Cliodynamics, or the study of historical processes

Cliodynamics and science-fiction

Cliodynamics is a field which very existence is controversial. Among all subdisciplines of cultural evolution, “history as a science” is the only one that consistently draw criticism when it is discussed - while its claims are by no mean stronger than those of paleoanthropological history or cultural attraction theory. Two points may stroke as odd in such arguments : the first is that the very same thesis receive very different treatment if they are labeled “cliodynamics”, “cultural evolution”, or “historical anthropology” ; the second is that critics of cliodynamics rarely answer any specific point made by cliodynamicians, instead attributing them claims akin to Asimov’s psychohistory.

Part of this misunderstanding can be attributed to the mix of dubious communication choices by proponents of cliodynamics - who are eager to present their work on the evolution of States or secular cycles in political integration as some sort of uncontextual, physics-like knowledge - and paradigmatic reactance from its opponents. However, the recurrent invocation of psychohistory, a fictive discipline that is widely remote from anything that could meaningfully be articulated in a scientific enquiry, suggests the existence in both communities of deep-rooted misconceptions regarding the actual scope and capabilities of the natural sciences.

Indeed, the defining element of these sciences is to understand the world by identifying what causal relations shape what we can observe. In the case of highly complex systems such as living organisms or their societies, it implies to let go altogether of any claim of platonic “truth” concerning any observable entity, and instead focus on integrative theories of what processes shape them. This epistemological leap from collecting facts about things to inferring what drives their change is, for example, the defining feature of post-Darwinian theoretical biology.

We will discuss how this “process ontology” relates to cultural evolutionary theory, and how it relates to the epistemic modes of “mathematical modelling” and “prediction” through a crossed analysis of Turchin’s theory of social complexity, and Turing’s early account of morphogenetic development.

The birth of mathematical morphogenesis

While better known for his foundational model of computing machine, Turing’s fascination with the diversity and recurrence of life either helped structure or reproduced major findings regarding the physics of life, the development of organisms, and the study of self-organisation. He lived through a time of renewal for Darwinian thought - in fact, he was contemporary to the modern definition of genetic determinism in Schrödinger’s “What is life ?” (1946), which introduced the notion of genetic “code” before the ADN molecule was even discovered. This definition led to the paradigm of population genetics which came in the 60s to define mainstream evolutionary science, and still hold today a position of weak dominance.

Yet, Turing was critical of Schrödinger’s computer metaphor. As the person who defined modern computing theory, he was indeed well placed to understand its limits. Indeed, the top-down deterministic architecture of computers had little similarity to the messy, unruly physical world of the living. Unlike a computer code, deterministically enforced by a specialised processor, a gene has little to no coercive power over the information it is supposed to code : it can only manage to regulate some reaction time parameter in an irreducibly complex and stochastic metabolic system, and only if it happens to be expressed by the course of this very system.

Turing, whether in response to that fallacy or because of an unrelated interest, started to draw the mathematical formalism of earlier theoretical biologists such as Waddington, who accented the role of development and dynamical relations within an organism on the emergence of its mature form. The following work on morphogenesis formed the core of the mathematical theory of reaction-diffusion. Contrary to the intuition from equilibrium or linear modelling, it showed that the steady supply of energy to a chemical system may cause it to break symetry spontaneously : for example, to grow an head and a tail, a front and a back, a set of members…

Reaction-diffusion theory showed nothing less than the inability and irrelevance of genes to explain some key features of organisms’ development, and the centrality of much simpler physical processes to define their forms. Yet, these results came out of heterodoxy only after significant development and replications in the 90s, and their consequences are yet much less widely discussed than those of genetic adaptation. They indeed lacked the notion of direct causation and mathematical tractability of Schrödinger’s computing metaphor.

The power of models

The point of this account of Turing’s work is not to stress the promethean aspect of his figure, nor is it to deplore the appeal of simplistic explanations even in mainstream sciences. It is to illustrate the uncanny power of theoretical models in driving empirical research. Indeed, by defining the process through which biological causation was to be understood, the “genetic code” metaphor came to delimitate the mainstream in evolutionary science, in such an extent that both mathematical proof and empirical account of other means of causation were ignored or dismissed as speculation.

Human cognition works in such a way that objectivity is illusory. We have priors that shape our perception, a general process studied under the term of Bayesian inference. We have a social structure that echo the opinion of powerful people, and grant power according to how much one’s opinion is echoed. The models through which we understand the world are cultural constructs which once contracted proceed to shape every single one of our decisions - including those about the relevance of said models.

Among those models, there are some that formally transcribe alleged causal pathways : in Turing’s morphogenesis model example, it is the interplay between the passive diffusion of chemical species and the active production of those species by cellular metabolism which drives the emergence of asymmetry in developing organisms, interplay which is captured in part through his reaction-diffusion equations.

Those allow an epistemological manoeuvre that, one could argue, make up for the absence of individual or collective objective rationality. This manoeuvre consists in testing of the formal model by drawing its expected consequences in some particular conditions, and consequently trying to observe whether those consequences are indeed verified. This strategy, which we will call “instrumental prediction”, is the core of the mainstream popperian delimitation between science and pseudo-science.

Models and instruments in the natural sciences

In light of Popper’s deliberate attempt to frame the historical development of natural sciences as “true science”, and of significant relaxations of his theory in later works, it is wiser to let go of this pretension to define scientificity. Instead, we will just claim that instrumental prediction happens to be what defines the natural sciences, and discuss what it means about the dynamics and capabilities of these sciences.

The epistemology of prediction, at its core, implies that scientific theories must be systematically tested in controlled settings where we can precisely interpret our observations in the language of the theory, and discarded the moment we make contradictory observations. Together with the principle of parsimony, it means that natural sciences must define only the very entities that are necessary to predict the phenomenons they study. The principle of instrumentality, as we define it, means to convey two additional features of natural sciences : first, that contradictory observations mean nothing if they can be attributed to a failure to actually measure what we intended to measure ; second, the predictions a theory can make mean nothing where its core bricks cannot be found.

To get back to Turing’s morphogenesis example, a natural scientist should test that some real world dynamics can be accounted for with reaction-diffusion modelling before they take it seriously. They could for example identify reactants which dynamics follow Turing’s model, and manipulate their concentration in order to assess their role in development. Once it is established that reaction-diffusion drives important dynamics in morphogenesis, they should abstain to attribute causality to the specifics of genetics in their discussion of these dynamics (and conversely, abstain from recruiting reaction-diffusion in their explanation of evolutionary dynamics) unless they can observe such causality. Finally, they should delimitate clearly what this model is able to explain : reaction-diffusion successfully predicts some spatial features of morphogenesis, but not much regarding how organisms react to stress.

“Prediction” is in this context unrelated to the temporality of events, in the extent that it is not a statement about the future. Rather, the term is recruited here in its cognitive meaning : a prediction is an agent’s belief about a yet unobserved state of the universe, which stems from the union of preexisting conception of the world and a priori unrelated observations, and serves both to validate or refute these conceptions and to guide this agent’s behaviour in context of limited access to information. In the case of scientific predictions, the agent happens to be a scientific community, and the conception of the world a scientific theory, but this general characterisation still holds. All auxiliary means recruited by scientists to try and test their theory are understood as “instruments”, be it microscopes or protocols to assess whether a specific chemical work as a reactant.

Process ontology, or how models become true

The characterisation of natural sciences I drafted here has some interesting features. For example, the entities scientists recruit are judged to be existing or not only based on their relevance to models predicting observable phenomenons - if I need evolution theory to account for such and such observations, then there is such a thing as evolution. On the other hand, any single observable fact may be rendered obsolete at any point by a change in standard methodology or mistakes in its interpretation, and cannot be describe as scientifically “true”.

However, mathematical model are by construction concerned with how things change and relate to each other. The same model may apply in similar extents to structures very different in their substance, a feature that is only limited by our ability to instrumentally assess what are the relations between elements of these systems. The very same reaction-diffusion equations that Turing defined in his study of morphogenesis, for example, are recruited in accounts of neurocomputational mechanisms of perception and decision, and of agglomeration formation (accounts which, again, could be proved mistaken at any time). In other words, where the substrate of reaction-diffusion may change, its defining dynamics remain.

Therefore, natural sciences as characterised here mostly deal with mathematical understanding of relations and changes, rather than of observable things in themselves. In other words, the natural scientific world is defined by regularities in how changes happen, rather than the things it is home to. This conception is equivalent to the philosophical position known as “process ontology”, stating that the existence of change is causally precedent to the existence of objects or substance. In synergy to this principle, natural sciences are structured by their understanding of some categories of change (such as “gravity”, “reaction-diffusion”, or “evolution”) rather than their understanding of any particular category of things (such “weights”, “rats”, or “the Holy Roman German Empire”).

This argument is not intended to suggest that systematic observations are meaningless or marginal. On the contrary, natural sciences have been concerned with an extensive, detailed observations of living organisms before it became involved with grand theorising about how they change or maintain their structure. Today, evolutionary studies of any sort are still supposed to occur in association with a dense, ideally atheoretical observation of their subject. Yet, evolution is to be understood as ontologically prevalent to any observation regarding living organisms, because it is evolution that drives the existence of all living organisms rather than the contrary. Even if evolution theory is very much dependant on our systematic observation of the world, it is still its theoretical formulation that inform us most directly on what exist and why.

Cliodynamics : a science of processes

Cultural evolution, as an instance of the natural sciences, is very much defined by this focus of what processes drive observable facts. The main debate in present literature regards the relative importance in shaping human culture of cognitive bias in transmission of representations versus selection of most efficient social forms. In this line, cliodynamics itself is concerned with determining what causal factors are involved in the stability and changes of human institutions, rather than with stating facts about the future in the style of Asimov’s psychohistory.

As an example, we could invoke Turchin, P. (2010). Warfare and the Evolution of Social Complexity: A Multilevel-Selection Approach. Structure and Dynamics, 4(3). https://escholarship.org/uc/item/7j11945r. Turchin, a leading cliodynamicist, discuss at length of what has driven the emergence of large-scale hierarchic societies. In his opinion, it is warfare between ethnic groups that allowed the emergence of more and more intricate webs of strongmen. Indeed, his statistical analysis of historical data showed that proxies of warfare intensity did predict strongly the emergence and size of empires, in particular the geographical proximity of pastoral and agricultural societies and the military use of the horse.

It is important to note that this finding has no clear bearing on the history of any single empire. The recruited epistemological framework, later systematised in the SESHAT project, is only suited to establish causal relationships within structural patterns by testing competing theories of social change against available data. It is true that any single invasion could be prevented by the timely death of some strongman or random strife among his lieutenants. It is true that at some point some empire emerged while there were no horses, and that at some point there were horses but no empire emerged. These isolated facts have in turn no power to falsify a theory that states simply that warfare intensity is one of the structural drivers of political complexity.

If they were transposed to areas where the naturalist epistemology is more established, such claims of falsification would strike anyone as immensely bizarre. Would evolution theory be disproved by a meteorite-caused mass extinction events, which it could not have predicted ? It would not, as the proper domain of evolution theory is to describe how heritability of forms drive changes in living organisms. In the same manner, cliodynamics is by definition the study of structural patterns in history, and disproving theories about structural patterns in history implies to recruit observations of structural patterns in history, rather than merely stating that there are historical contingencies in additions to these patterns.

But why would cliodynamics, rather than any other discipline within cultural evolution, be confronted to repeated claims that contingency forbids study ? Turchin’s “history as science” catchphrase was certainly the most effective in convincing that cliodynamics is driven by tech bros, at the exclusion of historians and anthropologists - especially in institutional contexts where history is classified as social science rather than as humanity. Maybe the temporal nature of history, or the spectre of psychohistory, or the mention of a discipline that is somehow “mathematical” could lead to believe that cliodynamics pretends to hold some objectively true facts about the future of societies.

But when compared to linguistic evolution, cultural attraction, or adaptive theories of human culture, cliodynamics appears to be among subdisciplines of cultural evolution the closest in its thematics to established social sciences. Just like the infamously pseudoscientific neoclassical economic theory, it presents mathematical analysis as a necessary preconditions to empirical or theoretical enquiries, and bears a potential threat to depopulate departments of social science to fill them with computational scientist unable to contest or even conceptualise the status quo. While legitimate, this anticipation fails to account for key features of cliodynamics.

Unlike neoclassical economics, cliodynamics is led by social anthropologists, historians, and other social scientists. As its models are meant to capture properties observable phenomenons, they are generally nothing more than formal translations of preexisting social scientific theories. Even if historical models could stem from pure abstraction, they could never be tested without contextual expertise regarding specific periods, institutions, or social processes. The only significant way in which cliodynamics differs from the social sciences it was born from is in its association to the epistemology of “instrumental prediction” we defined earlier. By federating specialists of seemingly unrelated subjects toward the construction of rather general theories of social change, cliodynamics has succeeded in its (agreeably questionable) founding intention. Importantly, it did not imply to compromise on the political dimension of social scientific research : some of the main results from cliodynamics are far from aligned with the interests of dominant institutions, as for example the diagnostic of an ongoing crisis of the governance, or the association of States and elites to economic exploitation.

In addition, the question of whether natural sciences could replace social sciences any time soon has already been answered, and it is a clear negative. Ogien, A. (2011). Les sciences cognitives ne sont pas des sciences humaines. Une réponse à « Vers un naturalisme social » de Laurence Kaufman et Laurent Cordonier. SociologieS. http://journals.openedition.org/sociologies/3635 has already stated an argument which is decisive in assessing the autonomy of sociology in regard to cognitive science. While he recognise the full authority of cognitive science to define what “cognitive architecture” drives human behaviour, he also remarks that this architecture has no direct bearing on the “epistemic architecture” of socialised individuals’ beliefs and representations - an architecture that cognitive science, unlike sociology, has few tools to address. The argument of process ontology we developed earlier leads this argument to radical extents, as it implies natural sciences cannot claim unsituated knowledge of any single fact - let alone empirical facts of everyday belief it hardly knows how to measure.

Conclusion

Through this piece, we answered some common objections about cliodynamics by stressing how a focus of processes rather than things is a key element of the natural sciences and their reliance on mathematical modelling. In addition to these, we argued the “instrumental prediction” epistemology forbids in principle the ability of natural sciences to detain “truth” regarding any single object, rather than the classes of processes that shape them. Those arguments are indeed general to the natural sciences, and are an account of normative rather than descriptive principle. It is not intended as a definitive delimitation of the scope of cliodynamics (or any other science), validating its results until proof of the contrary. It is intended however to state precisely what cliodynamics is about, and what it is not about.

It is not true that cliodynamics intends to state facts about the future, or that it’d be falsified by the mere existence of contingencies in human societies. It is true the cliodynamics claims to be able to identify dynamical relationships in historical process. It is not true that cliodynamics involve few specialists of social sciences, or that their work is intended to be replaced by math. It is true however that its mainstream can be criticised based on its data collection strategies, its accent on statistical analysis, or its focus on macroscopic patterns rather than mesoscopic scales of analysis.

Beyond these points, it is important to note that cliodynamics is not a field of study in isolation, but is only a subdomain of the more general discipline of cultural evolution which studies patterns of change and stability in human culture. While the emergence of empires is labeled cliodynamics and the emergence of language is not, cultural evolution is all about the interplay between ultimately contingent social structures, and its claims are not meaningful in isolation from each other. For example, Scott, J. C. (2017). Against the Grain: A Deep History of the Earliest States. Yale University Press. underlined how sedentarity and probably patriarchy were essential in the emergence of early States. His claim could be pushed way further, as States arguably recruited social structures previously established to enforce equally social norms within a society in order to do the exact opposite.

The distinction between theories of processes and theories of things, we hope, may help delimitate how cultural evolution theory may reflect on current societies, and how it may not. In the end, its ability to meaningfully inform political action may reduce to what we choose to draw from it, and we hope we will be able to pick wisely. In any case, the cultural evolutionary research program has unmatched potential for studying the historical development of the human species is unmatched, and we are committed to furthering its course.