On cliodynamic realism

Is there such a thing as the “dynamics” of history?

Cliodynamics, in substance, is just what it says on the package. It is the application of concepts and methods from dynamical systems theory to the study of history (clio). In its contemporary form, it was founded by Goldstone (1991)’s study of the 1789 French revolution. His work led him to articulate a general theory of the dynamical interplay between various structural and demographic factors, the Structural-Demographic Theory (SDT). SDT predicts the existence of 200-300 years pseudo-cycles in political violence, State integration / political instability, which could be traced to empirically verifiable dynamics in demographic growth, elite production, median wage / alimentary insecurity, etc. Later, former evolutionary ecologist Peter Turchin undertook the massive work of reframing SDT (Turchin 2007; Turchin and Hoyer 2023) and testing its validity in a wide array of historically / archaeologically documented cases, ranging from the early Roman Republic to modern Russia (Turchin and Nefedov 2009) - and later the contemporary United States of America (Turchin 2016). In parallel, Turchin attempted to define the broader program of cliodynamics (Turchin 2011), and investigated many relevant questions associated to role of warfare in the development of State societies (Turchin 2009; Turchin et al. 2013). On his impulsion, a wide array of social scientists collaborated to build SESHAT, an extensive database integrating many relevant sources to enable the test of cliodynamic hypothesis (Whitehouse et al. 2016; Brennan et al. 2016; Turchin, Whitehouse, et al. 2018). This led to further studies on the definition of “social complexity” (Turchin, Currie, et al. 2018), the development of organized religion (Turchin et al. 2019), chariot warfare (Turchin et al. 2020), etc.

Although SDT plays a very important role in the history of cliodynamics, its effective definition is not tied to any theory of historical dynamics. Cliodynamics is by construction a methodological commitment to the use of statistical and computational methods to uncover certain patterns in historical dynamics. Although some of his public interventions may lead to the idea of a certain determinism (Turchin 2019), Turchin himself has been very careful to repeatedly clarify that cliodynamics cannot predict the future in the same manner that a futurologist or Asimov’s psychohistory claim they can (Turchin 2012; 2018). It is well known that even deterministic dynamical system escape any practical predictability when they display positive Lyapunov exponents (what is known as the “butterfly effect”), and that real life comes with noise which bars even the abstract possibility of predictability associated to the deterministic case. In other words, any claim of determinism within cliodynamics refers to the discovery of a measurable cause-to-effect relationship, not to the determination of future outcomes based on a sufficient array of information. While we recognize that, we will suggest here that cliodynamics commits what may be the worst sin known to scientific theory: the choice of a slightly sub-optimal ontology. Indeed, all the methodology deployed by cliodynamics works from the assumption that historical dynamics do in fact constitute dynamics, i.e. that their structure follows the formal assumptions of dynamical systems theory (or the associated but implicit assumptions of the specific tools cliodynamics use for inference).

This remains to be established, and we will make the case here that there is strong ground to refute this premise with direct effect on the methodology of cliodynamics. In particular, I will bring into question Turchin’s repeated claim that cliodynamics constitutes “history as a science” (Turchin 2023), as well as the associated ambition to identify “laws of history” - or even underlying mechanisms. This is not to say that I refute the use of mathematics in social science, or the overall relevance of cliodynamic models. After all, model are nothing but useful fictions (Wimsatt 1987), and the approach I will propose may outshine Turchin’s in terms of mathematical fundamentalism. However, I will argue here that the gap between the structure of dynamical system theory (as well as the structure implicitly assumed by cliodynamic methodology) and the structure of historical dynamics fundamentally constrains cliodynamic’s ability to explain or predict the latter. More precisely, I will show that the dynamical structure underlying human societies is transient at best and non-existent at worst, leading to a very weak ground to explain or predict historical dynamics from dynamical structure or statistical inference alone. In my perspective, this should cast doubt regarding any fundamental difference between the scientific status of cliodynamics and this of other established theories of social structure and historical change. Perhaps most importantly, this should motivate us to explore other avenues to investigate the underlying structure of human activity, preferably in a way that remains coherent with established sociological and historical methodology.

Prediction, explanation and ontology in cliodynamics

To properly frame our argument, we need first to establish exactly what methodological commitments underlie the research program of cliodynamics. The core claim of cliodynamics is not that, deep down, historical dynamics are dynamics – in the formal sense defined by dynamical systems theory. In fact, and as we will see soon, the concrete methodology of cliodynamics seems to conflict quite directly with this interpretation. The core claim of cliodynamics is that treating historical dynamics as if they were dynamical system theoretic objects, amenable to the type of modeling and statistical methodologies usually used in the study of complex systems, may lead to useful insight in the form of prediction or explanation. This remains a quite committing statement – what if the proposed methodology fails to provide meaningful prediction? what if historical dynamics behave more like animals, whose activity is more directly amenable to qualitative interpretation? To the credit of cliodynamics, it seems that this approach generally worked. The SDT was applied to many cases on the basis of the best data historical sociology could offer, and this didn’t lead researchers to identify to clear violation of its premises. Many interesting effects regarding the emergence of State societies and their historical developments were demonstrated, and those remained generally coherent with the theoretical explanations cliodynamists put forward. This should lead us to careful optimism about the validity of cliodynamics as a research program, or even to “mild realism” about its theoretical concepts – understood here in Dennett (1991)‘s sense, i.e. the manifest reality of concepts that simply provide a better explanation than their known alternatives.

However, this still fails to meet the standards by which we generally judge a model’s ability to explain and predict. In the context of life sciences, the gold standard of explanation is known as the structural-mechanistic model (Bechtel and Abrahamsen 2005). In substance, it states that a scientific model explains a given phenomenon if and when it adequately captures the structure in the world which brings it about – hopefully, together with a dynamical or computational model which demonstrates how the postulated equations of movement lead to the target phenomenon. This does sound a lot like what cliodynamics set to do. However, a major subtlety lies in the formulation of “adequately captures the structure in the world”. The structural-mechanistic model of explanation requires what we call an ontological commitment, i.e. an active engagement by the theorist to match the structure of its model to the known natural object it aims to represent. This may be implemented by the hardcore realist ontic model of explanation (Craver 2006; 2014), or by the looser standard of coherence (Colombo, Hartmann, and van Iersel 2015). The fact remains: a mechanical model can only be explanatory if we can demonstrate that its ontology (i.e. the objects it postulates) is valid (up to the standard of correspondence or coherence) with its target system. Therefore, assuming we accept the structural-mechanistic standard, cliodynamic model can only be explanatory if historical dynamics actually are dynamical system theoretic objects - or at least behave as such at the adequate scale of observation.

However, were we to look at the specific methodology developed by cliodynamics, we could conclude that the description of history as a dynamical system is more a statement of principle than an actual methodological commitment. In its overwhelming majority, cliodynamics research leverage some database (usually SESHAT or socio-historical data relevant to SDT) to measure patterns in the statistical relation between specific data vectors, without worry to provide a specific dynamical model to explain those patterns 1. Sensu Bechtel and Abrahamsen (2005), this simply constitutes the empirical description of an effect, which (barring circular logic) cannot stand for the explanation of the same effect. We’ll concede this argument is a quite abstract one: maybe cliodynamics does not provide explanation in the strict sense life scientists set out for themselves, but still provides useful insight in the form of prediction. However, we would argue that the goal of predictability requires similar conditions. Of course, the prediction of single trajectory is generally not possible even for actual dynamical systems for the reasons of chaos and stochasticity, and cliodynamics clearly does not commit to that goal. It remains that there is a clear difference of status between failures of prediction which stem from the intrinsic complexity of a given system, and failures of prediction which stem from our lack of knowledge of that system. Even aggregate prediction (in a sense that reduces to the valid inference of some form of causal influence, or dynamical modes in a target system) requires a quite adequate knowledge of the structure of a system.

For example, I may be able to describe quite precisely the trajectory of a ball of the atmosphere (accounting for stochasticity), and provide an adequate probabilistic map of its possible paths for the purpose of prediction. Were I to describe a bird, I would have to account for the fact that it can do as it wants for reasons beyond my understanding, and I would fail to provide such a map. I could of course give it a snack and try to map out how long it takes for the bird to get to it from such and such position, but this would be diverting quite radically my initial goal of predicting its behavior as a dynamical system. And even so, I would fail to integrate many important conditions that are probably more relevant to my predictions than my statistical model: at this moment in time, is the bird hungry? Is it scared by something near the snack, that could lead it to delay its approach or fly away? Does it even like this snack? In this sense, the study of human societies is much like the study of birds. Social structure routinely changes, sometimes within centuries, sometimes within one’s lifetime. Even statistical effects that are measured robustly (or dynamical structure that is adequately inferred) in one context may very well be reversed in some other, or even in the same context shortly after. On this basis, we may wonder whether the ontology of dynamical systems and the methodology of statistical inference is in fact adequate for the study of historical dynamics, and consider what alternatives there are.

Contextuality, dynamical systems, and the architecture of historical change

Mathematically, a dynamical system is defined as a function Ф: (T, X) → X ; where T represents a time dimension 2 and X represents a state space (i.e. the set of possible states of the system). Furthermore, two coherence conditions are imposed: Ф(0T , x) = x, meaning that the system does not move in zero time ; and Ф(t2 , Ф(t1 , x)) = Ф(t2·t1, x), meaning that the trajectory of a system in two consecutive segments of time is the same as the trajectory in the aggregated segment of time. Together, those conditions provide a maximally general representation of systems that have actual states, and which have determined laws of movement. It seems intuitive that human societies, or any other natural system, could in principle be represented in this framework. Of course, their complexity may make it impossible to provide a fully accurate model of their dynamics, but surely no system in nature can have non-actual states or laws of movement that they make up along the way! The problem with intuition is that many systems in nature can, in fact, have non-actual states and/or laws of movement that they make up along the way. For example, quantum systems have states that are brought about (rather than revealed) by measurement, in a manner that is irreducibly constrained by the frame of the measurement. This property is called contextuality, and it reduces formally to non-commutativity in the measurement operators (meaning that two measurements applied in different orders, assuming they do not share a frame, lead to generate two different states in the target system).

Rather than being a simple quirk of physics at infinitesimal scales, contextuality constitutes a fundamental property of human cognition (Guénin–Carlut 2024a) – and a fortiori of sociocultural evolution. Let us consider the simple case of reading. I may learn to read and then look at a page from a book ; or I may look at a page from a book and then only learn to read. It seems obvious that both scenarios lead to different states of the world, one where I have integrated the content of the page and one where I have not. This constitutes a rupture of commutativity, and therefore an instance of contextuality. In a more intuitive manner, we may describe cognitive contextuality as the fact that (more often than not) identifying the relevant dimensions of the world to observe to regulate one’s behavior requires specific skills that are developed throughout an individual’s development. An experienced reader may see esoteric printed symbol as meaningful, and be able to enact the required ocular movement to sample them as efficiently as possible and reconstruct words and sentences in their mind. An experienced climber may see the minute details of a cliff that would enable them to climb, and be able to move in the precise way which enable them to climb it without excessive effort or danger. In a formal sense, such skills correspond to the “context” of observation, and they underlie our experience of the world at its most basic level (Bruineberg and Rietveld 2014; Guénin–Carlut and Albarracin 2024).

Assuming the reader concedes this point, they may still object that it applies in any meaningful way to cliodynamics. The perception of letters or footholds may be reliant on prior skills, but surely social structure exists regardless of our opinion about it! However, social structure is distinctive from letters or footholds in the fact that it has no material instantiation, it purely exists in virtue of the fact that we imagine it to exist. The reproduction of social structure and the manner in which it is integrated and enacted by individual people has been a staple of social sciences for a long time, and is for example present in Goffman’s work on role theory (Goffman 1956) and frame analysis (Goffman 1974). In this line, the “context” of an interaction - as is defined by prior cultural skills, but also by specific communicative acts within the interaction – is precisely what enables the accurate perception of social situations. This is not only an abstract argument about the relation between culture and cognition, but this constitutes a concrete causal driver of cultural evolution. Indeed, contextuality in the reproduction of cultural skills and social structure constitutes a necessary basis to account for the open-ended character nature of sociocultural evolution (Guénin–Carlut 2024b). As elaborated in Heyes (2018)’s work on social learning (see also (Heyes 2012c; 2012a; 2012b; 2016), cultural transmission is itself reliant on prior cultural skills, leaving us with no coherent basis to ground our study of sociocultural evolution on a specific mechanical or dynamical basis.

It appears hard to imagine how a phenomena so fundamentally defined by the permanent change in its causal structure and in the states in brings about, may be explained or represented through inference about the dynamical structure or statistical regularities in historical dynamics. But one may argue that, while contextuality and open-ended evolution are real and important dimensions of sociocultural evolution, they are not the target of cliodynamics. In my opinion, this would be side-stepping the core issue raised by the present argument. By construction, cliodynamics attempts to explain historical change by calling onto a supposedly permanent dynamical structure. This is the central premise that grounds the assumption that the statistical regularities measured by the SESHAT project or Turchin’s work on SDT can be generalized as “laws of history” that can be applied between social contexts and time periods. In the absence of a prior reason to believe in the existence of such a structure, cliodynamics fails to explain historical dynamics (in the structural-mechanist sense of explanation), and fails to provide a strong ground to predict (or post-dict) historical dynamics anywhere but in the specific settings in which it was already validated. However, this argument does not entail that the research program of cliodynamic is irrelevant as a whole, or that we can downplay the possibility of representing or explaining historical dynamics as a physical phenomena. On the contrary, we have proposed elsewhere a quite direct avenue to build an arch-physicalist description of social organization and its evolution, potentially grounding a new kind of cliodynamics: Active Inference.

From Active Inference to public practical premises of social action

Active Inference is a neurocomputational framework describing human cognition as a predictive process, where neuronal activity self-organizes so as to predict upcoming sensations and action (Friston et al. 2017; Parr, Pezzulo, and Friston 2022). Generally, predictive approaches to the mind offer a coherent response to the fundamental problem of perceptive inference, as e.g. described in Varela (1989). Simply speaking, constructing a complete representation of the world by inferring what are the causes of stimuli (i.e. the objects we are trying to perceive) entails an immense computational cost, as well as extreme vulnerability to noise and ambiguity. Hence, it is not adequate to guide the behavior of a living, active creature. By relying on a prior model of how the world affects their sensations, predictive agents can instead circumvent that cost entirely by generating minimal, action-oriented predictions regarding the possible trajectories of their sensations given their choice of action (Clark 2015; Hipólito 2022) - which underlie perception in the predictive model. Active Inference takes a step further by postulating that the same predictive process also drive action, solving similar problems for motor planning (Hipólito et al. 2021). However, the main basis of legitimacy for the Active Inference Framework is its close association with the Free Energy Principle. Initially formulated as an equivalent to what we now call Active Inference (Friston, Kilner, and Harrison 2006; Friston 2010), the Free Energy Principle then evolved into a mathematical theory describing how bare dynamical system may (under specific conditions) self-organize to minimize an informational metric known as Variational Free Energy, functionally implementing Active Inference (Friston 2019; Da Costa et al. 2021; Friston et al. 2022; Parr, Pezzulo, and Friston 2022).

Active Inference overall appears to be a very successful model of the human mind. It uniquely explains away some of the hardest problems know to cognitive science (e.g. the eternal debate around representation (M. J. Ramstead, Kirchhoff, and Friston 2020; Constant, Clark, and Friston 2021)), while seemingly flowing from very basic physical assumption, and remaining coherent with neurophysiological evidence (Walsh et al. 2020). For the purpose of the present argument, we will therefore assume its validity to demonstrate its consequences regarding the social sciences in general and historical dynamics in particular. As outlined in Bruineberg and Rietveld (2014) and Clark (2020), Active Inference entails a dissolution of agency-as-we-know-it. Rather than computing decisions and actions given a neutral representation of the world, Active Inference agents directly perceive the world as a landscape of cultural affordances (i.e. perceived or actual possibility for action) defined by the structure of our environment as well as our developed world- and self-model (M. Ramstead, Veissière, and Kirmayer 2016; Veissière et al. 2020). For example, we do not see a book as a complex array of visual information which we have to process, we see it as something that affords or invites the action of reading – given the context set by our prior skills, our physiological states, and the nature of our self-expectations in a given moment. Guénin–Carlut and Albarracin (2024) describe this phenomena as embedded normativity, in the sense that the normative logic of our action does not simply result from our brain activity but is in part embedded in our material, biological, and social environment – given the context of interaction.

Critically, this applies just as well to the perception of social interaction, and therefore provides useful insight in the cognitive dynamics underlying social structure. Assuming the validity of Active Inference, we may derive the existence of social constraints (Guénin–Carlut 2024b) – causal constraints over human perception and action, born from (or more precisely, dual to) the shared expectation over social behavior embedded in a given cultural context. The deep integration of social expectations in cognition entailed by this model may suggest an extremely structural account of social activity, overall coherent with the permanence in dynamical structure assumed by cliodynamics. However, this would be discounting the role of contextuality in the enaction and integration of social constraints. Indeed, social constraints (as conceptualized in Guénin–Carlut 2024b) are by construction a creature of context. The same material or social environment may be understood by different agents in different ways given their respective context, leading to the application of different norms ; perhaps most importantly, such constraints can only be integrated by reconstruction, and may find themselves having a quite different meaning for different agents acting from different contexts. For example, a man may relate to patriarchal norms quite differently than a woman does, leading to a different context of action within what is ostensibly the same system of constraints. The contextual, reconstructive nature of social constraints enable them to be mapped quite directly to the causal structure underlying historical dynamics, but at the same time it prevents their description in terms of dynamical system and requires the use of more esoteric mathematics (Guénin–Carlut 2024b; 2024a).

While it is quite strictly grounded in the known physics of human cognition, this framework may seem too abstract to provide useful insight into social or historical dynamics. However, despite their formal motivations, those concepts map quite directly onto existing methodologies within ethnography and pragmatic sociology. In particular, they generalize the intuition of a linguistic “common ground” (Stalnaker 2002), of “public practical premises” for social action (Táíwò 2018), or perhaps more generally the symbolic interactionist framework developed by (Goffman 1956; 1956). The common thread of those approaches is they postulate humans generally accept certain norms (be they explicit linguistic premises, or implicit / embodied attitudes) as a prior while engaging with given social contexts. Such norms act as the “structure” of this context, but also provide an open-ended context that different individual or groups may attempt to negotiate or bend to their advantage. This framework naturally lends itself to an ethnographic approach, which is tuned to map the accepted structure within a given community, but also the situation of different actors within this structure and the potential coalitions that may emerge to renegotiate its frame. In particular, I have demonstrated its relevance to study the manner in which the Unification Church builds social influence (Guénin–Carlut and Dubourg 2024). By construction, this approach lends itself to document the type of transient structure which underlie human social activity. This provides a much stronger mechanistic grip on what dynamics actually shape a given context, but also precludes the possibility of long-term prediction given the open-ended nature of linguistic (or epistemic) dynamics.

Conclusion

Our argument here reduces to the following: cliodynamics attempt to study human history as a dynamical system, which it is not. Worse, a critical driver of human sociocultural evolution is contextuality, i.e. the fact that social structure is actively brought about by human engagement rather than a independent structure uncovered by social activity. This is precisely what enables human social structure to evolve in time, and therefore generates the phenomena that cliodynamics attempt to study. However, contextuality (and a fortiori open-ended evolution) is by construction antagonistic to the framing of dynamical systems theory, which is built to describe the well-defined trajectories of well-defined states in well-defined spaces. Of course, cliodynamics does not blindly subscribe to the dynamical systems ontology, and in fact tends to be more focused on measuring statistical patterns in historical data. However, this does not brighten the perspective on the field brought by our argument. Unlike dynamical modeling, simple statistical association does not provide a strong ground to “explain” natural dynamics, for lack of any explicit description of the underlying causal structure. Furthermore, any predictive ambition of this methodology relies critically on the assumption of a permanent underlying causal structure, which we know (because of the role of contextuality in cultural reproduction, as well as the documented open-ended nature of sociocultural evolution) is not verified. Together, these arguments lead to a bleak picture of cliodynamics as a research program: it simply fails to be a natural science, or at least a science that provides anything like naturalistic explanation of its target phenomena, and its aptitude to prediction is at best ill-grounded.

However, in my opinion, such a one-sided picture fails to convey the adequate context to judge the relevance of cliodynamics as a research program. We have exposed in part 1 that the generally accepted mode of explanation in the natural sciences is the stuctural-mechanistic model, where a mathematical or computational model aims to map the actual causal structure underlying the target phenomena. We have also established in part 2 and 3 that social structure is by nature transient, and (perhaps most importantly) contextual – which means, by definition, that it does not have an “actual structure” that may be represented. This state of affairs is, in my opinion, well integrated in the epistemology of social science. In particular, Bhaskar (2014) has introduced a variant of the structural-mechanistic model of explanation which accounts for these properties of social phenomena: critical naturalism. In essence, critical naturalism acknowledge the existence of “real mechanisms” driving social phenomena, while also aiming to account for their contextual and transient nature in its practical approach to scientific inquiry and explanation. I consider this approach to be, whether explicitly or not, a de facto standard in social science. On this basis, it is hard to demonstrate any specificity of cliodynamics as compared to other approaches in social sciences. Like any other approach, it attempts to investigate some form of causal structure, defined by its specific ontology and methodology ; like any other approach, it is vulnerable to the dissolution of this structure through routine mechanisms of sociocultural evolution.

Therefore, we cannot invalidate the research program defining cliodynamics ; but neither can we accept at face value its claim to constitute an uniquely “scientific” approach to human history. If we were to attempt the articulation of a purely naturalistic approach to the social sciences, my suggestion would be to stick to the known physics of human cognition and attempt to reconstruct what social phenomena we can on this basis. This is the spirit of the approach we have articulated in part 3, which constitutes in essence a reframing of existing ethnography methodology into the neurocomputational framework of Active Inference. This approach lends itself to modeling just as well as cliodynamics does, for example through the description of semantic networks (Axelrod 2015) or their generalization as (embodied, enactive) Bayesian networks within Active Inference (Friston, Parr, and Vries 2017). Most importantly, it enables (at least in principle) a dual account of social structure and of the situation of various agents within this structure, therefore opening the way to account for its diachronic change. While this approach lends itself more directly to a naturalistic kind of explanation, it does so at the expense of its predictive power. Indeed, given the endless possibilities of recomposing different public practical premises for public action and the contextual (hence irreversible) nature of sociocognitive dynamics, any system may at any point evolve in an arbitrary number of directions. If reasonable predictions may be formed on this basis, they still fall well short of constituting anything like cliodynamic “laws of history”.

Bibliography

• Axelrod, Robert. 2015. Structure of Decision: The Cognitive Maps of Political Elites. Princeton University Press.
• Bechtel, William, and Adele Abrahamsen. 2005. “Explanation: A Mechanist Alternative.” Studies in History and Philosophy of Science Part C: Studies in History and Philosophy of Biological and Biomedical Sciences, Mechanisms in biology, 36 (2): 421–41. https://doi.org/10.1016/j.shpsc.2005.03.010.
• Bhaskar, Roy. 2014. The Possibility of Naturalism: A Philosophical Critique of the Contemporary Human Sciences. Routledge.
• Brennan, Rob, Kevin Feeney, Gavin Mendel-Gleason, Bojan Bozic, Peter Turchin, Harvey Whitehouse, Pieter Francois, Thomas E. Currie, and Stephanie Grohmann. 2016. “Building the Seshat Ontology for a Global History Databank.” In The Semantic Web. Latest Advances and New Domains, edited by Harald Sack, Eva Blomqvist, Mathieu Aquin, Chiara Ghidini, Simone Paolo Ponzetto, and Christoph Lange, 9678:693–708. Lecture Notes in Computer Science. Cham: Springer International Publishing. https://doi.org/10.1007/978-3-319-34129-3_42.
• Bruineberg, Jelle, and Erik Rietveld. 2014. “Self-Organization, Free Energy Minimization, and Optimal Grip on a Field of Affordances.” Frontiers in Human Neuroscience 8. https://doi.org/10.3389/fnhum.2014.00599.
• Clark, Andy. 2015. Surfing Uncertainty: Prediction, Action, and the Embodied Mind. Oxford University Press.
• ———. 2020. “Beyond Desire? Agency, Choice, and the Predictive Mind.” Australasian Journal of Philosophy 98 (1): 1–15. https://doi.org/10.1080/00048402.2019.1602661.
• Colombo, Matteo, Stephan Hartmann, and Robert van Iersel. 2015. “Models, Mechanisms, and Coherence.” The British Journal for the Philosophy of Science 66 (1): 181–212. https://doi.org/10.1093/bjps/axt043.
• Constant, Axel, Andy Clark, and Karl Friston. 2021. “Representation Wars: Enacting an Armistice Through Active Inference.” Frontiers in Psychology 11:3798. https://doi.org/10.3389/fpsyg.2020.598733.
• Craver, Carl F. 2006. “When Mechanistic Models Explain.” Synthese 153 (3): 355–76. https://doi.org/10.1007/s11229-006-9097-x.
• ———. 2014. “The Ontic Conception of Scientific Explanation.” Explanation in the Biological and Historical Sciences. Dordrecht, Netherlands: Springer, 27–52.
• Da Costa, Lancelot, Karl Friston, Conor Heins, and Grigorios A. Pavliotis. 2021. “Bayesian Mechanics for Stationary Processes.” Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences 477 (2256): 20210518. https://doi.org/10.1098/rspa.2021.0518.
• Dennett, Daniel C. 1991. “Real Patterns.” The Journal of Philosophy 88 (1): 27–51. https://doi.org/10.2307/2027085.
• Friston, Karl. 2010. “The Free-Energy Principle: A Unified Brain Theory?” Nature Reviews Neuroscience 11 (2): 127–38. https://doi.org/10.1038/nrn2787.
• ———. 2019. “A Free Energy Principle for a Particular Physics.” arXiv:1906.10184 [q-Bio], June. http://arxiv.org/abs/1906.10184.
• Friston, Karl, Lancelot Da Costa, Noor Sajid, Conor Heins, Kai Ueltzhöffer, Grigorios A. Pavliotis, and Thomas Parr. 2022. “The Free Energy Principle Made Simpler but Not Too Simple.” arXiv:2201.06387 [Cond-Mat, Physics:Nlin, Physics:Physics, q-Bio], January. http://arxiv.org/abs/2201.06387.
• Friston, Karl, Thomas FitzGerald, Francesco Rigoli, Philipp Schwartenbeck, and Giovanni Pezzulo. 2017. “Active Inference: A Process Theory.” Neural Computation 29 (1): 1–49. https://doi.org/10.1162/NECO_a_00912.
• Friston, Karl, James Kilner, and Lee Harrison. 2006. “A Free Energy Principle for the Brain.” Journal of Physiology-Paris, Theoretical and Computational Neuroscience: Understanding Brain Functions, 100 (1): 70–87. https://doi.org/10.1016/j.jphysparis.2006.10.001.
• Friston, Karl, Thomas Parr, and Bert de Vries. 2017. “The Graphical Brain: Belief Propagation and Active Inference.” Network Neuroscience. https://doi.org/10.1162/NETN_a_00018.
• Goffman, Erving. 1956. The Presentation of Self in Everyday Life. Penguin Books, Limited.
• ———. 1974. Frame Analysis: An Essay on the Organization of Experience. Harvard University Press.
• Goldstone, Jack A. 1991. “Ideology, Cultural Frameworks, and the Process of Revolution.” Theory and Society 20 (4): 405–53. https://doi.org/10.1007/BF00157321.
• Guénin–Carlut, Avel. 2024a. “Contextuality, Cognitive Engagement, and Active Inference.” OSF. https://doi.org/10.31219/osf.io/wtrgh.
• ———. 2024b. “From the Existential Stance to Social Constraints - How the Human Mind Becomes Embedded in Our Social, Cultural and Material Context.” OSF. https://doi.org/10.31219/osf.io/3kyzt.
• Guénin–Carlut, Avel, and Mahault Albarracin. 2024. “On Embedded Normativity an Active Inference Account of Agency Beyond Flesh.” In Active Inference, edited by Christopher L. Buckley, Daniela Cialfi, Pablo Lanillos, Maxwell Ramstead, Noor Sajid, Hideaki Shimazaki, Tim Verbelen, and Martijn Wisse, 91–105. Communications in Computer and Information Science. Cham: Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-47958-8_7.
• Guénin–Carlut, Avel, and Isaac Dubourg. 2024. “Power, Ritual, and Identity - A Cognitive Study of Conversion to the Unification Church.” OSF. https://doi.org/10.31219/osf.io/wg36c.
• Heyes, Cecilia. 2012a. “Grist and Mills: On the Cultural Origins of Cultural Learning.” Philosophical Transactions of the Royal Society B: Biological Sciences 367 (1599): 2181–91. https://doi.org/10.1098/rstb.2012.0120.
• ———. 2012b. “New Thinking: The Evolution of Human Cognition.” Philosophical Transactions of the Royal Society B: Biological Sciences 367 (1599): 2091–96. https://doi.org/10.1098/rstb.2012.0111.
• ———. 2012c. “What’s Social about Social Learning?” Journal of Comparative Psychology 126 (2): 193–202. https://doi.org/10.1037/a0025180.
• ———. 2016. “Who Knows? Metacognitive Social Learning Strategies.” Trends in Cognitive Sciences 20 (3): 204–13. https://doi.org/10.1016/j.tics.2015.12.007.
• ———. 2018. Cognitive Gadgets: The Cultural Evolution of Thinking. Harvard University Press.
• Hipólito, Inês. 2022. “Cognition Without Neural Representation: Dynamics of a Complex System.” Frontiers in Psychology 12:5472. https://doi.org/10.3389/fpsyg.2021.643276.
• Hipólito, Inês, Manuel Baltieri, Karl Friston, and Maxwell James Ramstead. 2021. “Embodied Skillful Performance: Where the Action Is.” Synthese, January. https://doi.org/10.1007/s11229-020-02986-5.
• Parr, Thomas, Giovanni Pezzulo, and Karl Friston. 2022. Active Inference: The Free Energy Principle in Mind, Brain, and Behavior. Cambridge, MA, USA: MIT Press.
• Ramstead, Maxwell James, Michael D Kirchhoff, and Karl Friston. 2020. “A Tale of Two Densities: Active Inference Is Enactive Inference.” Adaptive Behavior 28 (4): 225–39. https://doi.org/10.1177/1059712319862774.
• Ramstead, Maxwell, Samuel PL Veissière, and Laurence J. Kirmayer. 2016. “Cultural Affordances: Scaffolding Local Worlds through Shared Intentionality and Regimes of Attention.” Frontiers in Psychology 7:1090.
• Stalnaker, Robert. 2002. “Common Ground.” Linguistics and Philosophy 25 (5/6): 701–21.
• Táíwò, Olúfẹ́mi O. 2018. “The Empire Has No Clothes.” Disputatio. https://doi.org/10.2478/disp-2018-0007.
• Turchin, Peter. 2007. War and Peace and War : The Rise and Fall of Empires.
• ———. 2009. “A Theory for Formation of Large Empires*.” Journal of Global History 4 (2): 191–217. https://doi.org/10.1017/S174002280900312X.
• ———. 2011. “Toward Cliodynamics – an Analytical, Predictive Science of History” 2:21.
• ———. 2012. “Psychohistory and Cliodynamics.” Peter Turchin (blog). September 3, 2012. https://peterturchin.com/cliodynamica/psychohistory-and-cliodynamics/.
• ———. 2016. “Ages of Discord.” Chaplin, CT: Beresta Books.
• ———. 2018. “The Ginkgo Model of Societal Crisis.” Peter Turchin (blog). August 16, 2018. http://peterturchin.com/cliodynamica/the-ginkgo-model-of-societal-crisis/.
• ———. 2019. “A History of the Near Future: What History Tells Us about Our Age of Discord.” Peter Turchin (blog). November 6, 2019. https://peterturchin.com/cliodynamica/a-history-of-the-near-future-what-history-tells-us-about-our-age-of-discord/.
• ———. 2023. “Cliodynamics: History as Science.” Peter Turchin (blog). May 17, 2023. https://peterturchin.com/cliodynamics-history-as-science/.
• Turchin, Peter, Thomas E. Currie, Edward A. L. Turner, and Sergey Gavrilets. 2013. “War, Space, and the Evolution of Old World Complex Societies,” 2013.
• Turchin, Peter, Thomas E. Currie, Harvey Whitehouse, Pieter François, Kevin Feeney, Daniel Mullins, Daniel Hoyer, et al. 2018. “Quantitative Historical Analysis Uncovers a Single Dimension of Complexity That Structures Global Variation in Human Social Organization.” Proceedings of the National Academy of Sciences 115 (2): E144–51. https://doi.org/10.1073/pnas.1708800115.
• Turchin, Peter, and Daniel Hoyer. 2023. “Empirically Testing and Refining Structural Demographic Theory: A Methodological Guide.” OSF. https://doi.org/10.31235/osf.io/yrqw5.
• Turchin, Peter, Daniel Hoyer, Andrey Korotayev, Nikolay Kradin, Sergei Nefedov, Gary Feinman, James S. Bennett, Pieter Francois, and Harvey Whitehouse. 2020. “Rise of the War Machines: Charting the Evolution of Military Technologies from the Neolithic to the Industrial Revolution.” SocArXiv. https://doi.org/10.31235/osf.io/u5wba.
• Turchin, Peter, and Sergey A. Nefedov. 2009. Secular Cycles.
• Turchin, Peter, Harvey Whitehouse, Pieter François, Daniel Hoyer, Abel Alves, John Baines, David Baker, et al. 2018. “An Introduction to Seshat: Global History Databank.” Journal of Cognitive Historiography 5 (1–2): 115–23. https://doi.org/10.1558/jch.39395.
• Turchin, Peter, Harvey Whitehouse, Pieter François, Daniel Hoyer, Selin Nugent, Jennifer Larson, Alan Covey, et al. 2019. “Explaining the Rise of Moralizing Religions: A Test of Competing Hypotheses Using the Seshat Databank.” Preprint. SocArXiv. https://doi.org/10.31235/osf.io/2v59j.
• Varela, Francisco J. 1989. Invitation aux sciences cognitives. Seuil.
• Veissière, Samuel P. L., Axel Constant, Maxwell James Ramstead, Karl Friston, and Laurence J. Kirmayer. 2020. “Thinking through Other Minds: A Variational Approach to Cognition and Culture.” Behavioral and Brain Sciences 43. https://doi.org/10.1017/S0140525X19001213.
• Walsh, Kevin S., David P. McGovern, Andy Clark, and Redmond G. O’Connell. 2020. “Evaluating the Neurophysiological Evidence for Predictive Processing as a Model of Perception.” Annals of the New York Academy of Sciences 1464 (1): 242–68. https://doi.org/10.1111/nyas.14321.
• Whitehouse, H., P. Francois, J. G. Manning, R. Brennan, T. Currie, K. Feeney, and P. Turchin. 2016. “A Macroscope for Global History - Seshat Global History Databank: A Methodological Overview.” Digital Humanities Quarterly. https://ora.ox.ac.uk/objects/uuid:5b10c64e-51ae-4ba0-8766-a1bea932c805.
• Wimsatt, William C. 1987. “False Models as Means to Truer Theories.” Neutral Models in Biology, 23–55.