Another Round with Hoffman — Is Consciousness Primacy Incoherent?
Last post I discussed some good things and some weird things about Donald Hoffman's consciousness-based metaphysics. To quickly recap: placing consciousness as the fundamental reality is one way to do away with the Hard Problem of consciousness (the qualae problem basically). But it's a non-solution. It is really, in a simplistic but fairly robust sense, the inverse of the physicalists denialism. You either deny subjective conscious reality exists, (Dennett and denialism) or you deny physical reality exists (Hoffman's neo-Idealism).
Hoffman himself does not like to associate his ideas with classical idealism, but at the core there really is not much difference. Classical idealists did not have the apparatus of quantum theory to support some kind of non-determinism in nature (necessary to argue against physical determinism*). Hoffman does. But apart from this, combined with modern evolutionary theory and computer science, Hoffman is not saying anything fundamentally different to the classical idealists. All metaphysical idealists have implicitly adopted some non-physical ontology. Most were unable to develop a generally accepted "theory" (you cannot truly get theories in pure philosophy, so I use the term loosely, once you can mathematize your ideas they become proper science, a special branch of philosophy, not philosophy proper).
There are alternatives between these extremes of denialism, and I think a lot of rich rewards can be gained from taking a more agnostic point of view. In the end, one can always use semantics to come down hard on one side or the other, but it is not always helpful to do so, it is alienating and too rigid to rightfully go under the banner of philosophy. To his credit, Hoffman is not really an extremist. He is well aware and open about the fact his ideas are speculative and cannot be proven. He has not entirely lost scientific objectivity, even though paradoxically his whole metaphysical construct imagines there are no objectively real "objects". this is a curious sort of condition to be in philosophically, and I would love to explore it, but if feels too indulgent and pointless. If you can hold contradictory beliefs together then good luck to you. Some good can come of it too, but usually in the form of eventually abandoning one of the contradicting assumptions.
Today I will write a little bit about Hoffman and Prakash's supposed "proof" that evolutionary theory "predicts" organisms adapt for fitness traits, not for traits that gain "truth". Then I will finish with some notes on whether consciousness-based reality can even be coherent.
The Evolution Argument Problem
The idea is that evolutionary game theory can be used to show organisms are adapted for fitness, and this can be counter to being adapted to perceive the truth about reality. The idea is that evolution does not necessarily drive our conscious experiences towards "veridicality". Hoffman and Prakash draw from this the idea that we do not in fact perceive reality.
One problem with their ideas is that they draw a non-sequitur: they think they have proven conscious perceptions cannot be representing any objective "truths" because consciousness is not evolved to discern truth. Well, I do not buy that conclusion.
Why don't I buy it? Veridical perceptions are not disjoint from fitness perceptions! Indeed, although there might be vanishing homomorphic correspondences between perception and reality. To draw their conclusion you would need to also prove (which they have not) that the structures of our perceptions are nothing like reality. Indeed, it is clearly plausible that two dissimilar structures with no homomorphic correspondence can be as alike in infinite ways as we care to devise. Indeed, Hoffman's own analogies show us how. He is fond of the graphical computer interface metaphor: the structure of icons for files on your desktop computer display bear no relation to the structure of the binary storage bits that actually comprise the information in the files pointed to in the computer registry.
But I would say, "so what?" The icons are real. The underlying bits are real. They are different representations at different levels. Moreover, while there is no direct homomorphism from desktop icon onto binary storage, there are nonetheless several mappings from one to the other which convey aspects of veridical truth. The "fitness" of the desktop icons does not forbid veridical perceptions of the existence of binary bits somewhere in the computer memory. For sure, one can generate a "fake icon" representing a text file for which there is no text, but the icon is always generated by bits in the memory. Just observing the icon tells us (allows us to reasonably infer) there are bits underlying it.
It is just silly to think that lack of homomorphism is lack of veridical perception. Consciousness does not require mathematical homomorphic correspondences in this sense. The mathematics is misleading Prakash and Hoffman. They have over-thought things and placed a mathematical structure onto their thinking that is needlessly rigid. This is not surprising coming from well-educated academics. In recent history academia has become infested with a perverse desire to mathematize everything. It is the old aping of physics pathology, "physics envy". While useful, it is not always healthy (witness the economics profession, which has rigorous mathematical models, and yet none of them correspond to real world economies).
There are other problems with Hoffman & Prakash.
The proof by Prakash and Hoffman has nothing to do with what most people understand by "consciousness". That's the first problem. They assume a model of a "conscious agent" but this is mere word play. There is no subjective element in their model. They use measure theory on a set they presume captures the notion of conscious states (which is highly dubious), a Markovian kernel, and a probability measure. This gives them a model for supposed conscious agent behaviour, but it has nothing at all to do with two of the essential attributes of consciousness: qualae and mental abstraction (human mathematical ability for example).
In the main argument used by Prakash and Hoffman, which claims natural selection does not evolve towards veridical perception, but rather towards mere fitness, they think the vanishing probability of precise homomorphism from perceptual data structures onto the "real world" to be proof that we are not perceiving the real world in any truthful fashion. This is a gross logical error. It assumes (implicitly) that one needs a homomorphic map to gain true knowledge about the world. That's just not a valid assumption. Information theory and error correction, suggest quite the opposite assumption is justified. A system capable of inference can gain true facts about the world from filtered and distorted data structures. This is the basis of some fairly trivial lost information retrieval techniques in neural networks. One can even implement such retrieval in a few dozen lines of python code probing an SQL database (no fancy neural net required). Our brains are surely capable of vastly more sophisticated verisimilitude.
Prakash et al claim,
To our knowledge, there are no arguments for veridical perception based on genetic drift, pleiotropy, linkage, or constraints from physics and biochemistry. Such arguments seem unlikely. It is hard to imagine how neutral drift, for instance, could favor veridical perceptions.
Well, I've just given a very concrete example of how simple computation can lead to veridicality.
They also claim,
One might object that many payoff functions are close to being homomorphisms of the structures of the world in, say, the sense of a L 2 norm, and thus that natural selection will shape perceptions to be close to veridical, if not precisely veridical. We reply that they will also be close to being homomorphisms of countless other structures that are not in the world, and thus that natural selection will equally shape perceptions to be close to countless non-veridical structures. There is no argument here for natural selection favoring perceptions that are close to veridical rather than close to countless non-veridical possibilities.
How do they know, "There is no argument..."? This seems a bit pompous and presumptuous. It seems to me a fairly simple argument favouring evolution towards veridical perception is that all "nearby" homomorphisms may form an equivalence class under whatever measure of "closeness" one cares to construct. A structure "not in the world" can therefore still contain information about the noumena that is in our world. In fact, trivially so, it carries information about the equivalence class of homomorphisms for which the "real object" is a representation. If that information happens to be the best "truth" we can perceive, then we are none the worse off for it, our perception is at least perceiving this modicum of truth.
Hoffman could afford to be pedantic and retort: "Yeah, but part truth is not the whole truth." To which I'd say, "So what?" The whole proof that natural selection heads toward fitness but not truth is just logical nonsense without other strong and unjustified presumptions.
I think it's even worse for the Hoffman cause. Several respected evolutionary biologists (I know of at least Steven Rose and Richard Lewontin, but there are others) have compellingly argued that biological evolution is not simply "all in our genes" and that natural selection is only a small part of the story of evolutionary science. Molecular chemistry constraints are a far more significant aspect. So I think arguments based only on natural selection are not doing the work Hoffman requires. You have to include at least some chemistry to get a proper idea of how the mid perceives external sensory data. The effect of chemistry could very well be to shrink the space of perceptions to a vanishingly small set, on which Prakash and Hoffman's veridical homomorphisms could become non-zero measures.
Drawing the Wrong Inferences from Synesthetes
The use of synesthesia to try to argue that our senses are not adapted to veridicality reminds my of Julian Jaynes' arguments about split brain patients and blindsight to argue that humans were once not conscious! Neither type of argument holds water. Synesthesia is evidence that our mind is using a perceptual interface. That part Hoffman clearly has correct, it is a rather trivial observation, albeit from third person accounts. But it is not evidence that there is no external objective reality.
Is Hoffman's Brand of Idealism Coherent?
Another problem in a lot of Hoffman's work is that he seems to want to "have his cake and eat it too". What I mean by this is that he and Prakash are using evolutionary theory (which is based on biology, chemistry and physics) to argue that there is no such thing as physics. Are they pulling the rug out from underneath themselves? Or are they clever enough to avoid self-inconsistency and other paradoxes?
I think it is hard to tell. The problem being that their language is confused and intertwines physicalism with idealism sometimes in ways that I find (perhaps others do not?) very difficult to tease apart. They use objective scientific concepts to argue that there are no objective objects. It is a kind of confused Ouroboros proto-theorization. I cannot easily prove they are eating themselves, but that's partly because their thinking is so mixed and confused, switching from physicalism back to idealism at every paragraph.
The best I think I can do is point readers to Hoffman's interviews (there are dozens on the *tubes). In his oral interviews he is much clearer in denouncing objective reality. in his academic papers he uses too much physical science language, and objectively specifiable mathematics, to make any sense of his Idealism. So for now, let's just say I think Hoffman's Idealism is not obviously incoherent, but it is confused.
Perhaps this is a "feature not a bug"? Is not all metaphysical Idealism inflicted with such confusion? If you are going to purport to explain "reality" using the notion that nothing objective exists, that everything is just pure thought, then clearly you will run into problems trying to objectively explain things to other minds. It is never going to work in fact! But that's a feature! You only have to convince others that you are "onto something" and you will get allies, people who think the same way. And since, in your bubble, thinking is all there is to reality, you've in some sense accomplished your objective. If nothing objective exists then you've nothing to explain other than your own thoughts. In this sense I tend to think Hoffman is a neo-solipsist. Taking him seriously, it is quite plausible to suppose he is the only thing that exists. (I'm sure he would object, but by his own logic there is nothing to prevent such a conclusion, since he cannot prove that other minds exist. How so? It is because in his theory there is no objective reality, and his own perceptions are "almost surely" (his words) fooling himself.)
In a metaphysics where our perceptions are assumed to not have any veridical content, you cannot do science and philosophy coherently. I agree with Hoffman that evolution does not automatically tune our perceptions to "truth", but they do no obviously tune us actively away from truth either.
Here we can conceive a counter to Prakash and Hoffman's theorems. With the assumption that our universe is structured (by physical laws) to have fitness perceptions track or parallel (sometimes converge) with veridical perceptions, then the Hoffman and Prakash theorem becomes a mere curiosity with little practical importance for our reality. It describes possible alternative universes where fitness evolved perceptions could diverge dramatically from veridicality.
More importantly I think, there is spiritual merit in not giving up on truth. This is not a scientific objection to Idealism, it is a spiritual argument. Nevertheless, I find it compelling. The idea of a platonic sense of justice, honesty, love, truth, trustworthiness, wisdom, compassion, these are outside of the realm of science, but by virtue of transcending science they are powerful. In our modern consumerist-materialist age, people have largely forgotten the power of abstraction and spirituality, such "archaic notions" having been almost entirely replaced by logicism and utility theory. I am one of those scientists who instinctively rebel against such modernist movements. I think there is room in human civilization for both empiricism and idealism, for both logicism and for spirituality. We impoverish our minds when we fail to admit into our thought, discourse and action all that can be good.
Finally, what does it even mean to have "veridical perception" if there is no external objective reality? It seems like a reductio ad absurdem: first suppose external object exists, then show one arrives at a contradiction, so the supposition must have been false. But I do not think Prakash and Hoffman achieve this. In finding vanishing probability of homomorphisms from objects to our perceptions, they do not gain any contradiction. As I pointed out above, failure to find one-to-one mappings does not prove there are no useful information preserving mappings and does not prove our perceptions cannot capture a lot of the essential truths about a presumed external sensible reality.
Mental model: the equator of a sphere is a measure zero slice from the sphere, but it has a lot of information, indeed, given the equator I can generate the entire sphere.
Is Too Much Made from Natural Selection?
Well, yeah! I've already explained why above. But there is more to say. Natural selection is a very blunt instrument. It purports to explain so much, but in reality explains very little. Why? You have to look at the structure of the theory of NS. It is a theory of selective pressure on (random) mutations and hence adaptation. The problem is this is not a causal theory. The underlying causal theory has to be based upon mircophysical foundations, since in natural evolution the present cannot influence the past, all must be unidirectional in time. But once you have an underling time-oriented causal theory then you lose macroscopic causality, and hence NS becomes an epiphenomenal theory, it is a theory that merely tells a story, but is devoid of actual causal meaning.
To avoid such a trap (the trap of vacuity of natural selection) you have to find a way to get macroscopic causal influence. If the story of NS is to have any significant meaning it must have some extra ingredients that provide some kind of genuine emergent laws that cannot be explained by appeal to microphysics foundations. Hoffman and Prakash fail to provide such meaning to evolutionary theory.
However, this is one area where I think I could help them out. But that's for another essay. Here I want to just point out that if NS is a physical process (we all assume it is) then it cannot be used to derive any theorems about the primacy of consciousness and the illusion of veridical perception. At best one can make reductio ad absurdem arguments of the form: suppose NS is a valid physical theory and consciousness is physics... [fill in some arguments based on some set of facts X] ... therefore "NS is not a physical process", hence it cannot be the case that NS is physics AND consciousness is physics. The argument would be completed by didactically asserting it is more likely of the two alternatives that NS is physical and consciousness is not.
But what Prakash and Hoffman seem to be saying is that NS is not physical either! It is a purely mental construct. So, in my mind, their arguments fail to do the work they want to achieve.
A Slightly Better Computer Analogy
Now I want to sketch a brief counter-argument to Hoffman's basic thesis (the thesis that there is no external objective reality to the world, and that on the contrary, all is consciousness).
The idea counter to Hoffman is to use information theoretic concepts and principles of parsimony. These are surely "mental constructs" that I have perhaps imagined! But they are subjectively real to me, so I will exploit them to infer the existence of an objective external world. (Apart from the reasonable assumption I should entitle myself to make that the books where I read about parsimony and information theory were written by other people who were not my own mental constructs!)
The sketch of this inference is to note that in order for consciousness to construct reality this places a massive burden on the conscious mind. It is much simpler computationally, to assume the structure of things exists in the noumena, in the things themselves. Then in order to perceive reality my mind (or any other mind) does less work, less computational effort, since only a bare minimum of sensory data needs to be relayed and processed by the mind. Most of the information comes from external interaction with the presumed objects.
Now, if we alternatively hypothesise that no external objects exist, then computationally it is far more effort for any mind to construct reality. I do not even appear capable of perceiving that I am making such efforts! The most plausible inference is thus that yes, my mind exists apart from the physical world, but also the external physical world also exists and interacts with my mind. My mind is pretty lazy, it does only what it needs to do to gain some semblance of veridical knowledge about the external world. A semblance is sufficient for fitness. However, competition will obviously enhance this fitness, and more than a semblance of verisimilitude is one way for the competition to advance.
Whether Prakash's theorem is true then depends upon the structure of the laws of our universe: if they permit non-verdical fitness, then sure, our perceptions can be basically tantamount to "all lies". But there is no evidence whatsoever that our universe is structured this way. The more reasonable inference is that the structure of our universe closely matches fitness with veridical knowledge. One form of this inference is this argument from parsimony.
Blinded by Vision?
Like a lot of researchers in pursuit of consciousness I think Hoffman suffers from specialism biases. He puts far too much emphasis on analogies with visual perception, and thus (unknowingly perhaps) discounts a lot of other riches in the study of consciousness. The fact that we suffer optical illusions is not evidence that our sensory perceptions are failing to perceive an objective external reality. If anything it's the opposite, otherwise we'd never discover that we can suffer from visual illusions! What we are capable of dong is using additional mental faculties (non-visual faculties, like reason and scientific investigation) to discover in non-visual ways, that we suffer optical illusions. This point is worth another essay, so I will leave it for another time.
One important such source of richness is the idea that although we are all clearly connected in some way by interaction, we can also be relatively separated. Distinctions are a spice for life, and so at first bat I find is just silly to adopt an axiomatic assumption of denialism of any variety. In particular, I dislike both physicalism and idealism because they close the mind, and they present a false dichotomy. Existence is a One but also a Many.
Footnote
- When I wrote, "quantum theory is needed to support some kind of non-determinism in nature (necessary to argue against physical determinism)" I was being a trifle glib. In my own work I've written a lot about how classical general relativity (thought to be a deterministic theory) can be used to derive non-deterministic time evolution mechanics (and possibly quantum mechanics). The requirements for being able to generate non-deterministic time evolution is the existence of closed time-like curves. Unlike Newtonian mechanics, general relativity admits CTC's. So GR is one type of theoretical framework that shows us QM is not necessary for non-determinism. It is not implausible other classical-like physics frameworks could also generate non-deterministic time evolution without quantum postulates. Of course, any plausible such scientific theories would still need to "explain" quantum effects (account for them somehow and/or derive QM from simpler postulates). Indeed, tantalisingly the allowance for wormholes in GR does furnish us with CTC's and also a spacetime mechanism for entanglement (the ER=EPR Conjecture) and possibly also superposition. How might we generate superpositions from pure GR? One answer would be that the ER bridges that account for entanglement would have to be in some instances traversable bridges (in the ER=EPR conjecture they were originally thought to be non-traversable).