Abstract: Neurocognitive Mechanisms (2020), by Gualtiero Piccinini, is a philosophical and neuroscientific work, seeking to ground cognition and its phenomena in a comprehensive neurocomputational theory. But the philosophical treatment is wrong. The introduction of “aspects” of a single level of reality also introduces something that looks independently of that single level. We fall back on an unexplained dualism. An alternative is presented.
General theories of reality including the mind are rare. All are reductive rather than universal, because they are scientistic, philosophical or religious odedience. None of them synthesize our epistemic tools. The most successful attempts are those of multidisciplinary authors. This is the case of Gualtiero Piccinini, as well versed in the philosophy of mind as neuroscience. Looking for essays to compare with Surimposium, I recently discovered his Neurocognitive Mechanisms, Explaining Biological Cognition, published the same year, in 2020. The title does not honor Piccinini’s ambitions. It is much more than a neuroscience book, with in particular a first chapter that proposes a new solution to the philosophical controversy between reductionists and emergentists.
« Cognition is explained by multilevel neurocognitive mechanisms »
The introduction causes me the greatest excitement. This profession of faith is exactly that of Stratium, part of Surimposium devoted to the mind. Stratium, a contraction of ‘strata’ and ‘atrium’, represents the mind as a building of hierarchical and integrated neural services, with gateways on each floor. Although neuroscience is seduced by the “multi-level”, it is still very focused on the horizontal, geographical analysis of brain function.
Piccinini installs the roles of science and metaphysics:
« Sorting out what there is, what properties objects have, and how they relate is the business of science. Sorting out what sorts of objects there are as well as what sorts of properties and relations they have is the business of metaphysics. »
The text outlines a difference between the ontological look of science and the teleological look of metaphysics. But Piccinini is content to hold science and metaphysics as two epistemic tools, without worrying about their genesis and how to coordinate them. It lacks the formalism of the double look, which makes everything appear as constitution and phenomenon, which obliges to make ontology and teleology coincide.
Composition and production
« I will use the term “composition” for the relation between wholes and their parts: parts compose a whole; a whole decomposes into its parts. I will use the term “realization” for the relation between properties of a whole and properties of its parts taken collectively, or between its parts taken collectively and all of the parts’ parts taken collectively: higher-level properties are realized by lower-level properties; lower-level properties realize higher-level properties. »
The two directions of the look are present but their vocabulary is not made independent by Piccinini. He gives it to the ontological direction: everything comes from the components. To make the double look valid, it is necessary to say: “a whole sees itself broken down into its parts” and “the properties of the higher level see themselves realized by those of the lower level”. Impossible otherwise to understand the phenomenal aspect. The phenomenon appears to something at least as complex that concerns it, not to the components.
« Both composition and realization are asymmetric, irreﬂexive, and transitive. Both are relations of synchronic metaphysical necessitation: once you have the parts with their properties and relations, you necessarily have the whole with its properties and relations. Necessitation does not entail dependence! »
Piccinini establishes the principle of relative independence between the properties of the parts and the whole. The whole necessarily involves its parts but is not dependent on them. Piccinini’s explanation for this relative independence —aspects of a unique level of reality that reveal themselves depending on the context— is particularly original. Before getting there, he lays his fundamental foundations by dealing with the subject of causal exclusion. Unfortunately, he does so with an outdated philosophy. This bad start on complexity makes him fall into the reductionist camp, although he denies it. But first let’s see his presentation of causal exclusion:
The puzzle of causal exclusion
« Brain B is composed of a bunch of neurons […] which I’ll collectively label NN. […] Brain state SB and neural state SNN [are] such that SNN realizes SB. Consider a voluntary action E performed by an organism, and consider two causal claims :
SB causes E
SNN causes E
When we make this kind of causal claim, we generally assume that each cause is sufﬁcient for the effect under the relevant background conditions. That is to say, if SB causes E, then the occurrence of SB is enough, all by itself, to bring about E. By the same token, if SNN causes E, the occurrence of SNN is enough, all by itself, to bring about E.
The puzzle of causal exclusion arises because if SNN is already enough to cause E, there seems to be nothing left for SB to do. […] The problem generalizes. I picked neurons and brains among many other levels that I could have picked. The same puzzle can be run with neural systems, neural networks, molecules, atoms, and so on—not to mention whole organisms. This is the puzzle of causal exclusion: if any level is enough, all by itself, to cause an effect such as E, then other levels seem either dispensable or identical to the one causally efﬁcacious level. »
Strangely Piccinini renounces his idea of ‘necessity without dependence’ between whole and parts. He espouses the point of view of Kim (1998, 2005) who instead enacts a strict dependence between the causalities of the whole and the parts. In a field in constant reshuffling such as complexity, 20 years is a long time. The puzzle becomes badly posed. What for?
Let us accept to take “voluntary action E” as a singular event. It is in fact a composition of two looks: in E there is ‘action’ and ‘voluntary’. In the causal relationship, ‘action’ can be understood with NN, neurons; but ‘will’ can only be understood with B, the brain/whole.
Let’s position ourselves at the level of motor neurons, NNM part of NN that performs the action. They receive a pattern of specific signals and return their own coded signal. They “cause” physical action. But they are not the cause of the master plan. Nor the identification of the action by other neurons. There is a hierarchical organization within NN that makes it perfectly rude and unacceptable to say “SNN cause E”. It’s likening NN to a black box… without saying it. While by talking about the whole B, the brain, we implicitly admit this assimilation and its limits in terms of causality: B causes fusionally, as a whole representing the complexity of NN, the voluntary action E.
Double causality for a double look
It is essential to differentiate the two looks. For ontological, that of interacting neurons, there is a chain of causality sufficient to explain the excitation of neurons by others. Ontological causality, blind to the phenomena produced by this activity. No E is named. For the teleological look there is a causality from the result E. Teleological causality. E is observed as a phenomenon.
In other words, it takes both “SNN cause E (neural excitation)” and “SB cause E (voluntary event)” to complete the causality of E. The puzzle of causal exclusion is distorted by the postulate that if reality is ‘one’, the causality of each of its events is ‘one’ too. Not. Our mind detaches itself from reality, produces a causality of the other end of its complex dimension, which is not ontological. Since the reductive postulate is already present in the causality used by Kim, his reflection necessarily leads to a reductive conclusion.
Yet Piccinini does not present himself as a reductionist. On the contrary, he continues with a meticulous critique of the two reductionisms attached to each isolated look: the ontological which is materialism –everything comes from the quantons– and the teleological which is idealism –everything comes from the whole universe.
« [idealism] shares with atomism the problem that picking any level as a reduction base seems arbitrary. Aside from the puzzle of causal exclusion, it’s unclear why any level should be more ontologically fundamental than any other. And the puzzle of causal exclusion is entirely neutral on which level is more fundamental. It could be the whole universe, the level of fermions and bosons, the level of brains and other mid-size objects, the level of molecules, or any level in between. »
Piccinini sends back to back idealism and materialism. Concerned with respecting the importance of each scientific discipline, he refuses even to define the slightest causal priority at a level of reality. Which is excessive. It also means refusing to take into account the causal direction we feel about the world —it goes from the simple to the complex. It is also to refuse the History of knowledge, made up of a frank success of ontological science, opposed to the unproductive holism of idealism.
Against anti-reductionism too
The trial of the two reductionisms is followed by that of anti-reductionism. Here Piccinini’s condemnation is without appeal: the notion of strong emergence does not hold at all, he says:
« It is unbearably mysterious how, after a plurality of objects has become organized, the whole could possess causal powers that are lacked by the organized parts taken collectively. Yet this is what strong emergentists claim. Needless to say, no one has been able to make any sense of how this could be—how a higher-level property could possibly endow an object with causal powers that its lower-level realizers—considered collectively at the same time—lack. Or, if a strong emergentist were to deny that higher-level properties are wholly realized by lower-level ones, no one has been able to make sense of where the unrealized portion of the property comes from, or how it relates to the properties of the object’s parts. »
Again it is incomprehensible that Piccinini, after having put forward the idea of relative independence between levels of reality, should return to such a summary reductionism. The strong emergence is not “unbearably mysterious”. It is based on a simple observation: a multitude of states of the lower level can produce a stable, uniform property, as whole. That is, the whole is recognized as a single element, by entities capable of interacting with it, simultaneously with the fact that its parts form a host of different sets. A multitude becomes “one”. The whole merges, or approximates, all its possible varieties of constitution.
Of course, using a reductionist look it is impossible to grasp the difference, since the whole is reducible to all its parts (for ontological reduction) or the parts are reducible to the whole (for idealistic reduction). But anti-reductionism, precisely, is to refuse these two types of reduction. Forming a whole leads to a reversal of the look. A frenetic change in the state of the parts (as seen by them) does not alter the state of the whole (seen by the other wholes).
The argument of multiple realization, described by Piccinini as the most used by anti-reductionists, is a simple consequence of the relative independence of levels. This is not an argument for strong emergence, because it does not need it. The strong emergence is indeed a leap in the organization of reality because if it is possible to say to which everything leads a set of parts, a contrario it is impossible to say which is the set of parts from the whole. The upper level “buries” the lower level under its approximation. It is a “step” of the complex dimension.
One level to direct them all!
But then, if Piccinini denounces anti-reductionism after pointing out the shortcomings of both reductionisms, what is left for him to build his theory of reality? Well, he will return to good old dualism, with our virtual thoughts on one side, concrete reality on the other. To rid reality of its complications, he brings it down to a single authentic level, the rest being pushed back into our mind:
« To get out of the reductionism vs. anti-reductionism morass, it is tempting to get rid of multiple levels altogether. If we didn’t have multiple levels, we wouldn’t have to worry about which one is fundamental and which ones are not. […] The multiple levels might be a feature of our thoughts and descriptions, not reality.
If there are no multiple levels of being—no layers of objects standing in composition relations, no layers of properties realizing one another—the next question is what our discourse refers to; what makes it true or false. The simplest answer is that there are, in fact, some objects and properties in the world. They are just not organized in multiple levels of composition and realization. They form a single genuine level. »
Radical reductionism from which Piccinini wants to extricate himself by a pirouette: the levels of reality are perfectly safeguarded in the mind. It is only in concrete reality that they disappear completely, also erasing the problem on a fundamental level:
A subtle difference
« This single level picture sounds a bit like reductionism, but there is a subtle difference. Reductionism retains many levels as real, whereas the single level picture eliminates them. Reductionism claims that there are multiple genuine levels of reality: our talk of neurons, brains, etc., and their properties does refer to real things—it’s just that they all reduce to the one fundamental level. By contrast, the single level view claims that there are no multiple levels at all—no genuine neurons, brains, and higher-level properties: our talk of such things is just shorthand for talk of things at the one and only genuine level. »
From now on, Piccinini houses in the mind the whole hierarchy of levels. He makes them aspects, proprietors of our thoughts. To introduce this solution, he discusses the relationship between causality and property levels:
Categoricalism vs. dispositionalism
« What matters most here is the relation between properties and causal powers —the powers that allow objects to do things. There are two simple views.
Categoricalism is the view that properties are qualities. Qualities, in turn, are ways objects are that do not, by themselves, include any causal powers (Lewis 1986; Armstrong 1997). Paradigmatic examples include being square versus round or being a certain size. In order for objects to do things, their qualities must be paired with natural laws. Objects do what they do because they obey laws in virtue of the qualities they possess. If we could change the laws, we would change objects’ behavior—even if their qualities remained the same.
Given categoricalism, then, causal powers are not basic—they are byproducts of qualities and laws.
By contrast, dispositionalism is the view that properties are dispositions, or causal powers (Shoemaker 1980; Bird 2007). Paradigmatic examples include being soluble or fragile. In order to do things, all that objects have to do is possess causal powers and encounter conditions that trigger their powers’ manifestations. Laws are not needed. In fact, law statements are just descriptions of how objects behave in virtue of their causal powers. Given dispositionalism, causal powers are basic while laws are derivative ».
This description is simply that of ontological (dispositionalism) and teleological (categoricalism) views that are complementary and not contradictory. By making them incompatible, Piccinini is forced to choose between different solutions, illustrated below:
From left to right: 1) Identical causality between levels. 2) Aspect (the causality exercised by SB is an aspect included in SNN). 3) Low emergence (causality of SB partly independent of SNN). 4) Strong emergence (causality of SB including that of SNN).
The platist school
All this, as reflected in the figure, is extremely horizontal and traps causality in set theory: one causality is included or not included in another. The complexity of things is flattened. We could call this way of treating causality “the platist school of reality“. But let us remember that all mathematical formalism is based on postulates and has its counterparts, based on contradictory postulates, and sometimes concretized at other levels of reality. Thus Euclidean and non-Euclidean geometries are equally existential. Let us therefore beware of analyzing causality with an epistemic tool less fundamental than it, such as set theory.
Unfortunately this is the path taken by Piccinini, only following many authors:
The influence of set theory
« The aspect view is inspired to an extent by the so-called subset view of realization. Roughly, the subset view is that properties are (or “bestow,” or “confer” on objects) sets of causal powers, and higher-level properties are subsets of the causal powers of their lower-level realizers. Versions of the subset view have been defended by several authors (Wilson 1999, 2010, 2011; Shoemaker 2007; see also Fales 1990; Clapp 2001; Polger and Shapiro 2008; Watkins 2002; Yablo 1992). »
We are well in the heart of set theory. After Piccinini’s good start, I confess my deep disappointment at these gesticulations motivated by the lack of tools to manipulate the complex dimension, hence the need to erase it. This scrub comes at the price of an exclusion by the author of his own mind. How to include it in a reality made of aspects? Who or what becomes aware of these aspects? It must be something external to reality. The difficult problem of the phenomenon cannot be solved in this way. It was swept into an alternate space. An impassable boundary between real and virtual has been created. It is impossible to define an aspect, nor an illusion, in a monistic universe. It is indeed a question here of a real reductionism.
Frozen time… At what level?
This reductionism continues when Piccinini tries to freeze time to find a synchronous composition between the parts and the whole. “The instant”, a notion as mysterious as causality, is not defined here. While its value seems abysmal different between the quantum instant and the instant of a neural synchronization, Piccinini also flattens it into “a” instant common to all levels. Admittedly, the reflection is simplified, at the cost of being against the current of the disciplines of complexity, which gives back to the interacting elements the ownership of the rules followed by the system, including its elementary time. More annoying, it is also against the current of phenomenal time, the same abyss separating that of our mind from that of quantons.
To erase the problem of identity between levels of reality of a thing, Piccinini uses invariants, and takes the example of a statue:
The bronze statue
« Consider the classic example of the statue and the lump of bronze that constitutes the statue. Are they one object or two? Given that they occupy the same space, there seems to be just one object. But let’s suppose the original lump was later molded into the statue. Given that the lump existed before it took the shape of the statue, there seem to be two distinct objects. What gives? If we reframe the situation in terms of invariants, the puzzlement vanishes.
Squashing a statue destroys it; breaking a tiny chip from a statue does not. That is, statues are invariant under chip removal but not under squashing. Why? Because breaking a tiny chip from a statue leaves much of the statue’s shape intact, whereas squashing the statue obliterates its shape. Since one of the properties we use to identify and re-identify statues is their shape and shape is invariant (enough) under chip removal but not under squashing, we say that having (roughly) a certain shape is essential to being a statue. That’s why being a statue is invariant under small chip removal but not under squashing. »
The example is wrong. Suppose we remove one shard after another from the statue, until it disappears completely. From what shard is the statue no longer one? Very difficult to raise your hand. Each observer will do so at a different time. The invariance of the statue is a fuzzy notion, unrelated to its physical definition. In physics, invariance concerns a simple property evaluated in a specific context. While ‘statue’ is a complex property that takes on different values in people’s minds. Piccinini is trapped by the deliberate erasure of the complex dimension.
So how would it be possible to adopt his ‘egalitarian ontology of levels‘, which he summarizes as follows:
Egalitarian ontology of levels
« Neither wholes and their properties nor parts and their properties are more fundamental. Instead, wholes are invariants under certain transformations of their parts, and their properties are aspects of their realizers. This egalitarian ontology explains how multiple levels are causally efﬁcacious without being redundant. »
Critical flaws: The whole maintains its properties within certain limits of the properties of the parts, and these limits are blurred; No synchronization between levels. Who or what takes the aspects into account since reality has only one ontological level? How does the universe work in the absence of the human mind to consider these aspects? Or: how does the material universe manage on its own to use aspects in its organization?
The flat way
These questions are insoluble with egalitarian ontology, from which we are excluded. The “platist” misdirection, which I also call “horizontal thinking”, can be summarized in the following steps:
1) The complexity of reality as hierarchical levels poses insoluble problems, both for ontological and teleological reductionism.
2) Anti-reductionism seems to bring out the emergent causality of nothingness.
3) It is therefore necessary to get rid of this complexity, to reduce it to an epistemic aspect.
4) The real in itself becomes one-piece (a block). It is our mind that sculpts its levels and makes it a hierarchy, when in reality all aspects are in the block.
Truth be told, Piccinini is not wrong to say that most of our minds see things that way. The conceptual categorization of our conscious space is indeed a horizontal organization. There accumulates a host of representations whose hierarchy is perfectly arbitrary. Hierarchy of importance rather than causal. Soccer Ligue 1 results can be superior to the latest philosophical essays and scientific experiments. All these mental images are juxtaposed and their possible causal links are rarely transparent. They arise from the unconscious without being accessible.
Representation and organization of representation
This is why it is essential not to confuse the conscious mental scene with the organization of the mind that produces it. The stage is horizontal, flat; the organization is hierarchical. If the levels of this hierarchy do not possess an existence, and a causality , at least relatively independent, why can my consciousness not access their workings? Why can’t I correct a failed gesture perfectly?
This proven relative independence is a frank, direct denigration of a real one-piece. Everyone can have this experience: if I am part of a real one-piece, I should be able to experience myself in all my aspects, quantum, atomic, molecular, cellular, unconscious, conscious. But I can only experience one of these aspects, the last. And for what or who does this phenomenon, this specific experience, appear, if I am in the block? To whom am I supposed to show my soul to dispose of it?
Relief with Surimposium
Surimposium is an anti-platist theory. On the contrary, it makes complex hierarchy the fundamental dimension of reality. Ontological and teleological visions fit in without embarrassment. It is the double look: starting from distant points in the complex dimension, the micromechanisms constitute things, the mind represents them. The proper identity of each thing is proven when the two looks coincide.
The “points” of the complex dimension are attractors around which system stabilities are formed. Systems of related elements establish their own rules and entangle their causality at previous levels. No fundamental laws, no universal framework, only a meta-principle linking the laws and frameworks proper to each level of reality. Each transition from one level to another is a source of a new phenomenon. No need for the human mind for phenomena to appear. They are already recognized by matter itself. The peculiarity of the human mind is a depth of processing by neural networks such that it creates an incredible amount of levels. The phenomenon formed, consciousness, oscillates between a cottony emergence and this bluffing acuity when we are well awake and connected.
Thanks to its complex vertical dimension, and its two causal directions, Surimposium is a shelf library where existing theories are housed without them interfering with each other. The only thing they lose is their hegemonic claim. Unfortunately Surimposium, despite its 800 pages, has far fewer references to classical thinking than Neurocognitive Mechanisms, and is sorely lacking readers capable of understanding it.
Excellent ontological work
I end this reading of Piccinini in a much more positivist way. Despite his conceptual misdirection, Piccinini ultimately treats cognition as a hierarchical system. Having reduced levels to aspects does not prevent him from taking them into account in his neuro-computational approach to cognition. Piccinini is in line with the most recent concepts: neural computation is neither digital nor analog, but specific to neurons. It is the result of metabolic, hormonal, electrochemical levels, associated with rules of delay, synchronization, synaptic weight. Here are Piccinini’s final conclusions:
1. There are many levels of mechanistic organization. Parts compose wholes; properties of parts realize properties of wholes.
2. All levels are equally real. Higher-level objects are invariants under some changes in their parts; properties of wholes are aspects of their realizers (Chapter 1).
3. Higher-level functional properties are often multiply realizable, and some multiply realizable properties are medium independent (Chapter 2).
4. Some mechanisms, including cognitive and computational mechanisms, have teleological functions, which are regular contributions to goals of organisms (Chapter 3).
5. Some properties are functional—deﬁned in terms of the functional organization of a mechanism. Functions and structures are not separable: they mutually constrain one another (Chapter 4).
6. The original Computational Theory of Cognition states that neurocognitive processes are digital computations (Chapter 5).
7. Digital computation is one kind of computation among others, including but not limited to analog computation. Computation and information processing are medium-independent mechanistic processes (Chapter 6).
8. Constitutive explanation is mechanistic (Chapter 7).
9. Cognition is constitutively explained in terms of multilevel neurocognitive mechanisms. This is the type of explanation cognitive neuroscience seeks to provide (Chapter 8).
10. Neurocognitive processes are computations that process information (Chapter 9).
11. The Computational Theory of Cognition does not follow from the Church–Turing thesis (Chapter 10).
12. There are no compelling objections to the Computational Theory of Cognition (Chapter 11).
13. Neurocognitive processes operate on representations (Chapter 12).
14. Neurocognitive processes are neither digital nor analog—they are a sui generis form of computation (Chapter 13).
15. Consciousness may or may not have a wholly functional nature. Even if consciousness does have a wholly functional nature, its functional nature may not be wholly computational (Chapter 14).
Piccinini’s ontological work is remarkable here. It is really a pity that it has closed on this single direction of the look, because it lacks in the end the explanation of the phenomenon consciousness. Computational theory is not enough. It lacks this reversal of a level of reality on itself allowing it to test its constitution, in an independence that can not be a simple “aspect”. For this impression is what is most substantial in conscious experience, persisting even ridding it of all content, of all representation. The complex hierarchy must be elevated, not flattened, to reveal an experience of this nature.
Nevertheless, I salute the originality of Gualtiero Piccinini’s thought, which achieves a coherent synthesis of classically conflicting approaches, both among philosophers and scientists. Can it close the breaks? I hope he’ll make a response to this criticism which is not sterile, but based on an alternative.
Neurocognitive Mechanisms: Explaining Biological Cognition, on The Brains Blog, author’s website
Neurocognitive mechanisms, reviewed by Matteo Colombo