense, each nerve cell is sentient in some small way by virtue of its necessary functional organization and incompleteness. This fact creates, among the linked neurons, an affordance to one another that involves them in first-, second-, and third-order processes of a higher rank than that which is internally regulated within each alone.
In addition, a symbolic species such as Homo sapiens has further entangled the causal architecture of its billions of minds in a vast higher-order emergent semiotic web. This web is characterized by symbolic self-organizing and by evolutionary processes that are quite different from those at lower levels. In addition to the least-dissonant-remainder effects of the various underlying levels of genetic teleodynamic processes (including neurological, embryological, and evolutionary processes), the further distributive power of symbolic communication itself provides a multi-stage dissociation from specifc thermodynamic factors. A symbolizing mind has perhaps the widest possible locus of causal influence of anything on earth. Minds that have become deeply immersed in the evolving symbolic ecosystem of culture as are all modern human minds may have an effective causal locus that extends across continents and back millennia, and which grows out of a locally least-discordant-remainder dynamic involving hundreds of thousands of individual communications and actions. Each symbolically mediated thought is the emergence of a specifc 'constitutive absence'; each is a specifc variant instance of an evolved adaptation within this vast spatially and temporally distributed ecology. This immense convergence of causal determination is coupled with an equally vast capacity for selective amplification via the teleodynamics of neural processing. With so many levels of amplification and causal inversion mediating between brain chemistry, conscious cognition, and symbolic evolution, it is no wonder that we experience symbolically mediated causality as almost completely disconnected from thermodynamic causality, even though its very efficacy is founded upon it.
Human consciousness with its features of autonomous causal locus, self-origination, and implicit 'aboutness' epitomises the logic of emergence in its very form. Like something coming out of nothing, the subjective self is, in effect, a constitutive absence for the sake of which new constitutive absence is being incessantly evolved. In this sense, there is some legitimacy to the eliminativist claim that there is no 'thing' that it is. Indeed this must be so. The locus of self is, effectively, a negative mode of existence, that can act as an unmoved mover of sorts: a non-thing that nonetheless is the locus of a form of inertia a resistance to change with respect to which other physical processes can be recruited and organized. Consciousness is not exactly something from nothing. It merely appears this way because of the misdirection provided by the double-negative logic of the least-dissonant-remainder processes involved. It is, nevertheless, a form of being that is constituted by what it is not, and yet remains a locus of physical influence. It is the hole at the wheel's hub.