In working slowly through a gigantic backlog of blog comments, I met with one that goes back to an old school question, about my project to demarcate the natural and the supernatural. The comment by Enlightenment Liberal is here. He is asking questions about the conclusion I argued here and in print here (with a followup here). The first, Defining the Supernatural, supports the others, Defining Naturalism I and II. His perspective can be summarized as “If we grant your definitions of ‘natural’ and ‘supernatural’, I think that all hypotheses of the form ‘X is supernatural’ entail absolutely zero observable predictions about the world,” in particular because “I think that I have absolutely no basis to conclude that there is any relation or correlation at all between the fundamental nature of things and the observable nature of things,” in accordance with Logical Positivism.

So, is he right? Let’s explore…

To understand his remark, note the very elegant refutation of strict Logical Positivism by Hilary Putnam, which is smartly summarized on Wikipedia. One point there relevant to here:

Putman also alleged that positivism was actually a form of metaphysical idealism by its rejecting [any] scientific theory’s ability to garner knowledge about nature’s unobservable aspects. With his “no miracles” argument, posed in 1974, Putnam asserted scientific realism, the stance that science achieves true—or approximately true—knowledge of the world as it exists independently of humans’ sensory experience. In this, Putnam opposed not only [logical] positivism but [any] other instrumentalism—whereby [a] scientific theory [is] but a human tool to predict human observations—filling the void left by positivism’s decline.

Though I have been very influenced by logical positivism and retooled some of its most useful aspects for use within my own philosophy, I agree with Putnam that it is not coherent. I am now (and was already upon writing Sense and Goodness without God) a scientific realist. I had been influenced as much by Michael Polanyi (Personal Knowledge) as by A.J. Ayer (Language, Truth and Logic), and my semantics and epistemology is a synthesis of them (I still consider both their books just named as essential reading for any philosopher today). And my metaphysics follows from that, in conjunction with the facts of science.

In result, I do believe there is a meaningful, and testable (and thus verifiable and thus falsifiable) distinction between a natural world and a supernatural one. Tautologically a natural world is a world with nothing supernatural in it, and a supernatural world is a world with at least one supernatural thing in it. So we aren’t just talking about whole worlds, but the properties of individual things: there can be a supernatural person, object, force, etc. These are all coherent and meaningfully different from those same persons, objects, forces being entirely natural. Epistemic access to the difference might not always be available. But it can be available enough. And more importantly, the differences are meaningful, in that, we can understand what would be different about the world if one were true and not the other. Which entails we can in theory observe those differences.

A natural object (or person or force etc.) is an object that is entirely causally reducible to a nonmental thing (or system of nonmental things). So, for example, if all human conscious thought and experience can be entirely causally reduced to just the neurological biochemistry of the brain, then the human mind is a natural thing. A supernatural object (or person or force etc.) is an object that cannot be causally reduced to nonmental things. At some fundamental level of its existence, it is dependent on at least one irreducibly mental property. Souls, God, magic potions can all be supernatural in this sense. Now, the supernatural as thus defined might be logically impossible, in which case naturalism is necessarily true (and God, as traditionally conceived, necessarily does not exist). But no one has adduced a deductive proof of that, so we can’t yet say that it’s so. (For a full exploration of the possibility, though, and why it might be more than just a possibility, see The God Impossible.)

To understand why this makes sense—and doesn’t have the consequence that ‘X is supernatural’ entails no observable predictions about the world—just follow along with us as I briefly fisk EnlightenmentLiberal’s case:

-:-

Perhaps the disagreement is here: I think that I have absolutely no basis to conclude that there is any relation or correlation at all between the fundamental nature of things and the observable nature of things.

So you think “Leprechauns cause all the observations of the Periodic Table” is just as probable as atoms causing it? Somehow I doubt that. You therefore cannot really believe there is no correlation between observed evidence and inferred causes of those observations.

So my first question is, are you confusing probability with absolute certainty? Because the Leprechaun theory of the Periodic Table is possible and could well be true. But it requires adding so many assumptions not in evidence that its prior probability drops to a vanishingly small value (see Proving History, pp. 80-81, 104-06). Whereas that does not happen for the “atomic” theory of the Periodic Table. Therefore it is highly probable atomic theory is the cause and not leprechauns.

And this holds all the way down—like trying to explain why atoms exist by appealing to leprechauns rather than the Standard Model, or trying to explain the Standard Model with leprechauns rather than, for example, Superstring Theory, and so on. This is why a fatal problem with logical positivism is that it cannot answer any Cartesian Demon (or CD) argument. It is therefore logically defunct. If a theory can’t explain why a CD explanation (or indeed a solipsistic CD explanation) of all observations is not likely, that theory is probably hosed. Or else you should be declaring you have proved it is likely only you exist and the whole world is your invention. And if you think that would be a weird thing to say, you’ve noticed the problem.

The reason CD scenarios are epistemically improbable is that they require assuming too many things are true for which there is no evidence. That then translates into a high likelihood that it is false in fact, because if it didn’t, then it would be incorrect to say it was epistemically improbable if you know that in fact it’s, say, 50/50. See chapter six of Proving History, especially the second half. The reason a highly needlessly complex theory is epistemically improbable is that it is indeed metaphysically improbable, because it requires more things to exist (or a thing with a much more specified set of properties) than need to exist to cause the actual observations.

If you don’t yet see how that translates, just imagine building a “many worlds” tree in which every logically possible world exists, and then counting the number of those worlds that have our observations but no leprechauns and only atoms (etc.), and realize that means even if those worlds all exist (a la Tegmark), the actual physical (not just epistemic) probability that we are still not in a leprechaun world is extremely high.

This is one reason Eliezer Yudkowsky correctly articulated my demarcation principle in terms of information theory (in this video, timestamp 12:38 to 20:44, which especially makes sense in the context of 5:45 to 9:30): the Kolmogorov complexity of leprechaun worlds is so high, compared to mere “atom” worlds, that even if all possible worlds exist that look exactly like ours, the number of them that are caused fundamentally by leprechauns is extremely small. And that even assumes leprechaun worlds are logically possible. That might not even be true (see The Argument from Specified Complexity against Supernaturalism).

This allows falsification, because things didn’t have to turn out that way. And still could turn out differently…

Again, just to be clear, I fully agree that if the usual model of Christian young Earth creationism were true, there would be plenty of actual evidence we could find, and it is totally conceivable that we could gather enough evidence to have a firmly supported conclusion that the usual model of Christian young Earth creationism is true. However, under your definitions, I think we could never conclude it’s “supernatural”.

Here we can turn the table around to illustrate my point: just as a leprechaun world is less probable as an explanation of what we observe in this world, so also, conversely, would a naturalistic world that explained observations that corresponded to Young Earth Creationism. The elaborate number of sub-hypotheses you would have to invent out of whole cloth to get that to fit would be just as staggering as for the leprechaun-based Periodic Table. Naturalism would therefore at that point be the one with a vanishingly small prior (not a zero probability; just a really, really small one), rather than the other way around.

In other words, at some point, the number of putative supernatural things observed reaches a point where to explain them naturalistically requires a way more complicated theory than just accepting that you are in a supernatural world. At some point, after living at Hogwarts long enough, holding out for a Forbidden Planet explanation ceases to be rational. Yes, maybe it will be that. But it’s really, really unlikely. Because it requires assuming, without evidence, a vast quantity of things to be the case (the whole Forbidden Planet scheme, from entity to creation to function, all the way to its coincidentally being undetected), far more than would be required by a supernatural theory (even accounting for its full Kolmogorov complexity). In other words, the simplest hypothesis then would be that magic just exists, and mind overrides matter.

It just happens that we haven’t observed ourselves to be in such a world. Our world looks the opposite: it looks exactly like a non-supernatural world would look. To credit that as merely a coincidence is to import a massive improbability into your explanation, an improbability that is removed if you just accept that this is not a supernatural world. Therefore, naturalism on present observations is the far more probable theory. And we conclude this on the evidence of the world (all the information in b, the set of all background knowledge, in a standard Bayesian equation—upon which all the probabilities in Bayes’ Theorem are conditioned). Had the evidence of the world been different, so might our conclusion have been.

Therefore…

To use an example: materialism might be true to all possible tests we can employ, but our observable reality might…

Stop there. “Possibly, therefore probably” is a fallacy.

You are done.

No argument of any value can proceed from “it might be that…”

What you need is an argument of the form “it is just as likely that…” (or, even better, “it is more likely that”). But notice how very much harder it is to adduce evidence for that proposition. Arguing for something being likely is a bitch. Arguing that it merely “might” be is easy. Because the result is virtually useless. All things possible are not equally likely. So what is merely possible, is of no use knowing. (Except for its one practical use, which is in ruling out the impossible.)

Because you need evidence for something being likely (even as likely as 50%, for anything that has any theoretical complexity at all). And in the absence of evidence for a proposition, that proposition can only be maintained on a raft of ad hoc assumptions, each of which is not 100% probable but at best 50/50 either way, and often considerably less (after accounting for the evidence accumulated in b…evidence that might have turned out entirely differently, but alas, didn’t). And all those probabilities multiply against each other, creating a diminishing epistemic (and thus metaphysical) probability. The more random stuff you have to throw on it to gerrymander it to fit observation, the less probable it becomes, not more.

You could try to leverage that prior probability up by having a mountain of prior evidence. But we don’t have any (so, that option is not available now). And even if we did (or suddenly found some), it’s really hard for it to be enough. For example, if the movie Inception were a true story and you lived it, then you would have a much better reason than we do now to suspect all reality might be someone’s dream. The prior probability in that case would go way up, because you would have tons of data in b that make it more likely. But that entails our lacking that data makes its prior probability go way down. In fact, down by precisely as much as having that data would have brought it up. Thus, it is our not having that data (even though, logically, we could have by now) that is the test that falsifies the Inception hypothesis. The more we live in the world without encountering any evidence of Inception capabilities, the less likely it becomes that Inception capabilities exist. That is the proper effect of evidence on belief.

And yet, even if you had all that Inception evidence, still counting against your theory that it’s all just some guy’s dream is that such a hypothesis adds one extra ad hoc layer of explanation over on top of an explanation without it. It thus violates Occam’s Razor—which means, it cuts the prior probability down tremendously. Because there still has to be a universe in which the dreamer resides, and not just a universe, but one that explains the dreamer’s existence, the dreamer’s powers, and the particularly bizarre and amazingly coincidental way those play out so as to look exactly not at all like a dream but very exactly like a material world would look by itself. So now you are hypothesizing two universes (a complex and bizarre real one you don’t observe and have to invent out of whole cloth, and the dreamed one you do observe) to explain observations that are perfectly well explained by just one universe (which, all told, is vastly less complex in its required component propositions). Statistically, therefore metaphysically, therefore epistemically, the latter explanation is vastly more likely to be true. On the evidence. And in that case, even with Inception evidence.

However, if you lived Inception, though your “it’s all a dream” hypothesis would be unlikely to be true (unless you gained access to evidence supporting it, e.g. you encountered and met the supreme dreamer and got a convincing explanation from them of how they can exist without a material universe to contain them and why they exist and why they have the weird powers they do and why those powers are playing out in just precisely coincidentally the least expected way so as to look exactly like a universe without any such dreamer), that different set of evidence would entail naturalism is probably false. Because to explain Inception powers and events on naturalism would require an immensely complex Rube Goldbergesque system of assumptions. So if you have no evidence for any of those assumptions (like they did in Forbidden Planet), epistemically (and thus metaphysically) their conjunction is simply improbable. Supernaturalism would be the far simpler theory, and thus on the evidence the one far more likely to be true.

The remaining “possibility” that you are deceived in all this is still accounted for by the probability of that explanation not being zero. But though it may be nonzero, it is still extremely small. You therefore have no rational reason to believe it’s the case. And every rational reason to believe the most parsimonious supernatural theory is very probably true.

It’s just, again, we don’t find ourselves in such a world. No Inceptions for us.

Thus, when I said “the underlying mechanics of quantum phenomena might be physically beyond all observation and therefore untestable, but no one would then conclude that quantum mechanics is supernatural. Just because I can’t look inside a box does not make its contents supernatural,” I was already taking into account the vast body of evidence in b that makes it extremely unlikely that supernatural things are in that box. I don’t have to look in the box to know that. It’s a statistical certainty already, given the evidence outside the box. Evidence that could have been different, but alas wasn’t. If we were Incepting people at Hogwarts, then I’d have a really good reason to suspect what’s in that box is indeed in some way supernatural. I certainly would then have no basis for saying it’s very unlikely to be. But we aren’t Incepting people at Hogwarts. Or anything of the kind. The evidence thus tells us the world is probably the exact opposite of that. The box almost certainly has just mundane natural stuff in it.

Thus…

Just because I cannot look outside the box which is this universe and see the sleeper who is dreaming – that would not make the world any less supernatural under your definitions.

My definitions fully allow that there is a small nonzero probability that this world is indeed just the dream of a disembodied psychic leprechaun, and therefore supernatural after all. But that being possible is irrelevant. Do we have any reason to believe it is likely? Well, no. Even so much as 50% likely? Not even close. And that’s the end of it. And the reason this is so is because the evidence has tracked that way. Logically, it could have tracked differently. It didn’t.

And trying to get around that by inventing an elaborate Rube Golderbesque explanation for why the evidence tracked entirely differently than a dreamworld would be expected to look and behave, in fact tracked in precisely coincidentally the one way that looks exactly not like a dream world but in fact an “atomic” one, doesn’t escape the consequences of probability theory. All the unevidenced assumptions you must then add are vastly greater than any you must add to explain the world just “atomically,” with the effect that their collective improbability sinks the ship.

So…

Thus far, the only actual useful and meaningful definitions that I can find to back “natural” and “supernatural” are based on the modern scientific notion of materialism.

There is no modern notion of materialism. The term has been supplanted with physicalism. Because matter is actually just a condensed (cold) form of energy. Energy is now the fundamental physical unit. It is defined as a vibration (a movement, and thus atemporally just a geometric shape) either of a fundamental massless material (like strings), or of space-time itself (and I think the latter is the more parsimonious explanation). But there is nothing logically necessary about that. Physics could have remained fundamentally nonmental yet reduced to completely different stuff. Unless, of course, there is indeed only one logically possible physics! Which would be a lovely discovery. But alas. No one has yet discovered such a thing.

Does some thing X count as “natural” or “supernatural” under your definitions if X is wholly causally reducible to some mind thing Y which is itself wholly causally irreducible, e.g. Y is (wholly) supernatural?

Any such X is supernatural. Because if X can’t be causally reduced to the nonmental, then X is supernatural. By definition.

The question is, at what point does our failure to causally reduce X make its non-reducibility the more likely explanation of that failure than our lack of epistemic access? When we don’t have epistemic access and it’s obvious why, then the priors hold (what’s in the unopened boxes is probably just like what’s in all the opened ones). When we have to start making up elaborate ad hoc excuses for why we still don’t have epistemic access even though statistically we should have by now, then it becomes a matter of what is more probable: that there is no natural explanation, or that we haven’t discovered it yet. And at some point those probabilities will begin to shift.

And that’s what my many examples illustrated: the more complicated you have to make a causal reduction with unevidenced assumptions, the less likely that causal reduction is. Until evidence changes that assessment (e.g. discovering a far simpler causal reduction, or discovering evidence making the assumptions it depends on very likely). So if a natural reduction has to be inordinately complex in its ad hoc parts, it then becomes the less likely explanation. If a supernatural reduction has to be inordinately complex in its ad hoc parts, then it becomes the less likely explanation. (And here complexity means in terms of information theory, not “actual physical parts,” e.g. God is extraordinarily complex in his specified complexity even if he has no parts at all).

Right now, we observe a world the simplest explanation of which by far has been naturalism. That was not logically necessarily the case (so far as we know). Had we found ourselves in fact to be in the world exactly as described throughout the Bible, then the simplest explanation by far would be a supernatural one. Albeit not the one Christians might be comfortable with, given that it would entail God is actually frail, fallible, and evil.

§

To comment use Add Comment field at bottom or click a Reply box next to a comment. See Comments & Moderation Policy.

Discover more from Richard Carrier Blogs

Subscribe now to keep reading and get access to the full archive.

Continue reading