Reflections on the Naturalization of Relevance Realization

Dialoging with Tim Jackson in response to a new paper by Jaeger, Riedl, Djedovic, Vervaeke, and Walsh

The article "Naturalizing Relevance Realization: Why Agency and Cognition are Fundamentally Not Computational" (Front. Psychol., 24 June 2024) attempts to articulate a theory of relevance realizing organismic agents that is fully naturalistic. The video above records Tim and I sharing our reactions after reading it.

Below are some of my notes written before our conversation:

Jaeger, Riedl, Djedovic, Vervaeke, and Walsh argue that the ability to realize relevance is observable in all living organisms, from bacteria to humans. But despite being perfectly natural, organismic relevance realization transcends formalization. Computational models may partially simulate some aspects of it, but they can never realize or fully instantiate this core competency of living beings.

To explain the nature of relevance realization, the authors present a "trialectic" or triadic dialectic, which encompasses metabolic, ecological, and evolutionary dynamics. These dynamics are involved in the realization of relevance through a meliorative process that allows agents to maintain and optimize their grip on their environments, achieving a “transjective” agent-arena relationship. This relationship is described as a form of “embodied ecological rationality,” a fundamental aspect of life that the authors argue categorically distinguishes organisms from non-living matter.

From my perspective, this categorical division between non-living matter and organismic agency leaves us with another intractable explanatory gap, making it impossible to rationally understand the origin of life and its broader place in the cosmos. If life is fundamentally (ie, ontologically) different from non-living matter, the origin of life becomes not just a challenging problem, but analogous to the "hard problem of consciousness" as framed by Chalmers. In this sense, the origin of life becomes impossible to resolve without looking under the hood and tinkering with the presuppositions of physicalist ontology.

I believe we can generalize Vervaeke's understanding of relevance realization beyond the living world, beyond the agency of single-cells. One way of approaching this is to translate the core of relevance realization—ie, opponent processing—into the terms of Whitehead’s categoreal scheme. Whitehead's conceptions of physical and mental poles in the process of concrescence involve two different forms of prehension or ways of feeling reality. Physical prehension involves an inheritance of already past actualities, reiterating what has come before, while conceptual prehension ingresses possibilities, exploring alternatives to what has occurred. This dialectic between physical and conceptual prehensions gives rise to what Whitehead calls “propositional feelings,” allowing multicellular organisms with nervous systems to anticipate future trajectories.

Propositional feelings in Whitehead's terms are “lures for feeling,” similar to predictions, except unlike in Bayesian approaches, they are generally non-statistical, rooted rather in aesthetic contrasts than numerical calculations and precision weights. The authors distinction between logical inference and relevance realization is similar to Whitehead's distinction between conscious logical judgments and mostly unconscious propositional feelings. Consciousness is itself an especially intense form of propositional feeling, but most propositions propagate instinctively from organism to organism down the generations.

Concrescence is not just a non-Bayesian, but a non-Boolean form of growth (see Tim Eastman’s Untying the Gordian Knot). The organismic agency that Vervaeke and his co-authors describe, involving opponent processing and relevance realization, aligns well with Whitehead's account of the physical and mental poles of concrescence. By incorporating these Whiteheadian metaphysical categories, which apply across physical, biological, psychological, and cosmological domains, we can at least distinguish (but not ontologically bifurcate) biological organisms capable of entertaining novel propositions from physical and chemical processes, which are far more conformal in their inheritance of the past.

But there is evidence already of potentiality and creative emergence in the physical world. Even physical processes are engaged in evolution as a learning process, long before biological organisms emerged. Cosmic evolution, from plasma to atoms, from stars to galaxies and planets, indicates a slow but significant evolutionary agency has been at work from the get go. Energy does not simply run down, it emanates from a negentropic singularity. This process involves an interplay between both global aim and locally emergent aims interacting in a dialectical way.

Organisms live in a world overflowing with potential meaning. Given their limitations, they must select, filter, and realize relevance moment by moment in order to stay alive and flourish. This stands in stark contrast to what algorithms can do. Again, algorithms, defined as sets of symbols and coding operations executable on a universal Turing machine, can mimic or simulate certain aspects of organismic agency. But they cannot realize relevance. They cannot achieve genuine organismic agency.

The difference lies in the nature of these processes. Algorithms operate within a predefined formalized ontology, handling well-defined problems in a "small world." In contrast, organisms must navigate a "large world" filled with ill-defined problems, where relevance must be continuously realized to make sense of their environment.

The authors contrast their approach, which they call “agential emergentism,” with an approach they label “pancomputationalism.” Pancomputationalism assert that all physical processes that can be actualized (not just cognitive ones) must be Turing-computable. Pancomputationalism argues there is no fundamental boundary between the realms of the living and the non-living. Panexperientialism (Whitehead's view) also denies a fundamental boundary between life and physical nature, but it does so by “[fusing] them together as essential factors in the composition of ‘really real’ things whose interconnections and individual characters constitute the universe” (Modes of Thought, p. 150). From a Whiteheadian point of view, concrete physical processes are also noncomputable, not fully captured by any deterministic model.

It is interesting to note that, while the authors argue that pancomputationalism fails because “physical processes and phenomena are generally not discrete by nature,” Whitehead insists on the reverse: that actuality, including physical actuality, is “incurably atomic,” meaning it is fundamentally discrete, while the realm of possibility (ie, his “eternal objects”) is continuous.

What I appreciate about Vervaeke’s project is his attempt to naturalize a form of immanent teleology, which differs from extrinsically imposed teleology. This means recognizing purposes realized by organisms themselves, rather than purposes projected onto them by external observers or designers. However, this raises a significant challenge. If we acknowledge real purpose—genuine purpose, not just ‘as-if’ teleonomy—then the explanatory gap between the realm of physics and chemistry (where such teleology is said not to exist) and the living world becomes even more pronounced.

To bridge this gap, I think it is necessary to identify some cosmological precursor to the purposiveness that becomes more fully realized in the living world. Without such a precursor, the leap from non-living to living entities, from a world devoid of purpose to one rich with it, remains an unresolvable “hard” problem.

The authors claim that relevance realization requires an organism to assess potential outcomes of its behavior and to correct for errors. This reminds me of Whitehead’s comment in Process and Reality (p. 168):

“Error is the mark of the higher organisms and it is the schoolmaster by whose agency there is upward evolution. For example, the evolutionary use of intelligence is that it enables the individual to profit by error without being slaughtered by it.”

Whitehead’s insight highlights that higher organisms, through their capacity for modeling error instead of actually committing it (a capacity afforded by perception in the mode of symbolic reference and explained below), are able to adapt more effectively.

In Whitehead’s theory of perception, there are two pure modes of perception that cannot make errors: “causal efficacy” and “presentational immediacy.”

Causal efficacy is not merely a mode of perception but also of will and action. It conveys our sense of temporality, the feeling of receiving impulses from the past and launching them into the future. This mode of perception involves both inheritance and anticipation, embodying dynamic movement and the exchange of force, always with an emotional vector quality to it.

Presentational immediacy, on the other hand, constitutes our qualitative experience of the present moment as it is extended in space, what philosophers typically refer to as sense perception. Whitehead argues that presentational immediacy is not the most primal mode of perception, contrary to traditional philosophical thought. Instead, causal efficacy is primal, constituting our bodily reception of energetic emotional vectors from the past that compel us to reiterate prior experiences.

Causal efficacy is thus a perception-action mode of experience, while presentational immediacy is displays a static picture, lending itself to classical empiricist analysis like that of David Hume. Hume argued that if we only consider sense perceptions (qualia or sensory universals), we do not directly experience causation. Causation, however, is directly experienced through causal efficacy, encompassing both influences from without and actions in response.

Human experience predominantly occurs in the mixed mode, which Whitehead calls “symbolic reference.” Symbolic reference can make errors because it involves connecting our sense perceptions of the present with our bodily reception of the origins and aims of feelings transmitted to us from our environment. Symbolic reference thus associates vector qualities (causal efficacy) with scalar qualities (presentational immediacy), allowing for cognitive predictions and predictive modeling of a shared future. These models are recipes for action, not just static pictures, engaging the organism in the transformation and metamorphosis of its environing situation.

In Whitehead’s view, lower-grade actual occasions, such as those associated with physical and chemical processes, primarily involve causal efficacy with minimal if any symbolic reference. Cognitive judgment and logical inference require symbolic reference and are dramatically enhanced and extended by a second-order symbolism, namely, language. Nevertheless, from a panexperiential perspective, even at the level of physical processes there is a transmission of feeling, and thus a modicum of purposiveness is at work even in the vibratory reiteration of electromagnetic radiation.

As we move to biological organisms, propositional feelings and thus anticipation and prediction become possible, massively accelerating evolutionary learning. This acceleration is not only due to natural selection in the Darwinian sense but also because of organismic selection. Organisms actively select and modify themselves and their environments to suit their needs and desires.

The authors argue that what is relevant to an organism in its environment is never entirely subjective or objective. Instead, it is “transjective,” arising through the interaction of the agent with the world. The term transjective initially sounds similar to Whitehead’s concept of the “superjective,” but there are some differences.

For Whitehead, a subject is always also a superject, so he refers to the “subject-superject.” Unlike the Cartesian view of subjectivity, where a substantial subject has experiences as its accidental properties, Whitehead describes momentary subjects emerging from their objective data. In other words, a larval subject’s physical prehensions of the already actualized past (objective data) grow together with conceptual prehensions of possibilities during the process of concrescence, which aims at and achieves satisfaction.

This process involves an anticipation of the superjective, which provides an occasion will an intuition of how it will affect the future. In the context of a living organism, and particularly conscious humans, this superjective influence is a core aspect of our stream of consciousness as well as our sense of moral agency.

The subject-superject is internally related to the prior occasions within its nexus of becoming or historical route. It is externally related to the occasions who will succeed it. Moment by moment, subjective immediacy transitions into superjectivity, becoming an objectively immortal part of the changed environment that is inherited by the next concrescing occasion. Thus the world builds itself by way of the same concrescent process by which we build ourselves.

The authors claim that predictive processing models, including the free energy principle, remain embedded in a thoroughly computationalist framework. This framework views the allocation of relevance—and thus resources or precision weights—as a iterative and recursive algorithmic process. Such models, therefore, cannot fully account for relevance realization. From a Whiteheadian point of view, however, there may be a way to approach this iterative recursive process that offers a more nuanced understanding.

In Whitehead’s philosophy, the process of concrescence is both iterative and cumulative. Iteratively, it involves a conformal relationship to the past through physical prehensions. Cumulatively, it entails that each act of concrescence changes the universe forever, as the satisfaction achieved by a concrescing occasion perishes and becomes part of the environment inherited by future occasions. Consequently, each concrescence inherits a slightly different universe, reflecting the creative evolution of reality.

Crucially, concrescence is not an algorithmic process. It is an aesthetic process, not yet subject to the logical rules of non-contradiction or the excluded middle. These logical rules apply only to the satisfaction—the product—of the process. In this sense, concrescence is non-Boolean (see Eastman), resisting digital representation as 1s and 0s and any algorithmic formalization.

Similarly, relevance realization is a non-Boolean process. It cannot be captured by algorithmic or digital means, nor can it be fully represented by Bayesian statistical methods. Bayesian approaches, which involve calculating inferences, changing priors, and adjusting weights based on probabilities, may simulate aspects of relevance realization but do not actually achieve it. This is because relevance realization, like concrescence, involves a dynamic, context-sensitive, and creative engagement with the environment, which cannot be reduced to purely computational or statistical terms.

Thus, both relevance realization and Whitehead’s concrescence highlight the limitations of computational models in capturing the essence of these processes. They emphasize the importance of understanding these phenomena as inherently aesthetic and emergent, involving continuous transformation and adaptation that cannot be fully formalized or predicted by algorithmic means.

The authors discuss Robert Rosen’s approach to defining closure using category theory. Rosen argues that organisms are closed to efficient causation and identifies this efficient causal closure with final causation, referring to this as a form of “immanent causation.” The authors suggest that Rosen may have muddled Aristotle’s causes, and as a result is missing the importance of formal causes. To incorporate what is missing in Rosen’s framework, they draw on Hofmeyr’s extension of Rosen’s work, as well as Louie’s contributions.

Hofmeyr developed what he calls a fabrication-assembly (F,A) system, which extends Rosen’s metabolism-repair (M,R) system. These FA systems are also closed to efficient causation but remain open to formal causation, allowing for heritable variability. This introduces a distinction between mere recursion, where processes feed back on each other, and co-construction, where processes build on each other through continuous constraint-building.

As we’ve seen, the authors express concern that algorithmic and computationalist approaches result in a graded, rather than a categorical, difference in the complexity of living versus non-living systems. They affirm a categorical difference between non-living and living systems, arguing for a stronger claim: “It no longer makes sense to ask if organisms are computable if they are not completely formalizable in the first place.” Further, if even arithmetic is not completely formalizable, then why should we expect living processes to be? While discussions about computability focus on our limited ability to predict and simulate organismic behavior and evolution, their argument about formalization reveals even deeper limitations concerning our ability to explain living systems.

The authors next discuss the notion of operational closure, which implies that an organism responds to environmental perturbations based on its own internal norms, effectively becoming its own final cause. This concept of closure has always concerned me, because it seems to suggest a kind of biological solipsism, where the organism is informationally isolated from its environment. However, the authors illustrate that this is not entirely the case by using the example of a paramecium swimming up a glucose gradient. The paramecium’s cilia beat as a consequence of its metabolism and the maintenance of its internal milieu. Yet, the effect of the cilia extends beyond the cell, causing the organism to move toward its food sources through the turbulent flow of its watery surroundings.

This example demonstrates that the constraints subject to closure within the organism exert effects beyond the organism’s boundaries, indicating a kind of extended cognition. The boundary between the organism and its environment becomes fuzzy, highlighting that the organism’s agency cannot be fully understood without considering its embeddedness in its environment. The environment provides the context in which the organism operates, and the organism, in turn, affects and modifies this environment.

Extended cognition can serve as a bridge to a more integrated understanding of living processes, emphasizing their inseparability from their environments. While the authors maintain a categorical distinction between the autonomy and agency of organisms and the non-semantic, non-agential, purposeless nature of mere physics and chemistry, I believe this division needs to be re-evaluated. The intimacy of the tranactions between an organism and its environment indicate that agency and cognition are not confined within the organism but are distributed across the whole organism-environment nexus. In short, agency is in the relating and not in one of the relata.

Of course, the concept of operational closure is not inconsistent with thermodynamic openness—organisms are open to energy flows but are said to maintain closure concerning semantic information. Semantic information is realized within the organism, while the external environment is perceived as a landscape of affordances.

In my opinion, once we recognize that the entire environment conspires with the organism to facilitate its operations, it becomes impossible to reify the physical world as something categorically separate from the organism’s semantic network. The environment and the organism are intradependent (ie, internally related), with the organism’s ability to perceive and act upon affordances being deeply rooted in the real potentiality of its surroundings.

The idea of affordances, introduced by J.J. Gibson, suggests that the environment offers various possibilities for action that are directly perceivable by an organism. These affordances are not intrinsic properties of the environment alone but emerge from the intra-action between the organism and its environment. This relational perspective aligns with the notion of extended cognition, where cognitive processes extend beyond the organism’s boundaries to include the environment.

So in this sense, the environment is not a passive backdrop but an active participant in the organism’s cognitive and operational processes. The landscape of affordances is shaped by the organism’s needs, capabilities, and experiences, and the environment continuously intra-acts with the organism’s enactions. This intimate and indeed seamless organism-environment relation suggests a more integrated view of life processes, where the distinction between the living and non-living becomes less clear-cut. In effect, the body of an organism includes in degrees the entirety of its earthly and cosmic environment. There is no ontological rift between living beings and their surroundings.

In Process and Reality (p. 119-120), Whitehead elaborates on our cosmic embeddedness, the bodily basis of perception, and how our experience is fundamentally characterized by a process of inheritance of environmental vector-feelings. He states:

“…the human body is to be conceived as a complex ‘amplifier’—to use the lan­guage of the technology of electromagnetism. The various actual entities, which compose the body, are so coordinated that the experiences of any part of the body are transmitted to one or more central occasions to be inherited with enhancements accruing upon the way, or finally added by reason of the final integration. The enduring personality is the historic route of living occasions which are severally dominant in the body at suc­cessive instants. The human body is thus achieving on a scale of concen­trated efficiency a type of social organization, which with every gradation of efficiency constitutes the orderliness whereby a cosmic epoch shelters in itself intensity of satisfaction.

The crude aboriginal character of direct perception is inheritance. What is inherited is feeling-tone with evidence of its origin: in other words, vector feeling-tone. In the higher grades of perception vague feeling-tone dif­ferentiates itself into various types of sensa—those of touch, sight, smell, etc.—each transmuted into a definite prehension of tonal contemporary nexus† by the final percipient.

In principle, the animal body is only the more highly organized and immediate part of the general environment for its dominant actual occa­sion, which is the ultimate percipient. But the transition from with­out to within the body marks the passage from lower to higher grades of actual occasions. The higher the grade, the more vigorous and the more original is the enhancement from the supplementary phase. Pure recep­tivity and transmission give place to the trigger-action of life whereby there is release of energy in novel forms. Thus the transmitted datum ac­quires sensa enhanced in relevance or even changed in character by the passage from the low-grade external world into the intimacy of the human body. The datum transmitted from the stone becomes the touch-feeling in the hand, but it preserves the vector character of its origin from the stone. The touch-feeling in the hand with this vector origin from the stone is transmitted to the percipient in the brain. Thus the final perception is the perception of the stone through the touch in the hand. In this per­ception the stone is vague and faintly relevant in comparison with the hand. But, however dim, it is there.

This passage underscores the complexity and layered nature of perception in Whitehead’s philosophy. It highlights the process by which objective data is physically felt, transformed, and integrated into higher-order cognitive experiences, retaining traces of its origins while being modulated by the organism’s neurophysiology.

The authors are careful to clarify that their explanation of teleology does not involve future events causing present ones. Rather, anticipatory systems pull the future into the present. Organisms engage with possibilities always already latent in the present. This view closely aligns with Whitehead’s understanding of propositional feelings, which act as lures for feeling, functioning both as conveyors of past truths into the present, and as predictions of future truths.

Whitehead’s propositions involve a comparison between the data of physical and mental prehensions, hybridizing actualities with eternalities to afford an anticipation of relevant future possibilities. This process is not about the future exerting a causal influence on the present (in a process ontology, the future does not exist yet), but about the present entertaining and realizing adjacent possibilities. Teleology is immanent, in this sense. It occurs within the horizon of the experiential present, manifesting as the organism’s anticipation of future possibilities.

For Whitehead, propositional feelings are just as much about the potentiality of what could be as they are a repetition of what already is. As he says, “In the real world it is more important that a proposition be interesting than that it be true. The importance of truth is, that it adds to interest” (PR, p. 259).

The authors of the paper advocate for an organism-centered model of Darwinian evolution, which they refer to as “situated Darwinism,” drawing on the research of Walsh. They view their own trialectic model as a reformulation and extension of relevance realization into evolutionary theory. In an effort to bring him out of Darwin’s shadow, I should mention that their emphasis on the agency of organisms in ecological contexts aligns well with the approach of Alfred Russell Wallace, the co-discoverer of evolution by natural selection.

The authors argue that their claims about organismic agency are firmly grounded in empirical knowledge of the physical and biological world. They are seeking a naturalistic account of relevance realization that requires no mysterious new forces or fundamental laws of physics. They propose a shift in focus from the predictive laws governing state changes in physico-chemical processes to the unpredictable historical succession of dynamic constraints operative in the living world. These constraints are refined into a combination of closed, efficient, and open formal causes, which continuously restrict and channel the degrees of freedom, as well as the rate and direction of change.

But even in physics, models describing the motion of particles in phase space are abstractions that do not fully capture the creative potency on evidence in the history of the physical world. While the biological world leverages relevance realization to radically accelerate the ingression of novelty, the physical world, though more predictable and amenable to mathematical modeling, is still pregnant with unprestatable possibility.

For instance, if we were a disembodied intelligence observing the early universe and devised a precise mathematical model to account for the plasma dynamics, even with complete knowledge of the phase space dynamics of every particle and field, we still would not be able to predict the formation of hydrogen atoms, much less the emergence of stars and galaxies. This inability to predict the complex and creative evolution of the universe highlights the limitations of purely mathematical models, even in physics. These models simulate physical processes but do not fully capture the real, creative dynamics at play. This suggests to me that the boundary between the non-living and the living may not be categorical after all. It is a difference in degree.

The authors conclude their paper by recuperating a contemporary version of Aristotle’s contribution to the study of different levels of nature: the physical, vital, sentient or psychological, and intellectual. They are not advocating for sharp divisions but rather important distinctions between these levels.

Understanding nature as comprised of distinct and irreducible yet interconnected levels of organization is an essential part of post-reductionistic science. When considering the intellectual layer of our being, we are contemplating thinking itself. For Aristotle, this culminated in the concept of “thought thinking itself,” his definition of God, the prime mover. Aristotle posited that all things are, in some sense, mimicking or being lured by this Active Intellect, with the human intellect being made in its image.*

Non-human animals partake of soul or psyche, which endows them with sentience, while plants are said to have a nutritive capacity, their vitality. (Contemporary plant scientists increasingly affirm that plants, too, are sentient.) Viewing the universe as a macrocosmic archetype for its own microcosmic recapitulation in living beings allows us to see these levels as dependently co-arising rather than linearly caused. Our evolving cosmos can no more be explained simply by a bottom-up stack of emergences than it can be explained simply as a top-down hierarchy of emanations (to use Vervaeke’s favored neoplatonic terms not mentioned in this article). Cosmogenesis involves an interplay of both emergence and emanation.

Considering the Earth, most of the so-called “dead matter” on its surface has been transformed by living organisms over billions of years. The minerals we find on the surface of the Earth today have almost all passed through the metabolisms of living organisms at some point. Thus, the relationship between matter and life is not a linear emergence of life from matter but a complex interplay wherein matter as we know it must be considered coeval with life.

---

*Aristotle describes the active intellect as a pure form of actuality (energeia), similar to his description of God in Metaphysics (Book XII). God, in Aristotle’s view, is the Unmoved Mover, a perfect, eternal, and immaterial being whose existence is necessary and whose primary activity is pure thought or contemplation (theoria).