Does Evolution 'Breed Reality Out of Us?' Addressing Donald Hoffman's Metaphysics
James A. Lindsay, @goddoesnt
"Metaphysics: Just Say No!" -Peter Boghossian
For those who haven't heard of Donald Hoffman, this is about to get weird. If you have heard of this particular cognitive scientist from the University of California, Irvine, you already know it's going to get weird. If you aren't sure, but you're into what goes on here at Allthink, you may recognize the name from Dilbert creator Scott Adams, who is really taken with Hoffman's ideas, uses them to support his unpersuasive case that Trump will win massively in November, and recently talked to Dave Rubin on The Rubin Report. Whatever case you find yourself in, this is still going to get weird, maybe even a little weirder than if I were talking about Max Tegmark, another academic with a very strange take on metaphysics (that reality is really just math).
Why is it going to get weird? Because Hoffman has a weird take on metaphysics, and for some reason, as with Tegmark, people take it seriously enough to give it some press. The link to the piece about Tegmark in the previous paragraph, in fact, goes to my reaction to listening to Sam Harris's rather enthusiastic talk with him. Tegmark believes the fundamental nature of reality is math. Hoffman doesn't quite believe in reality, kind of, as he revealed in an interview with Amanda Gefter in The Atlantic, titled "The Case Against Reality" (note: the interview was originally for Quanta Magazine) and now he's had some press about his ideas in NPR, after talking to Adam Frank, titled "What If Evolution Bred Reality Out of Us?"
Evolutionary biologist Jerry Coyne took Hoffman's Quanta/Atlantic interview to task back in July, and he shared his criticisms on Twitter again today after finding out about Frank's NPR piece. You really should read what Coyne had to say just to get some idea of not only how wacky Hoffman's ideas about reality are but also how unclear they are. If Coyne can't make sense of what you're trying to say after considerable effort and the presentation of your ideas having the benefit of top-notch editors, there's a good chance it doesn't make much sense.
https://twitter.com/Evolutionistrue/status/773615215120379904I can't really do this justice without forcing you to suffer some of what Hoffman thinks, but Hoffman, like others before him, does a great job of vindicating Peter Boghossian's straightforward summary of almost everything metaphysics: just say no! It's difficult to know where to begin, however, in succinctly presenting a "case against reality."
Hoffman's idea isn't so crazy that it can't get off the ground, so I'll start with Gefter's summary of it in Quanta/The Atlantic:
The true reality might be forever beyond our reach, but surely our senses give us at least an inkling of what it's really like.
Not so, says Donald D. Hoffman, a professor of cognitive science at the University of California, Irvine. Hoffman has spent the past three decades studying perception, artificial intelligence, evolutionary game theory and the brain, and his conclusion is a dramatic one: The world presented to us by our perceptions is nothing like reality. What's more, he says, we have evolution itself to thank for this magnificent illusion, as it maximizes evolutionary fitness by driving truth to extinction.
Frank gives him a similar treatment for NPR:
For decades, Hoffman, a professor at the University of California, Irvine, has been studying the links between evolution, perception and intelligence (both natural and machine). Based on that body of work, he thinks we've been missing something fundamental when it comes to fundamental reality.
Fundamentally, Hoffman argues, evolution and reality (the objective kind) have almost nothing to do with each other.
Immediately in both of these summaries, we hit a roadblock. Actually, not a roadblock; the roadblock. We've already run smack into the roadblock that always kills us with metaphysics. How on Earth could Hoffman claim to know that?
The answer is that he cannot. Hoffman cannot know that "evolution and reality have almost nothing to do with each other," that "we've been missing something fundamental when it comes to fundamental reality," or that "the world presented to us by our perceptions is nothing like reality." In fact, if Hoffman's interpretation of reality, which he calls "conscious realism" (paging Dr. Chopra...), is right, he can't know those things because he has absolutely no mechanism for telling him so. What measurement, which is to say perception, could tell Hoffman that his perceptions are "nothing like reality," and how could he know that it's "like reality" to say his perceptions are not like reality?
This is always the problem with metaphysical speculation: metaphysics begins where epistemology ends. In English, metaphysical speculations about "fundamental reality" pick up right where we have to stop saying "I can claim to know whether or not this is true."
Don't take my word for it. My own philosophical descriptor is what nerds like myself refer to as "model-dependent realism," a term imported from two pretty sharp fellows, Stephen Hawking and Leonard Mlodinow, who formulated this savvy update to the position formerly known as "instrumentalism" in their 2012 book The Grand Design. I had the pleasure of discussing model-dependent realism over wine and snacks with Mlodinow back in March, and I got his seal of approval that I have a pretty good grasp of it, so I trust what I'm about to say isn't armchair poppycock.
Model-dependent realism posits that there is a real objective reality out there, but the limits of our knowledge about it are defined by what our data tell us, as interpreted through the models we have made to explain that data. The general take is that we know our models are guesses, we can grade them by their explanatory and predictive power, and so we might as well call the objects of our models provisionally "real" and get on with it, even though the nature of "fundamental reality" lies beyond the horizons of what we can know. (I know, too much philosophy of science, and I'm sorry!) The short take: we can't know if protons are really protons, but we can say something proton-like is happening, the model containing protons is decently resonant with how we think, and the model containing protons plays really well with data, so we might as well call them protons, say they're real, act as though the model describes them accurately, and get on with it until we have something better.
Crucially, Hoffman is making claims that lie beyond the knowledge horizon, and so he's engaging in metaphysical speculation. It's popular stuff too, especially with today's slightly New Age, slightly postmodern quasi-intelligensia. Hoffman's popularity and press took off after they saw his sermon in their church, a viral (read with a double meaning) March 2015 TED Talk about his interesting, crazy, and ultimately overreaching metaphysical ideas. (TED's tagline is "ideas worth spreading," not "correct ideas," by the way, and much of what's been said in those prestigious red circles needs to be taken with a healthy grain of salt.)
Hoffman's radical idea at the core of his bizarre metaphysics is probably a good one, or at least a slightly interesting phrasing of one. He postulates it as a theorem:
Given an arbitrary world and arbitrary fitness functions, an organism that sees reality as it is will never be more fit than an organism of equal complexity that sees none of reality but that is just tuned to fitness.
Theorem, of course, is a word that is usually associated for robust logical proof, especially in mathematics but also in philosophy, and it's a pretty poorly understood concept. A theorem is an abstract statement within what is known as an "axiomatic system" possessing a truth value of TRUE. What constitutes an axiomatic system is immaterial except that it is a system of ideas, of abstractions -- exactly the kind of place where logic can let metaphysics run away from reality. It's critical to understand, however, that our theorems operate only in mental space and have no bearing upon reality. Bad axioms (assumptions) lead to "true" theorems that have nothing to do with reality.
That said, Hoffman's "theorem" is probably pretty sound, but only in the despite of its hidden assumptions. Once we account for the hidden assumptions within it's also banal. The magic is all in the hidden assumptions, and without exposing them, it looks profound. It looks like Hoffman is saying "Organisms with access to a more useful illusion always outperform those that perceive reality accurately."
We'll come back to the theorem and its assumptions because we have to stare into the abyss of what Hoffman concludes from it first. Hoffman's pseudo-profundity echoes the similar ramblings of a theologian, who uses his to prove there "must" be spiritual stuff. Hoffman's conclusions are little more than Notre Dame theologian Alvin Plantinga's Evolutionary Argument Against Naturalism in a lab coat and safety goggles. The EAAN isn't right; in fact, it's ironically wrong in that it attempts to justify a relatively useful superstition (Christianity) by arguing that evolution would favor useful superstitions over getting things right. Hoffman, or at least Gefter, isn't even abashed about the fact that he's retooling that religious apologetic nonsense; his original interview in Quanta Magazine is titled "The Evolutionary Argument Against Reality." I'd give almost anything to be kidding right now.
How does it work? Hoffman explains, courtesy of Frank at NPR,
... an organism tuned to fitness might see small and large quantities of some resource as, say, red, to indicate low fitness, whereas they might see intermediate quantities as green, to indicate high fitness. Its perceptions will be tuned to fitness, but not to truth. It won't see any distinction between small and large - it only sees red - even though such a distinction exists in reality.
From which he goes on to remark,
We assume the 'predicates' of perceptions - space, time, physical objects, shapes - are the right ones to describe physical reality. And this theorem says that [such] predicates are [the wrong ones] almost surely.
Maybe what he's saying isn't so crazy. It doesn't seem completely nuts to say the two things he seems to be getting at. Hoffman appears to be suggesting (1) that if our perceptions are attuned to fitness rather than accuracy, we have a better chance of achieving goals (fitness) and that evolution would select for such successes (duh -- that which is attuned to succeeding will succeed more often than that which isn't), and (2) we don't have raw access to reality but have to interface with it via a neurological perceptual system (duh -- brains filter our sensory perceptions as they become part of conscious awareness) that evolved for fitness.
We'll come back to (2) in a moment, but look at (1) for a minute. We're on the cusp of augmented reality, and with the help of some rather clunky instruments, we already have it. We have cameras that can see infrared and let us see footprints left on floors after people have passed, for a few minutes anyway. We have apps that name the buildings or streets we're looking at through our cellphones and that can tell us how to get places or link us to more information on the Web. We have other apps that identify plants for us, among about a billion other uses.
So, it isn't a stretch to imagine that we could create augmented reality apps that identify how healthy or dangerous wild plants are, that make venomous snakes glow red, and so on, through our interfaces. All of those tools would increase fitness and, to the degree they work, provide a good result. This, of course, is the whole burrito laying at the root of Michael Shermer's argument that we've evolved hyperactive agency detection and are, on the balance, a bit jumpy, just in case the wind in the savannah grasses is really a leopard. Wrong guesses in such instances keep us safer but waste resources, and getting things more right would optimize the difficult situation (which is part of why we evolved highly discerning brains, as we have). Augmented reality would help us identify threats and opportunities more accurately -- a net win.
Hoffman's argument in (1), then, is little more than that our sensory systems that we are already equipped with -- eyes, ears, nose, and so on -- would benefit us more by giving us clues about reality as it applies to our fitness (as Coyne points out: they do in some cases, like that fats and sugars taste good, and I'd add that sewage and rotting flesh smell horrible). I can't see that Hoffman is saying anything particularly new or interesting here until he starts asserting that our perceptions have nothing to do with reality. That's only new in that it's almost certainly dead wrong and thus few, if any, before him would be remembered for saying it.
The real problem arises in Hoffman's conclusions. Hoffman concludes things that are plainly outlandish, like that there is no external reality, brains aren't real, and so on. No, really. He said that, here from the Quanta/Atlantic interview with Amanda Gefter.
The idea that what we're doing is measuring publicly accessible objects, the idea that objectivity results from the fact that you and I can measure the same object in the exact same situation and get the same results - it's very clear from quantum mechanics that that idea has to go. Physics tells us that there are no public physical objects. So what's going on? Here's how I think about it. I can talk to you about my headache and believe that I am communicating effectively with you, because you've had your own headaches. The same thing is true as apples and the moon and the sun and the universe. Just like you have your own headache, you have your own moon. But I assume it's relevantly similar to mine. That's an assumption that could be false, but that's the source of my communication, and that's the best we can do in terms of public physical objects and objective science.
. . . Neurons, brains, space … these are just symbols we use, they're not real. It's not that there's a classical brain that does some quantum magic. It's that there's no brain! Quantum mechanics says that classical objects - including brains - don't exist.
Well, kind of. In Frank's words, for NPR, maybe what Hoffman clearly seems to be saying is not quite what Hoffman meant, but it's hard to say since what's just below is Frank's summary of Hoffman's views (which, as they are far saner, are easy to attribute honestly to Hoffman as well).
While there clearly is a world separate from us, Hoffman says, evolution does not give us access to that. Instead, he claims, it's our interactions as conscious agents that give shape to the reality we experience. (bold added)
Okay, now we're getting somewhere. Now we can see what's going on with (2) above, that we don't have raw access to reality but have to interface with it via a perceptual system. Hoffman has rediscovered the subjective/objective dichotomy in new language and gotten lost over the fact that the subjective experience platform would have evolved to benefit our organisms' fitness (or, at least the fitness of our ancestors). Of course, if you know what models are for in science -- interfacing between data and the universe of ideas and experiences that lives inside our head, which is all we have access to -- model-dependent realism already does this for us, as do other attempts to deal with our ultimate epistemic gaps, all of which are quite a bit less crazy in that they don't throw out reality with the bathwater. Pardon me while I ...yaaaaawn.
So Hoffman's claims that we actually interact with a set of inputs that have been filtered by consciousness (the filter itself having been shaped by evolution to improve fitness) isn't crazy; it's boring. It's the kind of thing that would have been Earth-shattering in the 1880s, at least if he had done it more carefully than today's injection of postmodernism seems to have allowed him. His observation that conscious experience is all each of us really have is similarly not-newsworthy. What makes anything Hoffman is saying here interesting, then? His woo-laden assertion that reality itself depends upon consciousness, instead of the other way around. Maybe you're scratching your head thinking, "nah, he can't really think that, though...." He does, and so does Scott Adams (the link on "Adams" being more interesting than the one on "Scott" here). Hoffman writes,
As a conscious realist, I am postulating conscious experiences as ontological primitives, the most basic ingredients of the world. I'm claiming that experiences are the real coin of the realm. The experiences of everyday life - my real feeling of a headache, my real taste of chocolate - that really is the ultimate nature of reality.
Pete... he said, "that really is the ultimate nature of reality." What do you think? Yeah, me too... METAPHYSICS: JUST SAY NO!
So, Hoffman has confused himself desperately on two simple facts. First, our experience of the objective world is inherently subjective, and, second, perceptions that are a boost to fitness allow us to perform better. (We could argue that he's also perplexed by the interaction of these facts, that the nature of our conscious experience evolved with us.) Coyne was right: nothing new or interesting is here; that is, unless you count Hoffman's peculiar metaphysics and intellectual arrogance, a persistent requirement of believing one has determined the ultimate nature of reality.
This is far too long already, so let's wrap it up with the last nail. I promised we'd get back to the hidden assumptions that let us see through Hoffman's theorem, though, and now we shall. Let's do it by restating the theorem and adding in the hidden assumptions. I'll put Hoffman's original in the usual font and bold the additions.
Given an arbitrary world and arbitrary* fitness functions relevant to survival and/or reproduction in that world, an organism that sees reality as it is will never be more fit than an organism of equal complexity that sees none of reality but that is just tuned to fitness given the constraints upon fitness imposed by that world.
That's where Hoffman's theorem breaks, and it probably happened on the words "arbitrary world." In fact, to correct this theorem accurately, the first word of "arbitrary fitness functions" would have to be removed. There are no "arbitrary fitness functions" in any world. All fitness functions are contingent upon fitness in that world according to achieving certain goals within that world. And that is where Hoffman's theorem impacts reality, completely undermining his ability to claim that reality is irrelevant. All fitness functions in any world will be dependent upon the realities of that world. Always. Because that's what the notion of a "fitness function" entails. That means the "fitness functions" at the heart of our own evolution in our own reality are shaped directly by reality and thus take it into account, as seems as plain as the noses on our faces, which not only we can see, but everyone else can too.
So no, evolution does not breed reality out of us, and, no, Hoffman's EAAR (Evolutionary Argument Against Reality) metaphysics doesn't argue effectively against reality any more than Alvin Plantinga's EAAN argues against a naturalistic universe. Coyne is absolutely right here: Hoffman gets it exactly backwards, and there's nothing new or interesting here.