A number of spiritual and philosophical traditions hold that reality is very different from what it seems to be. Buddhism springs to mind, as does George Berkeley’s idealism. But I don’t think I’ve ever heard an argument in this vein that’s as distinctively disorienting as the one made by cognitive scientist Donald Hoffman in his recent book The Case Against Reality: Why Evolution Hid the Truth from Our Eyes. The Apostle Paul said that we see this world “through a glass, darkly,” but Hoffman would say that’s wildly optimistic. As he put it to me in a conversation on The Wright Show podcast (available as video on meaningoflife.tv): We have to let go of the idea “that there's any resemblance whatsoever between the nature of our perceptions—and even the language of our perceptions—and the nature of objective reality.” Below is Part I of that conversation (which we had several years before his book came out). Part II will appear in next week’s newsletter.
ROBERT WRIGHT: What we're going to talk about today is kind of at the intersection of cognitive science and philosophy. We're going to talk about the mind-body problem, the question of what consciousness is, and a question that's raised by your particular theory of consciousness, which is, so far as I know, quite distinctive—unlike anything I've heard before. That question is whether what we think we see is really real, or how close to real it is.
Your theory of consciousness, which has been getting attention among the people who think about these things, suggests that things are not as real as we think they are.
This bottle of water—
—it’s useful for me to think I see it, but it may not bear a very close correspondence to the underlying reality, right?
DONALD HOFFMAN: Correct. It's real as an experience, but it may not exist apart from my experience in that form.
Your theory builds on the following fact about natural selection… that, strictly speaking, natural selection doesn't build a brain that sees the truth. I mean, that's not what the criterion of natural selection is. The criterion is: natural selection will preserve traits that are conducive to the proliferation of genes.
And so it will build brains that have the kinds of perceptions and thoughts that are conducive to the proliferation of genes. And if those perceptions and thoughts are false but still are conducive to the proliferation of genes, then there will be false perceptions. I think that's actually uncontroversial in evolutionary biology.
And some mundane reflections of that are well-known.
For example, you could imagine a species that grows up amid poisonous snakes. When you're walking through the brush and there's anything that makes the kind of noise a snake makes, it probably makes sense to get alarmed and even think you see a snake. Even if 9 times out of 10, there is no snake, that still is on balance a healthy policy. So that will lead to the actual illusion of snakes.
You'll sometimes think you see a snake when it's just a lizard or something else. But natural selection favors that kind of perceptual bias because it helps you stay alive long enough to get your genes in the next generation.
Now, that's a mundane reflection of this. And so too are a lot of famous optical illusions. Those are also reflections of this in ways I don't think we need to get into.
But you're taking this fact about natural selection and making a much deeper claim about its implications for perception, right? …
So, the standard view in the field is the one that you just described: that, of course, natural selection is, in the first instance, only about propagating the genes, propagating the species; and perceptions that do that are the ones that will survive. And I think nobody argues with that.
But the assumption in the field has been that the perceptual strategies that will actually be favored by that kind of natural selection are perceptual strategies that see reality as it is. Not exhaustively—very, very few people would claim that we see all of reality as it is—but that those aspects of the world that we do see, we do see accurately; and we see the ones that we need to survive and reproduce.
And so the assumption has been therefore that veridical perceptions—perceptions that are accurate to the state of the world … — are the ones that are favored by natural selection.
That's the textbook assumption, it's the assumption by most researchers in the field, but it's an assumption that we don't have to just wave our hands about. We can actually go and check it, because evolution is a mathematically precise theory. We have the tools of evolutionary game theory, evolutionary graph theory and even genetic algorithms, where we can actually go in and ask: What does the mathematical theory of evolution do … to veridical perceptions? Does it favor them? Does it make them proliferate, or does it drive them to extinction?
Can I just get clear on terminology? In the example I gave, if it's a veridical perception in your terms, that would mean the snake is actually there. "Veridical" means "accurately representing reality", right?
Exactly. And most people in the field of perceptual science would say that there is a real three-dimensional world with real physical objects like snakes in it. And when you see a snake—when you have the perceptual experience that you describe as a snake—that perceptual experience is a veridical representation of a true snake in objective reality that would exist whether or not you perceive it.
As a rule. Now, it is conventionally acknowledged that there are the exceptions, like where you think "Oh, that's a snake" when there's not, because you err on the side of caution, and a few other examples like that.
Maybe this is a good distinction to make. People agree that natural selection will create brains that give rise to the illusion of a snake in certain cases—but that often, when you see a snake, the snake is really there. And what you're saying is no, the snake is not necessarily ever there, in a sense.
That's right. … The standard assumption is that … there are rare circumstances where our perceptions will fail to be veridical, but … that, under normal circumstances, we will see those aspects of reality accurately that we need to see accurately.
And that was the assumption that I decided to test. Initially I was skeptical about it because I thought, well, reality is pretty complicated, and there are all sorts of selection pressures to do things fast and cheap. And so maybe perceptual systems may not go after the veridical perceptions of the world simply because they're very expensive. …
It turned out that that's true. We try to do things faster and cheaper because every calorie you spend on perception is a calorie that you have to get somewhere, you have to kill something and eat it to get that calorie. So, other things being equal, we try to do things on the cheap and do them quickly. And that's certainly turned out to be true in the Monte Carlo simulations of evolutionary game theory that we ran.
But the deeper result that, in retrospect, I should have seen upfront (but I didn’t) was that fitness functions—which are the bread and butter of evolution, they're how you get your fitness points from your actions—are a function of the real state of the world, whatever that world might be, but they're not identical to the world. They're functions on the world. …
Let me make sure I understand. You can do computer simulations of evolution. And you did computer simulations, which showed what we just said: that when there's a difference between seeing things truly and seeing things that will get your genes into the next generation whether or not they're really there, the perceptions that will evolve are those that get your genes into the next generation. Right?
We needn't spend much time on this, but I have a question about that. It's almost like: Why do you bother doing the simulation, since it's true by definition that, if natural selection has to choose between a true perception and a perception that gets genes in the next generation, the latter will be the perception that transpires? It seems to me you could have just asserted that as an axiom that's true by definition, and then gotten on with your philosophy—which we'll get on with shortly—but am I missing something as to why you actually needed to do the simulations?
Pretty much the reason I did the simulations is because everybody else in the field of perceptual science actually stated in print that veridical perceptions were the ones that were more fit.
So they understood that it's all about fitness, but then they conflated fitness with veridical perceptions. They said that the reasons our perceptions are useful evolutionarily is because they're true.
It was that identification of the evolutionary usefulness of our perceptions with them being true that I wanted to test. … And there is just the fact that, although fitness does depend on the state of the world, it's not the same thing as the state of the world. The two diverge.
So it sounds like you needed to convince some people who perhaps were not sufficiently steeped in the logic of natural selection, and so you did a computer simulation.
In any event, let's definitely stipulate the premise that non-veridical perceptions will prevail in natural selection, if that's what gets genes into the next generation. And let's do what you do with that, which is to take it to a much deeper level than I had seen it taken.
Maybe you could explain to us your desktop computer metaphor for what the relationship is between what we think is reality and actual reality.
Okay, very good.
So the first thing I should say is that I'm not a so-called metaphysical solipsist, which is a person who claims that there's nothing that exists except for me and my perceptions. ... I think that there is some kind of objective reality, and it'll exist whether or not I exist—and I have life insurance, so I'm putting my money where my mouth is—but the question is: if evolution has not shaped our perceptions to resemble that objective reality, it's just given us a set of symbols that are species-specific guides to keep us alive, then how should we understand the nature of those perceptions?
The metaphor that I think is very, very helpful is the metaphor of the desktop interface on your computer.
Suppose you have on your desktop an icon that's, say, blue and rectangular, and in the lower right-hand corner of the screen, and it's for an email message you're creating, or a paper you're writing.
You can ask yourself: if that icon is blue and rectangular, does that mean that the email message itself, or the file that you're writing, is blue and rectangular?
And the answer is, of course not. That's silly. Anybody who thought that the color of the icons on the desktop, or their shapes, or their position … was meant to reflect the true colors and true shapes and positions of the file on the computer, just misunderstands the whole point of the interface.
The interface is there not to show you the complex reality of the computer, like the diodes and resistors and the voltages and magnetic fields. It's not there to show you that—in fact, it's there to hide that. We pay good money for the interface software because we don't want to know that stuff. It's irrelevant to what we really want to do, like sending emails and editing photos and so forth.
Because it's, for practical purposes, a more efficient description of reality. To see an icon is a more useful perception than to see a bunch of electrons bumping into each other on a circuit board, or whatever the underlying reality is. And as we've said, natural selection will favor perceptions that are efficient in the sense of getting genes into the next generation.
Exactly. So the idea then is that natural selection has endowed us with a species-specific user interface.
Our space and time, the three dimensions of space as we perceive them are the desktop, and physical objects are the icons.
So it isn't just objects that are in some sense not real. You're saying space-time is itself just a convenient projection.
Now, you said "species-specific." ... I would imagine that certainly dogs are seeing three dimensions and are in some sense implicitly assuming that some things come after others. I mean, is it really “species-specific?” Or is it more like “specific to life?”
Well, there may be some aspects of the user interface that is common across some species, but I would imagine that E. coli, for example, doesn't have very much of a 3D representation of space. It certainly is plausible to me that it might not. If it did, I would be interested to find out that it did; but it's certainly at least reasonable to me to think that it doesn't.
So I don't want to take any aspect of the human perceptual system, the human species-specific interface, as necessarily true of any other species. I think we need to look case by case and see what evidence we might have that the interfaces of other species are like our own.
Even within homo sapiens, there are mutations, and it's not at all clear that your interface is exactly the same as mine. I would think that they're relevantly similar, and that's why we can communicate pretty well. But evolution is continuing, and so your interface is probably a little bit different from mine.
So even when I speak about a species-specific interface, I'm being a little bit loose there, because your interface is not exactly the same as mine.
So let me try to pursue a computer-like analogy a little. …
The question that will naturally occur to people is, well, what is the real world like then? You do refer to a world. And I'd be tempted to say, on the basis of your metaphor, that reality is as different from what we perceive as a bunch of electrons bumping into each other on a circuit board and moving in certain patterns are different from file folders on our computer.
But I actually think your theory is a little weirder than that. When I've heard you describe what you think the world is, it’s not as straightforward an extension of that analogy as I might've hoped. You know what I mean?
And I would assume that, at this point, a lot of people start having trouble wrapping their minds around things. … Like, if I asked you the question: okay, what is reality like? … If the analogy ... is … I'm a creature in a video game, like Pacman … running around the screen, seeming to see things and responding to obstacles by moving around them—but in fact, those obstacles are illusory, they're just pixels, and really the action is being governed by these electrons on the circuit board.
It's my temptation to think of it that way. But I actually think things in your view are weirder than that. Right?
That's right. So, the starting point is the recognition that the very language of our perceptions—of space and time and physical objects with matter and shapes and colors and so forth—that very language has been shaped by natural selection simply to keep us alive, not to tell us anything about the nature of objective reality as it is.
So there's no reason to believe that any of that language, any of the predicates of that language, and anything that could be described in that language is going to be the nature of the objective world.
And that's pretty strong. What that means is that, whatever objective reality is, the chances that thinking about it in terms of space and time and matter and physical objects and colors and shapes is just the wrong language. You'll get nowhere.
The analogy would be: suppose I said, I would like you to tell me a detailed theory of the inner workings of a computer, but the only language I'm going to let you use is the language of the pixels on your desktop. So you have positions and colors of pixels—that’s your language.
Go for it. Try as best you can to describe how a computer works. Good luck. I've made it impossible for you because I've given you a language that's not the right language to do the job.
And so that's the same problem that we have when we start trying to think of using the language of space and time and position, momentum, spin, shape, mass and matter to describe the objective reality. We've given ourselves an impossible task, just like asking ourselves to describe the computer just based on the pixels on the desktop.
So what we have to do is to be willing to let go altogether of some very deeply held beliefs, that there's any resemblance whatsoever between the nature of our perceptions—and even the language of our perceptions—and the nature of objective reality.
That's a very, very difficult position for us to take. We seem to be very, very wedded to the idea that our perceptions are giving us some kind of insight into the nature of reality; at least the language of our perceptions is the right language to describe reality. And that's what I'm saying we need to call into question now.
By the way, one of the things you're calling into a question ... is causality itself.
If you see billiard balls bumping into each other on a computer, and it looks like the ball is causing the other ball to move—well, you know that's not causality happening on your computer screen. Rather, they're both just behaving in accordance to the script. …
And this has been suggested by philosophers like Hume and so on, that we can't really confidently infer actual causality, we never really know more than that certain things regularly precede other certain things. … We never know for sure that causality per se happens.
Anyway—that’s another weird thing about your theory.
That's right. ... I'm happy to contemplate the idea that the notion of causality itself is not a fiction, but that the specific causality that we all know and love—namely that a physical object like a billiard ball hitting another billiard ball and making it move—that that's genuine causality, that I think we will have to give up.
So when the white ball hits the eight ball into the corner pocket, we can say that the cue ball caused the eight ball to move, and for practical purposes, that's fine. It's a useful fiction. But strictly speaking, it's a fiction.
There is perhaps some causal sequence going on in the objective world, and all I can do as a humble member of the species Homo Sapiens is to represent it in terms of a cue ball hitting an eight ball and having a fiction of the causal interaction between those two.
And by the way, in some scenarios where causality is an illusion, people are actually talking about a completely deterministic universe. So what they're really doing is just watching a movie, and everything has been prescribed from the beginning. ...
Yours is not necessarily, as I understand it, a completely determined universe. At least you use terms like "choice" in your theory, right? So you're not talking about a deterministic universe necessarily.
Not necessarily a deterministic universe. And I should point out that the theory that I'm working on has two separate aspects, and that one could buy the one and not the other.
So, the first aspect, that I'm fairly confident about, is that evolution entails that we almost surely do not see reality as it is. If you buy evolution, then I think that you have to buy the conclusion that space and time and physical objects are species-specific desktop and not a depiction of reality.
Now, once you bought that, we have to give up our dearly loved theory that reality resembles what we perceive. And that's hard to give up. But that, I think, just follows from evolution.
The next step is to ask, what theory of reality shall we propose? And there's an infinite number of possibilities, right? We've thrown away one theory, the one that we happen to like a lot, but it's just one of an infinite number of possibilities. So when I go and propose a specific new theory... You know, I'm probably wrong. …
The evolutionary conclusion is we don't see reality as it is. The second step is: okay, now, as scientific theorists, what shall we propose as a new theory of that reality? And someone can buy my first proposal—that we don't see reality as it is—and not buy like my proposal about the nature of reality that I'm going to propose…
Illustrations by Nikita Petrov.
This site features only a fraction of the writing I publish in my newsletter.
Please, consider subscribing.