World, Model and Mind
The Magical Flower of Winter is an essay series exploring reality and our relationship to it. It deals with philosophy, science and our views of the world, with an eye on the metacrisis and our future. Sign up to receive new essays here:
…the mind and the world jointly make up the mind and the world.
Hilary Putnam - Reason, Truth and History
In the account of what science is in Science and Explanation I took as my starting point that
an essential part of experience is expectation, that an inseparable component to normal experience is the effort to predict what is to come.
In this essay I want to close the hermeneutic circle opened there, and to review a few of the leading scientific models of how the mind and our world is related to our brain. Given what we have learned so far in this project about the relationship between reality and representations of it, we can better understand these approaches and what they can and can’t say about the world and our experience of it.
Predictive Coding and Active Inference
A leading approach to how the brain works advanced by a large group of neuroscientists1 is a scientific model of brain and world built on «a form of radical neuro-representationalism - radical because the claim isn’t simply that our access to the external world is mediated by neural representations, but rather that the world of experience is itself a representational construct.»2 The claim is that the relation between brain and world is not the naive one where our senses and brain passively lets the world through to our experience, that experiencing is like looking out at the world through a transparent window, but one where experience itself is generated by our brain, that the world we see is a creation of our brain, to which the external world is integrated through sensory data and error correction. This general approach goes by several names, most prominently predictive coding, predictive processing and active inference. The reason why our world is a generated one in this view is exactly because «an inseparable component to normal experience is the effort to predict what is to come»: what we experience is an integration of what the world model in our brain predicts the world to be in the immediate future given the past, maintained by incoming sensory data. On this view we are not experiencing the world, but our model of it3. That this generative nature of our experience is transparent to us is what Metzinger (2009) calls the ego tunnel. But what about the external world? How can we know or say anything about it if our only access to it is through a model of it? This dualism should be familiar, the external world has once again become some inaccessible reality beyond our experiential (and generative) powers to reach. In order to untangle this dualism we need to get an understanding of the Bayesian dimension to neuro-representationalism.
The Bayesian Brain
Neuro-representational approaches like predictive coding and active inference are commonly implementations of a conception of the brain as Bayesian. This is the hypothesis that what the brain is doing is an approximation to Bayesian inference: the brain implements a mechanism by which sensory information acts to correct and update prior likelihood distributions about how the world is into posterior likelihood distributions. These likelihood distributions are as such models of what the world is like and constitutes the generative aspect of this paradigm. In simpler terms, the “prior distribution” is an expectation about how things may turn out before an event, while the “posterior distribution” is the expectation about how things may turn out after an event, where the “event” is some occurrence that leads to information the Bayesian mechanism can use to update its model (distributions) in line with Bayesian statistics (minimizing surprise or variational free energy, in the language of active inference). The updated posterior distribution will in the next time step be used as the prior distribution, i.e. the new expectation given the new data received about the world, and in such a way the Bayesian brain is a continuous and dynamic effort at accurately modeling the world.
Because the Bayesian-generative brain’s priors are always based on previous posteriors, our model of the world is always based on the empirical - we are always in it. I have previously used4 “in it”-ness to highlight the immersive aspect of experience, which is just what the model of the Bayesian brain leads to as well: we are eternally bound to our dynamic experiential context. There can be no “clean slate” or “view from nowhere” that acts as an objective divider between the a priori and a posteriori, between the theoretical and the empirical5. An additional facet that this conception of the brain itself highlights is the sharp distinction between the probabilistic model of the “laws of reality” that the brain represents, and the “actual laws of reality”. Our Bayesian model is only an epistemic approximation to “natural law” based on limited evidence, and it is not necessarily primed towards maximizing accuracy, for the evaluative standard against which accuracy would be measured also shifts continuously. Hoffman (2019) has developed a «theorem» that shows that from neuro-representationalism combined with evolutionary theory, the brain optimizing for fitness is no guarantee that it accurately represents reality, i.e. that optimization for survival and fitness can occur at the cost of inaccurate representation. How are such consequences to be interpreted?
Neuro-representationalism starts out as a materialist realism: there is an independent material brain which generates experience and the world we see, based on sensory data from an inaccessible external reality. But this representation of things must also be embedded in this model, because the model of a brain generating experience enclosed in a world is a representation that on this view results from just such a model. But how can we then ever go beyond our model? How can we even posit that a brain is responsible for experience, if we are led to the brain also being just a representation in our model? If we are enclosed to a model, what justification is there for positing an external world? We have gone in a strange circle and arrived at some form of idealism, not materialist realism where we started6. We also see here an example of how the hermeneutic circle is unavoidable: we are immersed in reality, and once we try to explain this immersion a leap must be made, some background must be assumed, for otherwise we have no ground to stand on at all7.
Self-Reference and Paradox
Self-reference and “loopiness” becomes unavoidable when that which is represented is also the ground for representation itself. In Gödel, Escher, Bach (1999) Hofstadter deals extensively with recursion and self-reference in relation to science and mind, and perfectly exemplifies the loopiness inherent in reductionist or emergentist approaches. A central component of the work is stepping out of the system. It is only by stepping outside of a given system that we can interpret the system, and it is by interpretation that meaning arises. Starting with some finite formal or logical system, Hofstadter finds that novelty arises when this system can implement or refer to itself, and he hypothesizes that consciousness requires self-reference and self-modification, exemplifying systems that are self-referential “tangled hierarchies” and “strange loops”. But what if this self-reference and tangledness is an artefact of our representations, and not of the reality they are supposed to represent? What if self-reference and having to step outside of systems to generate meaning are pointing towards our conceiving of things the wrong way around? Hofstadter is inevitably shaped by the particularist paradigm, where consciousness must emerge from the material, and down this path he finds a self-referential framework that he takes as a plausible path to explain experience. I keep coming back to the prevalence of confusing our model of reality for reality: self-reference is nowhere to be found in our experience, for from experience always being novel and unique, any reference to “self” will always be a reference to a different version of “self”. The very act of uttering “I” changes the “I”, closed self-reference is as such only an aspect of our abstraction, of our model of what is going on.
In Beyond the Limits of Thought (2002) Priest has developed a general schema (the “inclosure” schema) for paradoxes and contradictions due to self-reference, limits and infinity. The general form of contradiction comes about from the existence of a property that is both had and not had by a totality, often due to a form of self-reference. A much simplified, but illustrative example is the infinite application of “the thought of X” to all thoughts X. Will the thought of the infinite series contain itself? Because we are dealing with infinity, the answer is not yes or no, but yes and no!8 When we attempt to break out of an infinite totality, there is in a sense nowhere to go, so we are both inside (“closure” in the terminology of Priest) and outside (“transcendence” in the terminology of Priest) the totality simultaneously. Notice also how in order to ask the question whether the infinite series contains itself we have to somehow have stepped out of the system already, as Hofstadter focused on. Wittgenstein remarked on this: “in order to be able to draw a limit to thought, we should have to find both sides thinkable (i.e., we should have to be able to think what cannot be thought). It will therefore be only in language that the limit can be drawn, and what lies on the other side of the limit will simply be nonsense.”9 There will always be a tension between “inside” and “outside” when discussing representations, because we will always stand “outside” them, this is in the essential nature of them being representations we create and manipulate.
We should also consider the reductionistic and representational origin of these paradoxes and limits. Any and all attempts to represent the infinite or some totality inevitably reduce it to a sign or signs, from which the infinity we want to represent cannot be recovered. Furthermore, self-referential representations lead to an infinity that isn’t there in reality, it is merely that what we attempt to represent isn’t representable due to being a whole. By representing a whole we inexorably take something away from it, and if what we are trying to represent is supposed to have as its function that it itself is capable of representing itself, the loss from reduction is compounding and cannot be amended. This is what we see when we conceive of experience as representable by a brain: this is itself a representation, but experience is a whole, and by reducing it we are at a loss.
As explicated in previous essays10 the mind-body problem cannot be solved from a particularist11 world view, and the very existence of the problem is a consequence of the central claims of the view itself. To escape the closed loop of representationalism one needs to abandon the hope that our attempts at representing reality can reclaim reality as a whole, and that language maps onto reality in a clean and separable way. Bergson (2007) also saw this: we cross notations in trying to speak of mind in material language, and further, by isolating the brain as the substrate of experience, we have removed the brain from the whole of which it is part in experience. World and experience is an inseparable whole, the breaking down of which in an attempt to explain one in terms of the other leads to irresolvable trouble. Furthermore, the language of our experience does not translate invariantly to the language of physical reality: the signs do not correspond to each other one-to-one, such that a division of an idea of a thing into parts, do not correspond to the parts of the thing. The independent physical world must be seen as an ideal limit. We can talk about and agree about a shared world, but its existence is due to our talking and agreeing about it. It is the kaleidoscopic union of all our worlds, and it is neither the cause of all our worlds, nor is it prior to them.
The Computational Theory of Mind
…I began to understand to what great extent our language constrains the world, arranging it and placing its various elements in logical systems that are of such nature that we see neither the system nor the logic, only the world it presents to us.
Karl Ove Knausgård - The Morning Star
Predictive coding and the Bayesian brain, along with other approaches like Integrated Information Theory12 and Global Workspace Theory, are considered to be the leading theories towards the explanation of mind. What they have in common is that they are all based on the computational theory of mind (CTM), that mind is somehow reducible to computation. CTM is closely linked with the Language of Thought Hypothesis that thinking occurs in a “mental” language13. We have already seen some issues with such an idea, in the sense where such a mental language is private, in Wittgenstein and the Private Language Argument. Putnam (1991) critiqued this mentalism on three grounds: meaning is holistic, meaning is in part a normative notion and concepts depend on environment, i.e. are contextual. These imply the irreducibility of meaning to mental representations, as part of what makes meaning what it is exists outside any and all of its representations, whether physical or mental. Putnams aim is part of illustrating a philosophical attitude I fully agree with, an attitute «that gives up many traditional assumptions about Appearance and Reality; that gives up, for example, the assumption that what is real is what is "under" or "behind" or "more fundamental than" our everyday appearances, that gives up the assumption of The One in the Many, and that also gives up the assumption that every phenomenon has an "ultimate nature" that we have to give a (metaphysically reductive) account of.»14
The attentive reader might further ask whether CTM, generally being a reductionistic approach, can truly answer any foundational questions about experience and mind. Isn’t the computational theory of mind embedded in the particularist paradigm of independence, dualism and reductionism? How can we expect to extract an explanation of mind from a framework that has excluded mind from its presuppositions? What we can at most hope for from these approaches are an understanding of correlations, i.e. that certain patterns of neural activity correlate with certain aspects of experience15. But explaining experience and mind will forever be external to an approach that has built into its foundations that the thing we want to explain is of a different kind than the things we use to explain. This shouldn’t be called a hard problem or even an impossible problem, but a meaningless problem. We can’t get or explain oranges from using apples.
As is evident, this essay is a continued critique of the particularist paradigm, here manifesting as the mechanistic and computational machine model of experience16. The critique is once again not an attempt at showing the paradigm to be wrong, but at showing it to be limited, and at showing a larger whole fully coherent with our experience of reality from which this limitedness makes sense. Science is an approximation to the order of nature and our experience, while its dis-order, the fullness and abundance of experience, is irreducible due to epistemisation17, experience overflows any and all logic18. I want us here to contemplate the meaning of disorder as un-order, as that which cannot be brought into a lawful form. Un-order is utterly holistic. In the essay Experience I will continue to build on these conclusions towards a view of reality as a whole.
Thank you for reading! If you enjoyed this or any of my other essays, consider subscribing, sharing, leaving a like or a comment. This support is an essential and motivating factor for the continuance of the project.
References
Bergson, H. (2007). Mind-Energy. Palgrave Macmillan. [1919]
Clark, A. (2016). Surfing Uncertainty: Prediction, Action, and the Embodied Mind. Oxford University Press.
Fodor, J. A. (1975). The Language of Thought. Harvard University Press.
Frith, C. (2007). Making Up the Mind: How the Brain Creates Our Mental World. Wiley.
Hoel, E. (2023). The World Behind the World: Consciousness, Free Will and the Limits of Science. Avid Reader Press / Simon & Schuster.
Hoffman, D. D. (2019). The Case Against Reality: Why Evolution Hid the Truth from Our Eyes. W.W. Norton.
Hofstadter, D. R. (1999). Godel, Escher, Bach: An Eternal Golden Braid. Basic Books. [1979]
James, W. (1909). A Pluralistic Universe. Longmans, Green & Company.
Knausgård, K. O. (2020). Morgenstjernen [The Morning Star]. Oktober Forlag.
Lange, F. A. (1925). History of Materialism and Critique of Its Present Importance, trans. E. C. Thomas. Kegan Paul. [1865]
Metzinger, T. (2009). The Ego Tunnel: The Science of the Mind and the Myth of the Self. Basic Books.
Parr, T., Pezzulo, G., & Friston, K. J. (2022). Active Inference: The Free Energy Principle in Mind, Brain, and Behavior. MIT Press.
Priest, G. (2002). Beyond the Limits of Thought. Clarendon Press. [1995]
Priest, G. (2006). In Contradiction: A Study of the Transconsistent. Clarendon Press. [1987]
Putnam, H. (1981). Reason, Truth and History. Cambridge University Press.
Putnam, H. (1991). Representation and Reality. MIT Press. [1988]
Seth, A. K. (2021). Being You: A New Science of Consciousness. Faber & Faber.
Wittgenstein, L. (1972). On Certainty (G. E. M. Anscombe & G. H. Wright, Eds.; G. E. M. Anscombe, Trans.). HarperCollins.
Wittgenstein, L. (2001). Tractatus Logico-philosophicus (D. F. Pears & B. McGuinness, Trans.). Routledge. [1921]
Zahavi, D. (2017). Husserl’s Legacy: Phenomenology, Metaphysics, and Transcendental Philosophy. Oxford University Press.
E.g. Clark (2016), Frith (2007), Hoffman (2019), Metzinger (2009), Patt & Friston (2022), Seth (2021).
Zahavi (2017) p. 187.
Frith (2007).
See Science and Explanation for more on this “boundary”.
Zahavi (2017) p. 191-2 attributes this realization to Lange (1925).
Wittgenstein discusses this in On Certainty (1972): Certain propositions or assumptions are part of our «frame of reference» (context). Doubting these removes the grounds for all judgement, and as such these claims are not subject to doubt, because in doubting them we undermine the very basis for doubt itself. But these "background" claims are of course variable, they are not a "solid ground", but a varying background.
Priest (2002, 2006) embraces these contradictions (“yes and no”) and builds a “dialetheic” logic on true contradictions. These are forbidden in “normal” logic which obey the law of the excluded middle (“yes or no”).
Wittgenstein (2001), Preface.
I.e. materialist, reductionist and realist.
See Fodor (1975).
Putnam (1991) p. 4.
Neural Correlates of Consciousness (NCC).
See Artificial Intelligence and Living Wisdom for a critique of the machine model of intelligence.
James (1909).