A thought experiment about evolution

(An improved and updated version of this material has appeared in my book Why Materialism Is Baloney. The version below is kept for legacy purposes.)

Do we see the world through distortive glasses? Image source: Wikipedia.

I want to invite you today for a thought experiment. Let us suppose that the key tenets of our scientific, material-reductionist paradigm are all correct. According to this worldview, reality is objective and independent of mind; mind and its conscious perceptions are a by-product of the matter of the brain; and the brain, along with our ability to understand nature, has evolved through natural selection favoring survival of the fittest. Still according to this worldview, life has evolved within a space-time fabric where the interplay of matter and energy gives rise to the set of objective phenomena we call reality. Let us imagine this reality as a collection of objects in the canvas of space-time.

As the first living organisms evolved, they were immersed in the same space-time canvas populated by all the other objects that make up reality: rocks, water, sand, air, other living beings, etc. They also had perceptual mechanisms that gave them indirect access to these other objects: for instance, eyes that allowed them to form internal, subjective images of the objects populating the reality they were immersed in. The game of life consisted in optimizing one’s behavior in the dynamics of all those objects so to increase one’s chances of surviving and reproducing. Now note that, still according to the current scientific paradigm, because a living being only has access to its own internal images – not to the objects populating reality – its choices for implementing its survival strategy are entirely based on those images alone.

The images are constructed according to the architecture of the living being’s nervous system, which is itself, as postulated, a result of evolution through natural selection. An obvious question is thus: What would the optimal mapping between objects and subjective images be so to optimize survival? A mapping between two spaces – the objective space of objects and the mental, subjective space of images – can, mathematically speaking, assume infinite forms. One of these possible forms is the identity mapping: to each object in the space 'out there' corresponds a unique, analogous image in the subjective space 'in here.' Such one-to-one mapping, again, is just one possibility and should not, in principle, be assumed to be the most effective one as far as survival is concerned.

Indeed, many of the objects in the space 'out there' (that is, objective reality) may be irrelevant to survival to the extent that they cannot influence the physical body whose survival is being optimized for. For instance, my own work in the field of artificial neural networks has shown that nervous systems can evolve to advantageously discard the representation of objects whose corresponding images would just increase the amount of 'noise' in the nervous system. Other objects may indeed be relevant to survival in different ways, but mostly according to their relative differences, so that a mapping that altered and distorted their true attributes (like location, behavior, appearance, autonomy, intensity, etc.) so to highlight these relative differences could conceivably favor survival. Again, in another one of my earlier scientific works, it has been very clearly shown that certain artificial nervous systems perform much better when failing to fully or accurately represent the data available to them. Beyond my own work, a wealth of data on pre-processing systems for artificial neural networks shows that one-to-one mappings between objects and subjective images are often not optimal. Artificial nervous systems using these advantageous pre-processing schemes would, thus, ‘see’ a world very, very different from what is actually 'out there.' Their perception of reality would hardly resemble reality, but instead be set up, through evolution, to 'transform' reality and optimize their own chances of survival. In essence, they would live in a hallucinated theater.

You see, evolution would, most certainly, favor mappings between objects (that is, reality) and subjective images (that is, perceptions) that favored survival, whether such mappings would accurately or completely represent reality or not. After all, the variable being optimized for here is not representation accuracy or completeness, but survival.

And now here we are: highly evolved organisms with the unique ability to create scientific models of reality. And yet, we naively make an assumption that our own models seem to render highly suspicious: we assume that what we see, or otherwise perceive, can be accurately mapped one-to-one onto the ‘real reality out there.’ We assume that the subjective images in our minds correspond perfectly to the objects of reality. We assume, thus, that we have complete and undistorted access to that reality. This is a contradiction: there is no reason to believe that our brains would have evolved to represent reality completely and as-is; they would, instead, have evolved to represent it in whatever incomplete or distorted way favored survival the most. Therefore, as evolved creatures, we simply have no way to tell what is really going on. And although our technological instruments do broaden our perception mechanisms beyond what nature has provided us with, they are ultimately also limited to what we can perceive as far as our ability to build them, and perceive their outputs.

So we end up with a profound contradiction: if we are to be consistent with the scientific paradigm, we cannot trust that what we see is what is actually going on; we may, for all we know, be living in an elaborate, brain-constructed hallucination of reality that happens to maximize our chances of survival. However, the very scientific paradigm that tells us this was itself built upon the very assumption that what we perceive corresponds accurately to nature. If that assumption cannot be made, then can we trust the conclusions of our scientific paradigm to begin with?


  1. This is a trite and unoriginal observation.

    Darwin himself remarked, "The horrid doubt always arises whether the convictions of man's mind, which has developed from the mind of the lower animals, are of any value or at all trustworthy. Would anyone trust the conviction of a monkey's mind, if there are any convictions in such a mind?"

    And, of course, we know from extensive research that our perceptions are often wrong, and wrong in predictable ways.

    But none of this provides the "profound contradiction" you think you have discovered. Yes, there are problems in perception and cognition, but the map between the real world and our mental model can't be wildly wrong. If it were, organisms with a better model would have outcompeted us. And the scientific method itself provides self-checking through replicability and peer review.

  2. Prof. Shallit,
    Thanks for the comment and interest in the article.
    I disagree with your statement that "the map between the real world and our mental model can't be wildly wrong." Yet it’s hard to argue against it since you provided no substantiation for such statement; as it stands, it is merely arbitrary. Saying that “if it were [the case], organisms with a better model would have outcompeted us” merely begs the question, since the point in dispute is precisely whether accuracy or completeness of representation would lead to a survival advantage. Therefore, I ask you: On what logical basis can your position be substantiated? I dare to say that there isn’t one such basis and I’d challenge you to articulate one. The moment we open the door to the possibility that our representations are incomplete and/or inaccurate (as you openly and clearly do), any substantiation for such a statement from within the framework of our representations becomes logically void. We just do not know what we are dealing with as far as reality is concerned, and must look upon even our most innate intuitions with suspicion. So even if subjective intuition alone were sufficient to motivate the kind of statement you make (and it is not), that could still be easily invalidated on a straight-forward logical basis.
    Finally, the replicability and peer-review processes you allude to are entirely irrelevant in the context of the argument, since both take place within the postulated framework of a distorted/incomplete representation space shared by all humans.
    Regards, Bernardo Kastrup.

  3. You're right, Bernardo, we might be a hallucination in some guy's dream. That's unfalsifiable.

    So how do we make progress?

    By bootstrapping. We work hard to establish some knowledge that seems dependable (by having others repeat it). Then, we can use that knowledge to try to gain some more. It's not perfect; nothing in science is provable, in the technical sense. But it's clearly made a difference to how we live.

    I agree that we don't have a one-to-one mapping of real objects in our minds. In fact, we don't have a representation for MOST things. It's only a small fraction of objects that we can really grasp. We spend most of our time talking about THOSE things (like we are right now).

  4. That said, I find the neurological mechanisms by which our cognition fails to be utterly fascinating.

  5. Prof. Orchard,
    Thanks for the comments.
    My own tendency is to think of science as an enabler of technology development. This way, whether reality is a ‘hallucination’ or not, so long as the ‘hallucination’ is stable and shared by all people we can still predict how its ‘hallucinated’ objects will behave and thus manipulate them accordingly. As such, progress can be made, and is constantly being made, insofar as our technology progresses.
    However, I do not believe that science can provide us a solid basis for the construction of ontology, and I think this point can be cogently argued in several different ways. That does not, of course, mean that science is unimportant, only that we seem to have extrapolated science beyond its reasonable scope without thinking about what we were doing. It is this somewhat illogical extrapolation of science onto its ontological facet I find counter-productive for our understanding of reality. We’ve grown too accustomed, in my view, to mixing models with truths.

  6. You are assuming that perception is indirect; that we form images and perceive those images. This is not settled science, and some believe that J.J. Gibson's direct perception is a better account. I include myself among the direct perceptionists.

    You claim that we optimize our behavior. I see no evidence or need for that. Suboptimal is plenty good enough, as long as it is sufficient for survivability.

  7. I acknowledge the direct perception hypothesis. Rupert Sheldrake also suggests the same. Still, it's clear that indirect perception is by far the mainstream, paradigmatic truth in science; and the thought exercise was to start from the mainstream scientific tenets.
    I do claim that another accepted idea in science is that evolution leads to an optimization of behavior. 'Optimization' here is a process, not a result. In other words, it does not entail that behavior is 'optimal' at any specific point in time, but that there continuous evolutionary pressure (i.e. a 'tendency') in that direction. That is sufficient for my argument to hold.

  8. On what logical basis can your position be substantiated?

    A silly question. We are not discussing any specific logic, where rules of inference are clear.

    whether accuracy or completeness of representation would lead to a survival advantage.

    Who said anything about "completeness"? But it should be clear to anyone that it is much easier for a model to be wildly wrong than wildly right, and most of those wildly wrong models will quickly lead to extinction. To give a trivial example: if I am unable to recognize predators, and react quickly in a way to defend myself or evade them, I will not survive to pass on my genes. Are you really denying this?

    And, of course, you didn't even address my point that your argument is trite and unoriginal - having been made by Darwin and Plantinga, just to name two.

    1. The neuroscientist, Donald Hoffman has a lot to say about the idea that what we experience is far removed from reality. He likens it to the icons on a PC screen. You move a file to the waste bin rather than figuring out how to reconfigure the bits on the hard disk to achieve the same transformation.

      I'd imagine Darwinian evolution might well have simplified the recognition of a predator - perhaps an antelope simply experiences profound fear if a predator is close.

  9. > A silly question <

    Let me try to re-state my question in an easier way. Your original statement was that our model of reality cannot be 'wildly wrong.' In the context of my article, which is what you refer to, 'wrong' implies inaccuracy and/or incompleteness (more on completeness below). Therefore, in your view, our internal representations must, at least to a fair extent, accurately and completely match what is really out there in terms of appearance, behavior, etc. My original point was that such is not necessarily the case, since representations would be skewed towards survival, not accuracy or completeness. Objects irrelevant to survival would not take up representation space (even if they were the vast majority of reality), and objects relevant to survival would be represented in whatever skewed manner were most conducive to effective behaviors.

    Therefore, my question to you was: How can you substantiate your statement that our representations cannot be wildly inaccurate or incomplete? It's a perfectly well-formed and legitimate question. I can accept your statement as the pronouncement of an opinion or personal intuition, but I was (and still am) hoping that you had more thoughts behind it.

    > Who said anything about "completeness"? <

    It is in my article, which I assume you read fully before commenting. Again: Logically speaking, a representation space is only not “wildly wrong” if (a) it somewhat accurately represents the objects it represents, and (b) if it represents most or all objects of reality. The latter is the reason to talk about completeness. Personally, I am very interested in whether the subjective reality constellated by my perceptions is the whole story, of just a mere subset of what is really ‘out there.’

    > But it should be clear to anyone that it is much easier for a model to be wildly wrong than wildly right ... if I am unable to recognize predators ... I will not survive to pass on my genes. Are you really denying this? <

    Of course not; on the contrary. Your question appears to reflect a basic misunderstanding of my argument.

    If you define “wrong” as a representation that leads to a survival disadvantage, and “right” as a representation that leads to a survival advantage, then I’d entirely agree with you that our representations must be “right;” but that would be an utterly trivial point: of course they are “right” in that sense. Indeed, that is at the core of my argument above.

    In the context of my article, which is what you refer to, “wrong” means an inaccurate and/or incomplete representation. To stay with your analogy, the point is whether the predators one perceives really exist in the way one perceives them (i.e. as existing continuously in time, as discrete space-bound entities, solid, autonomous, obeying our known laws of physics, displaying a certain appearance and behavior, etc.), or whether what we call “predators” aren't actually distorted projections of what is really going on, so that we stand a higher chance to survive whatever is really going on.

    > And, of course, you didn't even address my point that your argument is trite and unoriginal - having been made by Darwin and Plantinga <

    Well, I don't have a problem with the fact that you find the article trite; it seems to me to somewhat contradict your clear interest in debating it, but it's totally OK as far as I am concerned. Not sure what else I can say here.

    Regarding originality, I didn't claim or disclaim it. Frankly, I am not too preoccupied with this kind of ego trip so long as my particular articulation of the idea adds some value; and provided, of course, that I don't infringe on anybody's rights. That said, if, as you state, other great thinkers did hold similar views to my own (although articulated in very different ways), then that gives me extra confidence that I am on a legitimate path of thought. Thanks for pointing that out!

  10. I think his essay has it right on. In fact our perceptions are most definitely wildly wrong -- we cannot unaided perceive the incredibly rich and vast particulate world around us, and even when we use machinery to count pollen, for example, we get only a fractional picture of the whole; we do not see magnetic fields or many detectible wavelengths of radiation, and these are only the forces of which we are aware; the may be many others which escape our detection because we are not evolved to perceive them, though they be around us all the time. We cannot see 'time' itself. Nor can we well perceive things occurring on highly macroscopic scales of space and time around us. We have evolved a locally useful tunnel vision permitting us to see the narrow band of things useful to us against our potential predators, and that is about it!!

  11. The key is the relationship between truth as correspondence and biological survival: our models allow us to survive means that they crews is an objective reality? Before I believed that our models could not be systematically wrong about reality, because if so we would not have survived, but now I think the connection between truth and survival is weaker than believed, because not only our models do not need to represent the reality as it is for our survival, but an accurate representation of reality would be a stumbling block to our survival, because survival models need to be practical and simple, not so complex and sophisticated to faithfully represent reality.

  12. One can imagine intelligent aliens evolved in a way that their models of reality, while fully enabling their survival, are so different than ours as to make communicating with them near impossible. Of course there a number of other assumptions built into that statement.

    Re some of the comments posted here, you can always tell a reactionary dogmatist when their comments open with insults, and then proceed to show that they read what you wrote through such heavy filtering as to essentially not have read it at all. Of course he might just take pleasure in being nasty. In which case this is just one more little piece of the public record he is building about himself across the net which will be there for his associates, children, students, posterity etc.. to remember about him forever.

    1. Yeah, it is entertaining to see the furious self-appointed defenders of 'reason' displaying their insecurity. :-) Rational Wiki, for instance, is pricelessly funny. No, really, it is. :-)

  13. In response to the earlier comments, I would argue that the common postulate that evolution would select for brains that cannot fathom reality due to selecting away from REASONING ABILITY to quite different from the thrust of this article. That is, if humans are a group of "dim wits", we can still test our models and OBJECTIVELY see if we are on the right track. On the other hand, if we are selected to see a distorted reality, then we are missing data to work with. We likely wouldn't be able to predict anything at all. It's just incongruent with the picture of reality we are currently living with, where we can fathom theories of relativity and test them.