The Enactive Approach to Agency

Introduction

Explaining the properties and behaviors of organisms leads us to introduce a repertoire of concepts foreign to the world of non-living things. Arguably the most important of these is agency. Agency implies that behaviors are structured by intrinsic goals or purposes, and this brings with it further notions, such as those of norms and functions. The flip side of agency is subjectivity. Having goals implies the presence of a point of view on the world, which imbues the world with values and meaning. This is what it means to be a subject. This also has connections to the idea of sentience and that of first-person experience, aka consciousness (or perhaps proto-consciousness). But I will leave discussion of consciousness for another time,[1] and focus primarily on a minimal notion of agency applicable to the simplest organisms. What is it about life that gives rise to agency?

For me, a satisfying natural explanation would show that agency follows from a distinctive type of causal organization, one which confers a kind of autonomy within the dynamic causal web of the world.  A similar idea is central for theorists who follow what is called the enactive approach to cognition (or simply “enactivism”). Enactivism, a subset of a wider group of approaches that emphasize the importance of embodiment, places particular emphasis on how cognition is rooted in way living things are organized and how they interact with the world. This focus leads them to see the capabilities of simple organisms as relevant to understanding human and animal minds. Below I look at some ideas from enactive theorists about the crucial role of autonomy and causal organization in explaining the origin of agency in living things.[2] Since enactive theorists do not provide a theory of causation itself, I will try to sharpen the discussion at some junctures by referencing my own development of causal process theory.

My approach is based on the notion of a causal process: a basic entity that possesses dispositions toward mutual interactions with other processes (a dispositional profile). These causal interactions produce changes in those processes (an altered profile).  Causal processes and their interactions can be pictured as a kind of web or graph, with the processes as the edges and the interaction events as the vertices. I then introduce the notion of a composite causal process to explain how larger scale individuals with their own causal profile come about.  These composites arise due to patterns of repeated interactions. A composite will form when the members of a group of processes (the composite’s sub-processes) interact with each other at a greater frequency than they do with non-members (see this prior post for more details).

But this is a very general definition: it outlines what is minimally necessary to form a composite system with its own causal powers. As a result, it encompasses every sort of natural individual, both ephemeral and long-lasting. Much more must be said to describe the complex causal patterns that confer highly specific powers on various sorts of composite individuals, such as a living thing.

Autopoiesis

Given that even single-celled organisms are extremely complex, can we say something concise about how their causal organization enables biological autonomy?  One answer, proposed by enactive theorists (beginning with Humberto Maturana and Francisco Varela), is the property of autopoiesis (“self-production”).  Here is a definition:

The autopoietic organization is defined as a unity by a network of productions of components which (i) participate recursively in the same network of productions of components which produced these components, and (ii) realize the network of productions as a unity in the space in which the components exist. Consider for example the case of a cell: it is a network of chemical reactions which produce molecules such that (i) through their interactions generate and participate recursively in the same network of reactions which produced them, and (ii) realize the cell as a material unity. Thus the cell as a physical unity, topographically and operationally separable from the background, remains as such only insofar as this organization is continuously realized under permanent turnover of matter, regardless of its changes in form and specificity of its constituent chemical reactions. (Varela, Maturana & Uribe, 1974, 188)

The cell is a paradigm case, whose self-production of components, featuring its semi-permeable barrier, marks it as a distinct individual in its environment. More recent enactive theorists have supplemented this definition, viewing it for various reasons as not quite sufficient to account for the properties of living things (I will discuss one of these additions below).  But for my purposes, a difficulty is that autopoiesis relies a bit too much on the biochemical example of the cell and I would prefer a more general approach to explicating autonomy if possible. All causal interactions “produce” change; the kind of “production” autopoiesis discusses is a particularly involved undertaking. Must components be continually and completely recreated in a manner akin to that of the complex workings of macromolecule reaction networks in a cell?

Operational Closure

Thankfully, enactive theorists have discussed such a more general notion, using the terms “organizational closure” and “operational closure” to describe the idea.  Di Paolo and Thompson (2014) define an autonomous system as one that is operationally closed and precarious. The idea of operational closure they use (I will return later to the “precarious” requirement) is based on a special kind of relation between components that confers an “enabling condition.” This is a relation whose existence makes not merely some kind of difference to a process, but one whose absence leads to the process ceasing altogether. They then describe a scenario involving a group of processes each of which is both the source of and the recipient of enabling relations with other members of the group (the authors picture the enabling relations themselves as uni-directional). When a network of processes displays this pattern, it has the property of operational closure (see pp. 69-71).

This definition highlights the key point: it is not that the components need to manufacture each other in a complicated way, but that component processes depend for their existence on the network, even as the network obviously only exists in virtue of the presence of the components.  In terms of causal organization, I note that this definition relies on a particular kind of directed difference-making relation, whereas I see the true ontology of causation as involving mutual productive interaction. However, one can view the “enabling relation” idea as a simplification, standing in for a pattern of interaction among component processes that realizes the same result. The idea they are driving toward is that the processes will not persist absent their presence in the network.

This is made clear when the authors say that autonomy also depends on the condition of “precariousness”:

A precarious process is such that, whatever the complex configuration of enabling conditions, if the dependencies on other processes in the operationally closed network are removed, the process necessarily stops. In other words, it is not possible for a precarious process in an operationally closed network to exist on its own in the circumstances created by the absence of the network. (Di Paolo and Thompson, 2014, 72)

If you remove an ion from a crystal or a microchip from a computer, it can find a new home without substantial alteration of its own causal profile. But sub-processes taken out of a biologically autonomous system will not retain their powers.  The network is in a constant dynamic battle against decay and dissolution.

While my discussion has emphasized internal organization, enactive theorists also highlight the distinctive interactions an “autonomous” network must maintain with its surrounding. It must of course exploit environmental resources to sustain itself in their battle:

A precarious, operationally closed system is inherently restless, and in order to sustain itself despite its intrinsic tendencies towards internal imbalance, it requires energy, matter, and relations with the outside world. Hence, the system is not only self-enabling, but also shows spontaneity in its interactions due to a constitutive need to constantly “buy time” against the negative tendencies of its own parts. (Di Paolo and Thompson, 2014, 72)

I think this is an attractive picture. For one thing it highlights what seems to make the organization of living things distinct from the inorganic world. It also helps make it understandable why explaining how such systems first got started—the origin of life—is so tricky. But does this definition of an autonomous system give us enough conceptual resources to account for agency?  The idea is that agency comes into the picture with this interactive dimension of the autonomous system. Because interactions with the environment are connected with the struggle to maintain itself in the face of precariousness, we can claim the “goal” of the system is to maintain its organization, and the “function” of the components is to contribute to this maintenance (see Virenque and Mossio, 2023, p.2).

This is the basic enactivist idea for how agency arises from autonomy. But a skeptic might wonder: could we describe this kind of interacting system without the concept of true (or intrinsic) agency? Might agency be something we merely attribute to the system for our own convenience?

I’ll discuss this further below. One possible ingredient for a solution is highlighted in the last quote when the authors mention “spontaneity”. This needs some further unpacking, and is linked to another enactive concept I want to discuss: that of “adaptivity.” But before turning to that, I want to make a couple more comments about the causal organization discussed in this section.

Brief Digression About Causation and Autonomy

I want to make three brief points regarding possible controversial aspects of this kind of causal organization. First, I note that the operationally closed system exhibits a kind of circularity – this is an oft-noted aspect of living systems. But it is not a paradoxical causal circle: each component process propagates forward in time, and undergoes some change with each interaction. So, there is no case where a cause is truly also its own causal antecedent. This is easier to see in causal process theory as opposed to a difference-making causal model featuring abstract, uni-directional relations. One can picture a spacetime diagram with the evolution of processes moving vertically: the circle is then properly seen as an extended spiral or braid.

Second, I disagree with those who would describe this distinctive organizational scheme as involving “downward causation”. This notion arises because there appears to be a dependency relation between the network and its components that goes in both directions. While the network clearly exists in virtue of its components’ persistent activity, one might also say a component depends on the network for its persistence. But while the former is a fact of causal composition, the latter is just a gloss that can be rephrased: the component causally depends for its persistence on the full set of interactions among all the components (and their surroundings) over time. There are no synchronic causal relations between things at difference scales (or, god forbid, “levels of reality”).

Finally, there is no need to attribute a mysterious or magical kind of “emergence”. The lower-scale components and their pattern of activity compose a larger scale network with its own causal profile. This can be labeled emergence, but it is really a consequence of how patterns of activity among causal processes form composites generally. In my preferred view, the world features a nested hierarchy of composite causal processes with causal interactions naturally occur at multiple scales (please see the last section of this post for more discussion).

Adaptivity

Contemporary enactive theorists posit that living things possess a further feature. The environment surrounding them is constantly changing, and can change in ways that are unprecedented (for not only an individual organism, but its entire lineage). In response, living things exhibit what is called adaptivity:[3]

“Adaptivity” refers to the ability of certain autonomous systems to regulate their operationally closed processes in relation to conditions registered as improving or deteriorating, viable or unviable. (DiPaoli and Thompson 2014, p. 73)

Living systems persist through such environmental changes by modifying themselves and their resulting dispositions toward external interactions. Enactive theorists believe that autopoesis or operational closure needs to be coupled with adaptivity in order to have the kind of autonomy that confers biological agency.^[4] The adaptive system perturbs itself in ways that serve to explore the possible avenues to survive amidst change:

An adaptive act is in itself not a conservation, but a modulatory deviation from an existing tendency that would lead to eventual loss of viability if left unchecked. (DiPaolo, Thompson and Beer, 2022, 21, emphasis original)

Surviving requires more than mere maintenance (even that of maintaining a state far from thermodynamic equilibrium). It requires adaptivity.

Is More Needed?

I’ll finish by revisiting the skeptical stance. The authors I am reading don’t always make this crystal clear, but the idea that adaptive acts are self-generated and thus “spontaneous” (as in the earlier quote) seems critical for establish that we have true intrinsic agency. But where does this capacity for spontaneity come from? It must either result from the system’s distinct form of organization or from capacities inherent in its sub-processes. From what I have read, it doesn’t appear to follow from operational closure per se. It therefore must come from the components. But then one can ask: where do they get it? Either from their lower-scale organization or a capacity of their own components (and so on). So, I think a bit more needs to be said to account for the spontaneity of adaptive activity.

The good news is that I can see how this story might be told, although more work on the details would be helpful. Adaptivity certainly seems to fit with some of the behaviors displayed by simple organisms, e.g. their ubiquitous oscillatory behaviors.[5] While these kind of behaviors can be helpfully captured using classical stochastic or chaotic dynamical models, ultimately, it would seem that living things must exploit and amplify the spontaneous, indeterministic character in their most basic components—that displayed by quantum systems. If this makes sense, then we can see our way to concluding that living things do have the causal character necessary for true naturalized agency.

References

Di Paolo, E., & Thompson, E. (2014). The enactive approach. The Routledge handbook of embodied cognition, 68-78.

Di Paolo, E., Thompson, E., & Beer, R. (2022). Laying down a forking path: Tensions between enaction and the free energy principle. Philosophy and the Mind Sciences, 3, 2.

Hutto, Daniel D. (2023). Enactivism. The Internet Encyclopedia of Philosophy.

Varela, F. G., Maturana, H. R., & Uribe, R. (1974). Autopoiesis: The organization of living systems, its characterization and a model. Biosystems, 5(4), 187-196.

Virenque, L., & Mossio, M. (2023). What is agency? A view from autonomy theory. Biological Theory, 1-5.

Wan, K. Y. (2023). Active oscillations in microscale navigation. Animal Cognition, 1-14.

---

[1] I summarized some of my thoughts about the problem of consciousness in a recent post. The present discussion can be seen as an supplement to that post, particularly the section titled “Subjectivity as a Biological Phenomenon.”

[2] A nice compact overview can be found in this paper: “What is Agency? A View from Autonomy Theory” by Louis Virenque and Matteo Mossio.

[3] Note this is distinct from an evolutionary notion of adaptation; it is a property of individual organisms.

[4] Virenque and Mossio call this “adaptive agency” (Virenque and Mossio, 2023, 3). I note that enactivists also link the idea of adaptivity to “sense-making”, another key concept: “’Sense-making’ describes behavior or conduct in relation to norms of interaction that the system itself brings forth on the basis of its adaptive autonomy. An adaptive autonomous system produces and sustains its own identity in precarious conditions, registered as better or worse, and thereby establishes a perspective from which interactions with the world acquire a normative status (Di Paolo and Thompson, 2014, 73).”

[5] See this great recent article summarizing some of these behaviors: “Active Oscillations in Microscale Navigation” by Kirsty Y. Wan.