Monday, May 31, 2021

Composing Natural Systems

An interesting feature of Relational Quantum Mechanics (RQM) is its implication that discrete measurement-like interaction events are going on between natural systems (unobserved by us) all the time.  It turns out that this offers a way to incorporate quantum phenomena into an attractive account of how smaller natural systems causally compose larger ones.  In this post I will discuss the general approach, including a brief discussion of its implications for the ideas of reduction and emergence. In a follow-up post, I will discuss the quantum case in more detail with a focus on molecules.

Composite Causal Processes

The ontological framework I’m using (discussed in the last section of the prior post) is a modified version of Wesley Salmon’s causal process account (Salmon, 1984). The basic entities are called causal processes, and these comprise a network characterized by two dimensions of causation, called propagation and production. Propagation refers to the way an isolated causal process bears dispositions or propensities toward potential interactions with other processes--aka its disposition profile. Production refers to how these profiles are altered in causal interactions with each other (this is the mutual manifestation of the relevant dispositions).

The entities and properties described by science correspond to features of this causal web. For example, an electron corresponds to a causal process, and its properties describe its dispositions to produce change in interactions with other systems.

Given this picture, we can go on to form an account of how composite causal processes are formed.  What is exciting about the resulting view is that it can provide a framework for systems spanning the microscopic-macroscopic divide.

For background, I note that neither Salmon nor others who have explored causal process views provide a detailed account of composition. Recall that Salmon’s intent was to give a causal theory in service of underpinning scientific explanations.  In this context, he did outline a pertinent distinction between etiological explanations and constitutive explanations. Etiological explanations trace the relevant preceding processes and interactions leading up to a phenomenon. A constitutive explanation, on the other hand, is one that cites the interactions and processes that compose the phenomenon:

A constitutive explanation is thoroughly causal, but it does not explain particular facts or general regularities in terms of causal antecedents. The explanation shows, instead, that the fact-to-be-explained is constituted by underlying causal mechanisms. (Salmon, 1984, 270)

However, while Salmon sketches how one would divide a causal network into etiological and constitutive elements, he doesn’t provide a recipe for marking off the boundaries that define which processes/interactions are “internal” to what is to be explained by the constitutive explanation (see Salmon 1984, p. 275).

Going beyond Salmon, and drawing on the work of others, we can offer an account of composition for causal processes.  They key idea is to propose that a coherent structure at a higher scale arises from patterns of repeated interactions at a lower scale. We should pick out composite causal processes and their interactions by attending to such patterns at the lower scale.

In Herbert Simon’s discussion of complex systems, he notes that complexity often “takes the form of hierarchy (Simon, 1962, 468)” and notes the role interactions play in this context:

In hierarchic systems we can distinguish between interactions among subsystems, on the one hand, and the interactions within subsystems—that is, among the parts of those subsystems—on the other. (Simon, 1996, p.197, emphasis original)

The suggestion to take from this is that differential interaction rates give rise to a hierarchy of causal processes. When a group of processes interacts more with each other than with “outsiders” then it can form a composite. For example, a social group like a family or a business can be marked off from others (at a first approximation) by the differential intensity with which its members interact within vs. outside the group.

As part of his discussion of analyzing complex systems, Bill Wimsatt also explores the idea of decomposition based on interactions, i.e., breaking down a system into subsystems based on the relative strength of intra vs extra-system interactions. (Wimsatt, 2007, 184-6).  And while he describes how different theoretical concerns lead us to utilize a variety of analytical strategies, Wimsatt makes it clear that patterns of causal connections are the ultimate basis for understanding complex systems:

Ontologically, one could take the primary working matter of the world to be causal relationships, which are connected to one another in a variety of ways—and together make up patterns of causal networks…Under some conditions, these networks are organized into larger patterns that comprise levels of organization (Wimsatt, 2007, 200, emphasis original).[1]

Wimsatt explains that levels of organization are “compositional levels”, characterized by hierarchical part-whole relations (201). This notion of composition includes not just the idea of parts, but of parts engaged in certain patterns of causal interactions, consistent with the approach to composite causal processes suggested above.

To summarize: a composite causal process consists of two or more sub-processes (the constituting group) that interact with a greater frequency than each does with other processes.  Just like any causal process, a composite process carries its own disposition profile: here the pattern of interacting sub-processes accounts for how composite processes will themselves interact (what this means for the concepts of reduction and emergence will be discussed below). Consider social groups again, perhaps taking the example of smaller, pre-industrial societies. Each may have its own distinctive dispositions to mutually interact with other, similarly sized groups (e.g., to share a resource, trade, or to engage in raids or battle). These would be composed from the dispositions of their constituent members as they are shaped in the course of structured patterns of in-group interaction. We can also envision here that the higher scale environmental interactions also impact the evolution of the composite entity, but its stability is due to maintaining its characteristic higher-frequency internal processes.

Let me add a couple of further comments about composite processes.  First, as already indicated, a group of constituting sub-processes may be themselves composite, allowing for a nested hierarchy. Second, the impact of larger scale external interactions can vary.  Some may have negligible impact. Other interactions (especially if regular in nature) can contribute to shaping the ongoing nature of the composite. At the other extreme, there will be some external interactions that could disrupt or destroy it. The persistence of a composite would seem to require a certain robustness in the internal interaction pattern of its components. Achieving stability (and the associated ability to propagate a characteristic higher scale disposition profile) may require the differential between intra-process and extra-process interactions to be particularly high, or else there may need to have a particular pattern to the repeated interactions.  There will clearly be vague or boundary cases as well.

Why go to all this trouble of fairly abstract theorizing about a web of causal processes?  Because this account fleshes out the notions that underwrite the causal explanations scientists formulate in a variety of domains.

In the physical sciences, the familiar hierarchy of entities, including atoms, molecules, and condensed matter, all correspond to composite causal processes. Of course, in physical models, what marks out a composite system might be described in a number of ways (for example, in terms of the relative strength of forces or energy-minimizing equilibrium configurations).  But I argue this is consistent with the key being the relative frequency of recurring discrete interactions in-system vs. out-system. (This will be explored further in the companion post.)

In biology, the complexity of systems may sometimes defy the easy identification of the boundaries of composites. Also, a researcher’s explanatory aims will sometimes warrant taking different perspectives on phenomena. In these cases, scientists will describe theoretical entities that do not necessarily follow a simple quantitative accounting of intra-process vs. extra-process interactions. On the one hand, the case of a cell provides a pretty clear paradigm case meeting the definition of a composite process. On the other hand, many organisms and groups of organisms present difficult cases that have given rise to a rich debate in the literature regarding biological individuality. Still, a causal account of constitution is a useful starting point, as noted here by ElliottSober:

The individuality of organisms involves a distinction between self and other—between inside and outside. This distinction is defined by characteristic causal relations. Parts of the same organism influence each other in ways that differ from the way that outside entities influence the organism’s parts. (Sober, 1993, 150)

The way parts “influence each other”, of course, might involve considerations beyond a mere quantitative view of interactions, and connotes an entry point where theoretical concerns can create distance from the basic conception of the composite causal process. In a biological context, sub-processes and interactions related to survival and reproduction may, for example, receive disproportionate attention in creating boundaries around composite entities. Notably, Roberta Millstein has proposed a definition of a biological population based on just this kind of causal interaction-based concept (Millstein 2009).

It is also worth mentioning that constitutive explanations in science will rarely attempt to explain the entire entity. This would mean accounting all of its causal properties (aka its entire dispositional profile) in terms of its interacting sub-processes. It is more common for a scientific explanation to target one property corresponding to a behavior of interest (corresponding to one of many features of a disposition profile).

Reduction and Emergence

I want to make a few remarks about how this approach to composites sheds light on the topics of ontological reduction and emergence. In a nutshell, the causal composition model discussed here gives a straightforward account of these notions that sidesteps some common confusions and controversies, such as the “causal exclusion problem.”

When considering the relationship between phenomena characterized at larger scales and smaller ones, the key observation is that a larger entity’s properties do not only depend not only on the properties of smaller composing entities. They also depend on their pattern of interaction.  This is in contrast to the usual static framing that posits a metaphysical relationship (whether expressed in terms of composition or “realization”) between higher-level properties and lower-level properties at some instant of time.  This picture is conceptually confused (if taken seriously as opposed to a being a deliberate simplifying idealization): there is no reason to think such synchronic relationships characterize our world.

Recall that, in the present account, a property describes a regular feature of the disposition profile of a causal process. A composite causal process is made up of a pattern of interacting sub-processes.  The disposition profiles of the sub-processes are changing during these interactions: they are not static.  The dispositions of the composite depend on this matrix of changing sub-processes.  Note that both the forming of a higher-scale disposition (and its manifestation in a higher-scale interaction) takes more time than the equivalents at the smaller scale.  No composite entity or property exists at an instant:  this is a fiction concocted by us facilitate our understanding. Unfortunately, contemporary metaphysicians have taken this notion seriously.  It is perhaps easiest to see the problem in the case of a biological system:  nothing is literally “alive” at an instant of time.  Living things are sustained by temporally extended processes.  Less intuitively, the same is true of inanimate objects.

Emergence and reduction are clearer, unmysterious notions when based on this dynamic conception of the composition relationship.  Properties of larger things “emerge” from the interacting group of smaller things. The “reduction base” includes the interaction pattern of the components and their (changing) properties.  The exclusion problem says that since higher-level properties are realized by lower-level properties at any arbitrary instant of time, they cannot have causal force of their own (on pain of overdetermination). We can see why this is a pseudo-problem once a better understanding of composition is in place. Causal production occurs at multiple scales.

This take on reduction and emergence is obviously not unique to the causal process model discussed here.  It is implied by any approach that recognizes that properties of composites depend on interacting parts. For example, Wimsatt discusses at some length how notions of reduction and emergence should be understood given his understanding of complex systems. He offers a definition of reductive explanation that shows a similarity to the causal process view of constitutive explanation:

A reductive explanation of a behavior or a property of a system is one that shows it to be mechanistically explicable in terms of the properties of and interactions among the parts of the system. (Wimsatt, 207, 275)

This approach to reductive explanation is perfectly consistent with a form of emergence, in the sense that the properties of the whole are intuitively “more than the sum of its parts (277).” The key idea here, again, is that composition includes the interactions between the parts. For comparison, Wimsatt introduces the notion of “aggregativity”, where the properties of the whole are “mere” aggregates of the properties of its parts. For this to happen, “the system property would have to depend on the parts’ properties in a very strongly atomistic manner, under all physically possible decompositions (277-280)”. He analyzes the conditions needed for this to occur and concludes they are nearly never met outside of the case of conserved quantities in (idealized) physical theories.

Simon had introduced similar notions, describing hypothetical idealized systems where there are no interactions between parts as “decomposable,” which are then contrasted to “nearly decomposable systems, in which the interactions among the subsystems are weak but not negligible (Simon, 1996, 197, emphasis original).” To highlight this distinguishing feature, Simon considers a boundary case: that of gases. Ideal gases, which assume interactions between molecules are negligible, are, for Simon, decomposable systems. In the causal process account, we would similarly point out that an ideal gas doesn’t have a clearly defined constituting group: the molecules do not have a characteristic pattern of interacting with each other at any greater frequency than they do with the external system (the container). An actual, non-ideal gas, on the other hand, with weak but non-negligible interactions between constituent molecules, would correspond to the idea of a composite causal process.

Some contemporary work in metaphysics, focused on dispositions/powers and their role in causation, has incorporated similar views about composition and emergence. Rani Lill Anjum and Stephen Mumford describe a “dynamic view” of emergence:

 The idea is that emergent properties are sustained through the ongoing activity; that is, through the causal process of interaction of the parts. A static instantaneous constitution view wouldn't provide this (Anjum & Mumford 2017, 101)

In their view, higher scale properties are emergent because they depend on lower-level parts whose causal properties are undergoing transformation as they interact, consistent with the view discussed here.  Most recently, R. D. Ingthorsson's new book, while not discussing emergence and reduction explicitly, also presents a view of composition based on the causal interaction of parts which is in the same spirit (Ingthorsson, 2021, Ch. 6).


I think composite causal processes provide a good framework for understanding how natural systems are constituted.  A puzzle for the view, however, might arise via its use of patterns of discrete causal interactions to define composites. How would this work in physics, where the forces binding together composites, such as the Coulomb (electrostatic) force, are continuous?  One possible answer is to point out that physical models employ idealizations, and claim their depictions can still correspond to the “deeper” ontological picture of causal processes. But I believe we can find a better and more comprehensive answer than this.  To do so, we must look more carefully at physical accounts of nature’s building blocks, atoms and molecules, and see if we can uncover a correspondence with the causal theory. I think we can, assuming we utilize the RQM interpretation.  This is the subject of the next post.


Anjum, R., & Mumford, S. (2017). Emergence and Demergence. In M. Paolini Paoletti, & F. Orilia (Eds.), Philosophical and Scientific Perspectives on Downward Causation (pp. 92-109). New York: Routledge.

Ingthorsson, R.D. (2021). A Powerful Particulars View of Causation. New York: Routledge.

Millstein, R. L. (2009). Populations as Individuals. Biological Theory, 4(3), 267-273.

Salmon, W. (1984). Scientific Explanation and the Causal Structure of the World. Princeton: Princeton University Press.

Simon, H. (1962). The Architecture of Complexity. Proceedings of the American Philosophical Society, 106(6), 467-482.

Simon, H. A. (1996). The Sciences of the Artificial (3rd ed.). Cambridge, MA: MIT Press.

Sober, E. (1993). Philosophy of Biology. Boulder: Westview Press.

Wimsatt, W. C. (2007). Re-Engineering Philosophy for Limited Beings. Cambridge, Massachusetts: Harvard University Press.

Photo: Alina Grubnyak via Unsplash

[1] This passage goes on to mention other, less neat, network patterns: “Under somewhat different conditions they yield the kinds of systematic slices across which I have called perspectives. Under some conditions they are so richly connected that neither perspectives nor levels seem to capture their organization, and for this condition, I have coined the term causal thickets (Wimsatt, 2007, 200).”