1.3 Dualism and the Five Levels of Existence

To review, an information processor or IP is a physical construction, e.g. a living thing, that manages (creates and uses) information. As I have noted before, the reason that living things develop functions to help them survive is that they can — the opportunity exists that disposes them to keep existing. In other words, they have a reason to live. If I wrote a pointless computer program that had no disposition to do anything or accomplish any function, we would say it was devoid of any information or function; it processes data but no information. Something only counts as information or as having function when it is useful, meaning that it can be applied toward an end or purpose. But use, end, and purpose are not physical things or events; they are at most ways of thinking about things. However, if we could think about an approximate future state of physical things or events, where this approximation was defined in terms of similarities to past things and events, then we could think about making it our purpose to cause that future state to happen. Information processors are physical machines that use physical mechanisms to model physical things and events in a nonphysical way and then apply those nonphysical models back to physical circumstances to change them. It all sounds wildly complicated and unlikely, except that is exactly what life does: it collects physical processes (genetic traits) that approximately cause future physical events to transpire in such a way that it can keep doing it. The purpose of the prediction and application is nothing more than to be able to keep predicting and applying: to survive. But the “it” that keeps doing “it” is always changing or evolving, because time doesn’t repeat. Living IPs only do things similar to things they did before, and instead of just maintaining themselves, they create new IPs as offspring that are similar but never quite the same as themselves.

Information and function make generalizations about physical (or functional) things. These generalizations are references to these things, and references are not physical themselves even though a physical mechanism exists in an IP to hold them. The information is not how the reference is physically implemented (in neurons or computer chips); it is the useful, purposeful, functional ends to which it can be applied indirectly through references. These ends exist but are not physical; they are a new kind of existence I call functional existence. The fabric of functional existence is capacity, not spacetime. Specifically, it is the capacity to predict what will happen with better than random odds. This capacity can be described in other ways, such as the ability to answer questions, resolve uncertainties, or be useful, all of which refer to generalized ways of guessing that new things can happen that were similar to past things. I will use information synonymously with functional existence, but there is a subtle difference: information is often spoken of independently of the information processing that can be done on it, but functional existence requires both the information and its accompanying information processing as a functional whole.

We are justified in saying that function (or information) actually exists because things exist when we can discriminate them, they persist, and we can do things with them. We can discriminate information, it persists, and we can do things with it, yet it is not physical, so this qualifies it as a distinct category of being. We can therefore conclude that interactionist dualism is true after all. The idea that something’s existence can be defined in terms of the value it produces is called functionalism. For this reason, I call my brand of interactionist dualism form and function dualism, in which physical substance is “form” and information is “function”. I hold that physical things except for IPs are best explained using physicalism and IPs are best explained using a combination of physicalism and functionalism. While this means I am endorsing physicalism and functionalism, I am only endorsing a version of each. Specifically, I endorse physicalism but require it to drop the restriction that everything IPs do is physical, and I endorse functionalism only in the sense that I describe here. Many variations of functionalism with different ontologies exist which I will not describe or defend. The version I propose says that function (aka information) exists in an abstract, nonphysical way, but that a physical world like ours can use functional through information processors. Consequently, although function itself is abstract and independent of physical support, the use of function in a physical world is quite concrete and dependent on elaborate feedback systems that all derive from living things. As an interactionist, I hold that form and function interact in the mind and that they do so via information processing.

Probably most cognitive scientists already consider themselves to be functionalists, in that they view mental states and processes functionally, but that doesn’t make functionalism a well-defined stance. While progress can be made without a coherent definition, a vagueness pervades the conclusions that creates uncertainty about what has been shown. By explicitly defining function and information as a kind of existence that is independent of physical substance, I hope to clarify both the physical and functional aspects of information processing to show how these two kinds of existence persist, interact, and influence the future.

To develop this idea, I’m going to further distinguish five levels of understanding we can have for each of the two kinds of existence, only the first two of which apply to physical things:

Noumenon – the thing-in-itself. Keeps to itself.

Phenomenon – that which can be observed about a noumenon. Reaches out to others.

Percept, from perception – first-order information created by an information processor (IP) using inductive reasoning on phenomena received . Notices others.

Concept, from conception – second-order information created with deductive reasoning, usually by building on percepts. Understands others.

Metaconcept, from metacognition – third-order information or “thoughts about thoughts”. Understands self.

We believe our senses tell us that the world around us exists. We know our senses can fool us, but by accumulating multiple observations using multiple senses, we build a very strong inductive case that physical things are persistent and hence exist. Science has increased this certainty enormously with instruments that are both immune to many kinds of bias and can observe things beyond our sensory range. Still, though, no matter how much evidence accumulates, we can’t know for sure that the world exists because it is out there and we are in here. But we can imagine that a thing physically exists independent of our awareness of it, and we refer to this standalone type of existence as the thing’s noumenon, or thing-in-itself (what Kant called das Ding an sich). The only way we can ever come to know anything about noumena is through phenomena, which are emanations from or interactions with a noumenon. Example phenomena include light or sound bouncing off an object, but can also include matter and energy interactions like touch, smell, and temperature. When we talk about atoms, we are referring to their noumena, or actual nature, but we don’t really know what that nature is. We only know noumena by observing their phenomena. So everything science or experience tells us of physical apples or atoms is entirely in terms of their phenomena. We believe they have a noumenal existence because it can be measured in so many different ways, and this consistency would be unlikely if the apple or atom was an illusion. We know this because an illusion of an apple, say a picture or projection of one, lacks many phenomena that real apples provide.

All knowledge based on phenomena is called a posteriori, which includes all knowledge we have of the physical world. We can have direct or a priori knowledge of noumena we can logically perceive in our own minds, which most notably includes things that are true by definition. A priori knowledge includes everything that is true by construction, which includes the logical implications of explicit deductive models. If we define rules of arithmetic such that addition necessarily works, then the rules are just part of the definition and all their implications are a priori even if we can’t easily see what all those implications are. While Kant’s greatest contribution to philosophy was the recognition that we can only know the world through phenomena, leaving physical noumena unknowable, he was perturbed that this implied that philosophers can never derive anything about the physical world from reason alone. Philosophers had always thought some truths about the world (e.g. the idea of cause and effect) could be known through thought alone, yet he had apparently proven that knowledge of the outside world must be the exclusive domain of natural philosophers, i.e. scientists. While some saw this as a fatal blow to philosophy, all it really did was clarify that philosophy is a functional affair, not a physical one. The recognition that we only know the world through phenomena was an important breakthrough because now we can readily accept that everything about the physical world is approximate, a posteriori knowledge that, far from being absolutely true, merely extrapolates about the future based on patterns seen before. Causes and effects are convenient generalizations about the world, not intrinsic physical essences.

Perception is the receipt of a phenomenon by a sensor and adequate accompanying information processing to create information about it. Physical things have noumena that radiate phenomena, but they never have perception since perception is information. A single percept is never created entirely from a single phenomenon; the capacity for perception must be built over billions of inductive trial-and-error interactions as life has done it. We notice a camera flash as a percept, but only because our brain evolved the capacity over millions of years to convert data into information. So if a tree falls in the forest and there was nobody to hear it, there was a phenomenon but no perception. Because IPs exploit the uniformity of nature, our perceptions can very accurately characterize both the phenomena we observe and the underlying noumena from which they emanate, even if complete certainty is impossible.

Perception includes everything we consciously experience without conscious effort, and includes sensory information about our bodies and the world, and also emotions, common sense, and intuition, which somehow bubble up into our awareness as needed. Information we receive from perception divides into two parts, one from nature and one from nurture. The nature part, innate perception, provides information in the form of feelings from senses and emotions that require no experience to feel and which don’t change given more experience. The nurture part, learned perception, provides information in the form of memories and impressions, and continually changes during our lives based on the contents of stored experience. Color and fear have innate parts independent of experience, but they are only meaningful to us because of our experience with them; we have to learn what our bodies are telling us. For convenience, I will usually call learned percepts or intuitions subconcepts, since perception comes just below conception. Our capacity to develop common sense and intuition as subconcepts is itself innate, but the experiences themselves were not anticipated by our genetics and are entirely circumstantial to our individual lives. Everything we perceive is influenced by both innate and learned perception, even though they originate from completely independent sources. So we see red using innate perception, but a lifetime of experience seeing red things then influences our perception with impressions we attribute to either common sense or intuition. All information created by perception is first-order information because it is based on induction, which is the first kind of information one can extract from data. Inductive reasoning or “bottom-up logic” generalizes conclusions from multiple experiences based on similarities, a trial-and-error approach. Entirely outside the purposes of brains, all genetic information is created inductively, so I am going to use the word “perception” more broadly than mental perception to the evolutionary capture or perception of functionality into biological traits. More slowly than our intuitive mind, evolution “perceives” patterns that can provide functionality and it captures them in DNA as traits. A few of those genetic traits are mental and create for the innate perceptions we experience as senses and emotions.

Conception approaches information from a different direction. Instead of looking for associations from patterns from the bottom up, it works from the top down by proposing the existence of abstract entities called concepts that interact with each other according to rules of cause and effect. Concepts idealize frequently seen patterns into discrete buckets that group things or events into chunks that engage in predictable sorts of interactions with related concepts to form a conceptual model. Conceptual models obey whatever rules of logic we imagine for them, but they will predict best what will happen if they use deductive logic, because then they can reach conclusions that are logically certain. (Although conceptual thinking can follow any brand of logic and not necessarily full-fledged deductive logic, I will often refer to top-down or conceptual thinking as deductive for simplicity. However, conceptual models are broader than deductive models because concepts are vaguer than the more precisely-specified axioms of deduction.) The challenge of concepts is in building concepts and models that correspond well to situations in which they can be applied. To help with this, our base concepts are strongly linked to percepts and our conceptual rules are heavily influenced by patterns we intuit from perception.

The transition from percept to concept is gradual and is arguably a matter of perspective. From a functional standpoint, it has more to do with how the information is used than how it is structured. Structurally, information in the brain is all just interconnections, but functionally, top-level perspectives are necessary, and this makes conceptual interpretations meaningful. We trust our perceptions based on our memory and familiarity with them, but they don’t tell us why things happen. Understanding, comprehension, grasp, and explanation generally imply a conceptual model that says why. Within the logic of the conceptual model, especially if it uses deductive logic that reaches inescapable conclusions, we can know exactly what will happen with perfect foreknowledge as a logical consequence, which gives us the confident feeling that comes with understanding. We know that models never apply perfectly to the physical world, but when they come close enough for our purposes we take our chances with them (and even trust them; more on this later). The implications or entailments of logic can be chained together, allowing conceptual models to take us with certainty many steps further than induction, which can basically only reach probable one-step conclusions. I call knowledge built from conceptual models second-order information because it gives us understanding, as opposed to the mere familiarity of inductive first-order information.

Like perception, our capacity for conception is itself innate even though our concepts themselves are all learned1. So, as with perception, I will distinguish innate conception and learned conception as different components. A big difference between perception and conception, however, is that learned perception (subconcepts or intuitive knowledge) only grows at a fixed rate with experience, while learned conception (concepts or rational knowledge) is essentially unlimited because conceptual models can build on each other to become ever more powerful. This means we can not only leverage up our conceptual models over our own lifetimes, we can pass them on from generation to generation. Notably, although our innate conception is probably not much different than two thousand — or possibly even 50,000 to 200,000 — years ago, language and civilization have dramatically transformed our understanding of the world through learned conception.

I mentioned before that I would discuss whether cause and effect are meaningful in the context of induction. Inductive information processing acts based on past experience, but only looks one step ahead instead of chaining causes and effects. We can call the circumstances before an inductive action the cause and the result the effect. However, this uses concepts, calling out specific causes and effects in a general way. Induction itself doesn’t need the concepts of cause or effect; it just happens. However, given that caveat, it is fair to describe inductive processes conceptually using single-step causes and effects. Although these causes and effects happen without any intent, purpose, design, or strategy, they are dispositional. Life is predisposed to survive, and that is why the inductive processes that produce these simple causes and effects happen. Survival is the ultimate cause that produces all the inductive effects of life. This is very different from loose rocks that fall from cliffs, which have no disposition to do anything. The laws of physics are deductive tools that describe behavior in causative terms, e.g. that the loosening of clumps called rocks along cracks in cliffs will cause falling. Physically, however, the same subatomic rules are followed everywhere, and there are no rocks or cliffs except as abstract groupings in our heads.

Metacognition is thoughts about thoughts, or, more specifically, deductive reasoning about thoughts. Conception is a first-order use of deductive reasoning in which the premises are groupings of percepts, while metacognition is a higher-order use of deductive reasoning in which the premises can be concepts themselves, or concepts about concepts abstracted any number of levels. All physical things, grouped to any level of generality, are still just first-order concepts because we have a strong perceptual sense of their scope. So apple tree, tree, and plant are all first-order concepts, but lumber source and fruit season are metaconcepts about trees. Looking inward, we have a concept of ourself that is based on our subjective experience of doing things, but we also have a metaconcept of ourself which holds thoughts we have thought about ourselves. Metacognition expands our realm of comprehension from matters of immediate relevance to matters abstracted one or more levels away. It extends our reach from physical reality to unlimited imagination. I call metaconcepts third-order information because this move to arbitrary degrees of indirection unlocks new kinds of explanatory power. Conception, both first and higher-order, heavily leverages perception but gives us a kind of window into the future.

Noumena and phenomena just happen. That is, they are not functional in that they do nothing to influence future events. All physical things are necessarily just noumenal and phenomenal because they just happen, which we believe means that invariant physical laws of the universe apply to them. We can also apply the idea of noumena and phenomena to functional things as well by referring to them. Percepts and concepts are about things and we usually know what they are about and don’t need to observe them to find out. But things can get complicated and it may be necessary to observe a functional system to learn more about it. This is the case when we are using our minds to figure out the functions of other IPs, or if we have created a conceptual model whose implications are too hard to figure out logically. In these cases, we isolate the functional entity as a noumenon called a black box whose internal workings are taken to be unknown, and we make phenomenal observations of its behavior. For example, if we find a calculator-like device, we can play around with its buttons and observe what happens. We can never know for sure the true function the device was created to have, but we can develop theories based on observations of it. In this way, functional noumena are ultimately unknowable just like physical noumena. Our theories can become increasingly accurate, but only as they pertain to aspects of their existence that are useful to us, which may be quite different than their creators had in mind (if they were even created for a purpose).

When our concepts are based on formal models like mathematics, we have access to their actual noumena because we defined them. In this case, all their logical implications are also noumenal by definition. But the implications of many formal models can be too complex for us to reason out (i.e. prove), so we may instead opt to gather information about them by induction. If we can run the model on a computer, we can do this by running millions of simulations and analyzing the results for patterns. For example, weather simulators are precisely defined, but we have no idea what all the implications of their rules might be except by running simulations and seeing what pops out. Similarly, in our own minds we can’t logically forecast all the implications of many conceptual models, so we run simulations that project what will happen using subconceptual and conceptual heuristics we have refined over time. In so doing, we have effectively built functional noumena that describe the world about which we then make phenomenal observations.

Evolved functions or traits can be referred to as noumena to which evolution makes adjustments based on phenomenal observations called natural selection. As with the above cases, the noumena themselves can never be understood directly, but only in terms of surmised functions of their phenomena. The noumenon of any gene itself blends information from all the feedback that created it, which, if you think about it, would take an infinite amount of information to describe because time and space don’t break down into quantized units, even if matter and energy do. Even a finite attempt to characterize all that feedback would be astronomically complex. But we can simplify a genetic function using a deductive cause-and-effect model the helps us understand it in the sense that it empowers us to make any number of useful predictions about what it will do. We know such conceptual models leave out details, but they are still useful. Daniel Dennett calls explanations of evolved noumena “free-floating rationales”2. This is a great way of putting it, because it emphasizes that the underlying logic is not dependent on anything physical, which is important because function is not a physical thing. All functional noumena are necessarily free-floating in the sense that they don’t have to be implemented to exist; they embody logical relationships whether anyone knows it or not. But, of course, all physical IPs (either through DNA or thoughts) are implemented and are ultimately concerned only with what can be implemented, because function must ultimately be useful. In other words, we can imagine information and IPs in the abstract, but they can’t actually process any information.

Perception, conception, and metacognition are purely functional modes of existence. They depend on a physical IP, but the same function could potentially be rendered by any number of different physical implementations. We could happily live our lives entirely within a computer simulation if it did a good enough job. This doesn’t mean implementations of information on computers would be indistinguishable from implementations using physical matter; they could always be distinguished. Scientific experiments within the simulation could expose differences between physical and simulated reality. Of course, we could either prevent simulants from doing such experiments, or change the results they obtain, or modify their thought processes to fool them. But I digress; my point is not whether simulation could be a feasible alternative to physical life, but that we live our lives entirely as functional entities and only require a physical environment as a source of new information. We may require physical IPs to exist and think, but we are first and foremost not physical ourselves.

Let’s review our three quandaries in the light of form and function dualism. First, the origin of life. Outside of life, phenomena naturally occur and explanations of them comprise the laws of physics, chemistry, materials science and all the physical sciences. These sciences work out rules that describe the interactions of matter and energy. They essentially define matter and energy in terms of their interactions without really concerning themselves with their noumenal nature. As deductive explanations, they are based in the functional world of comprehension and draw their evidence from our perception of phenomena. While the target is ultimately the truth our noumena of nature, we realize that models are functional and not physical, and also only approximations, even if nearly perfect in their accuracy. With the arrival of life, a new kind of existence, a functional existence, arose when the feedback loops of natural selection developed perception to finds patterns in nature that could be exploited in “useful’ ways. The use that concerns life is survival, or the propagation of function for its own sake, and that use is sufficient to drive functional change. But perception forms its own rules transcendent to physical laws because it uses patterns to learn new patterns. The growth of patterns is directed toward ever greater function because of the functional ratchet. It exploits that fact that appropriately-configured natural systems are shaped by functional objectives to replicate similar patterns and not just by physical laws indifferent to similarity.

Next, let’s consider the mind-body problem. The essence of this problem is the feeling that what is happening in the mind is of an entirely different quality than the physical events of the external world. Form and function dualism tells us that this feeling is a reflection of the fact that some of the underlying natural entities are physical while others are functional. Specifically, the mind is a construction of functional entities, and the nonliving physical world is constructed of physical entities, but all living things are a composite of functional and physical entities. This division is not reducible, as physicalists would have us believe, because function is concerned with and defined by what is possible, and the realm of the possible is entirely outside the scope of mere physical things. While function doesn’t reduce to physical, it does depend on cellular metabolism in the case of life, or on the brain in the case of the mind. The mind is a natural entity comprised of a complex of functional capacities implemented using the physical machinery of the brain. The language conveniently has one word, mind, to refer strictly to the functional aspects, and another word, brain, to refer strictly to the physical aspects. The conscious mind is the functional portion to which we attribute conscious traits like awareness, attention, feelings, and thoughts, but nonconscious parts contribute to function to comprise the overall mind. Some nonconscious thoughts can “rise” up to consciousness in one form or another, while others can’t, but either way, our overall state of mind depends on the operation of the whole mind.

Finally, let’s look at the explanatory gap, which is about explaining with physical laws why our senses and emotions feel they way they do. I said this gap would evaporate with an expanded ontology. By recognizing functional existence as real, we can see that it opens up a vastly richer space than physical existence because it means anything can be related anything in any number of ways. The world of imagination is unbounded, while the physical world is closely ruled by rather rigid laws. The creation of IPs that can first generalize inductively (via evolution of life and minds) and then later deductively and metacognitively (via further evolution of minds) gave them increasing degrees of access to this unbounded world. The functional part alone is powerless in the physical world; it needs the physical manifestation of the IP and its limbs (manipulative extremities) to impact physical things; there is nothing spectral going on here. Physical circumstances are always finite and so IPs are finite, but their capacities are potentially unlimited because capacities can be generalized beyond specific circumstances as broad as we like. So to close the explanatory gap and explain what it means to feel something, we should first recognize that the scope of feeling, experience, and understanding was never itself physical; it was a functional effect within an IP. So what happens in the IP to create feelings?

I’m just going to say the answer here and develop and support it in more detail later on. The role of the brain is to control the body in a coordinated way, and as a practical matter, it solves this using a combination of bottom-up and top-down information processing. These two styles, which have to meet somewhere in the middle, are usually called intuitive and rational. The rational mind is entirely conscious, while the intuitive mind provides the conscious mind with a wealth of impressions and hunches from some inner source I call the nonconscious mind.

The role of consciousness is to focus specifically on top-level problems that the nonconscious mind can’t handle by itself. To create this logical view of top-level concerns, the nonconscious mind presents information to the conscious mind by creating a theater of consciousness. Conscious experience is a highly produced and streamlined version of the information the nonconscious mind processes. The analogy to a movie is particularly good because movies are designed to simulate consciousness. What we feel as pain is really just part of the user interface between the nonconscious and conscious processes in the brain. Senses, feelings, and lessons from the school of hard knocks bubble up to consciousness through the intuitive mind. We are aware of our bodies, our minds, the world, and the passage of time, and we have a specific conscious feeling of them through senses and emotions. Rationally, we organize the world into objects and other concepts that follow rules of cause and effect. Consciousness merges the two styles together pretty seamlessly, but they are actually quite different entirely functional constructs. Some of our intuitive and rational knowledge, though itself functional, is about the physical world, and some of it is about our mental world or other nonphysical subjects. Most notably, our somatic sensory information is about our bodies and our emotions are about our minds. They are not about our bodies and minds in a physical way; rather, they tell us things our bodies and minds need. Whether about physical, mental or other things, knowledge serves functional purposes and so can be said to be a functional entity.

To summarize my initial defense of dualism, I have proposed that form and function, also called physical and functional existence, encompass the totality of possible existence. We have evidence of physical things in our natural universe. We could potentially someday acquire evidence of other kinds of physical things from other universes, and they would still be physical, but they may produce different measurements that suggest an entirely different set of physical laws. Functional existence needs no time or space, but for physical creatures to benefit from it, there must be a way for functional existence to manifest in a natural universe. Fortunately, the feedback loops necessary for that to happen are physically possible and have arisen through evolution, and have then gone further to develop minds which can not only perceive, but can also comprehend and reflect on themselves. Note that this naturalistic view is entirely scientific, provided one expands the ontology of science to include functional things, and yet it is entirely consistent with both common sense and conventional wisdom, which hold that a “life force” is something fundamentally lacking in inanimate matter. We also see evidence of that “life force” in human artifacts because we are good at sensing patterns with a functional origin. Some patterns that occur in nature without any help from life do surprise us by appearing to have a functional origin when they don’t.3 Life isn’t magic, but some of its noumenal mystery is intrinsically beyond a complete deductive understanding. But we can continue to improve our deductive understanding of life and the mind to give us a better explanatory grasp of how they work.

  1. People used to think core concepts could be innate, but this just doesn’t add up, as my explanations of concepts in Part 3 will make clear.
  2. Dennett, Daniel C., From Bacteria to Bach and Back Again: The Evolution of Minds, W. W. Norton & Co, 2017
  3. Death Valley’s sailing stones and the Giant’s Causeway are common examples.

Leave a Reply