What’s AS/PT symmetry exactly?  What does “AS” and “PT” stand for?  In any case, “ironclad” and “immune to all skepticism” are some strong words.  Certainly there are beliefs I believe correspond to the Real-World-Out-There (RWOT), e.g. the belief that the earth is older than 400 years old, and aren’t completely immune to skepticism.  I recognize that it’s possible that the universe was created five minutes ago exactly as it then was, with false memories, faulty history books, fossils, etc.  Nonetheless, I believe myself to be properly justified in believing that the RWOT isn’t like that. 

While the metaphysics of truth aren’t particularly important with EAAN, we can for the moment say that it’s a correspondence theory of truth; a proposition is true if and only if it corresponds to reality.

Is this statement metaphysical or epistemological?  If it’s epistemological it doesn’t have much relevance to the type of reliability being referred to here.  If it’s metaphysical, it doesn’t have much to do with the type of truth being referred to here.


Against (2.1 - N&E entails SE):

Well, not quite.  A certain kind of kinetic behavior is identical to heat. Heat is roughly the energy of random molecular motion.  For the analogy to hold, every relevant factor would have to be held constant, which wouldn’t work here.  It would be like starting with a container of 13 gallons of water and a few kilograms of hydrogen gas, and saying “remove every atom of hydrogen while keeping the amount of water constant.”  While heat and a certain kind of kinetics refer to the same property, syntax and semantics refer to two different properties; here it is conceivable to vary one and keep the other constant.

Perhaps so, but it doesn’t really address the issue here.  N&E entailing SE doesn’t entail that supervenience doesn’t exist.  Nor does the argument for it N&E entailing SE.  To recap that argument:

1.      If holding the syntax of beliefs constant while varying the semantics of belief would not change behavior, then the semantics of the belief are not causally relevant to behavior, even if the syntax of the beliefs are.

2.      Holding the syntax of beliefs constant while varying the semantics would not change behavior.

3.      Therefore, the semantics of the belief are not causally relevant to behavior even if the syntax of the beliefs are (follows from 1 and 2).

It might be that due to nomic supervenience, premise (1) is a counterpossible, i.e. a counterfactual with an impossible antecedent.  That doesn’t imply that the first premise is false, however.  We can easily conceive a given neural structure having the same neurophysiological properties but having different semantic beliefs.

If you do not believe the above argument is sound, which premise do you reject and why?

Thanks for clearing that up.  I think we can make the argument clearer and more rigorous by having “message syntax” and “message semantics” to distinguish between “belief syntax” and “belief semantics.”  The argument then goes as follows:

{1} Message semantics affects human behavior independently of message syntax. {premise}
{2} If message semantics affects human behavior independently of message syntax, then belief semantics affects human behavior independently of belief syntax. {premise}
{3} Belief semantics affects human behavior independently of belief syntax. {from 1&2}
{4} (2.1.2 - Holding the syntax of beliefs constant while varying the semantics would not change behavior.) is false. {from 3}

For premise {1}:
{1.1} The syntax of messages is their letter and word arrangement. {premise}
{1.2} The meaning of messages (message semantics) affects human behavior independently of letter and word arrangement (message syntax). {premise}
-example: when you discover a double meaning in a message, you react differently.
{1.3} The message semantics affects human behavior independently of message syntax. {from 1.1&1.2}

In this case the objector could attack premise 2 on the grounds that the relevant thought experiment confirms the antecedent and the corresponding relevant thought experiment for belief semantics/syntax disconfirms the consequent.  What’s the relevant difference between the two scenarios?  The message semantics ultimately translates into beliefs about what the message says.  These beliefs have two different kinds of properties: syntactic and semantic.  The semantic epiphenomenalist can say that message semantics are causally efficacious by virtue of their corresponding syntactic beliefs, not by their corresponding semantic beliefs.


Pr(R|N&E&SE) is low or inscrutable?

What I was talking about is conceiving of a different supervenience relation such that all semantic beliefs are false whereas the given syntax is held constant.  Under semantic epiphenomenalism, the semantic properties of the belief could be anything and it wouldn’t matter (e.g. if semantics supervenes on syntax, this supervenience relation that determines semantics could yield any semantic belief and it wouldn’t matter) with regard to how we behave.  For example, whether we believed that 2 + 2 = 4, 2 + 2 = 5, or 2 + 2 = 906 makes no difference.  Our beliefs could even be unrelated to the external world (as in dreams) and it wouldn’t affect our behavior at all.  If the semantic property of a belief could be anything and it wouldn’t matter, the probability that the substantial bulk of our beliefs is true would be rather low (or else, at best, inscrutable).

Sure, but the odds that a powerful correlation between true beliefs and adaptive behavior would appear rather low given N&E&SE alone, since thanks to SE the adaptive behavior could have any semantic beliefs at all and it wouldn’t matter (e.g. if semantics supervenes on syntax, the supervenience relation could produce any semantic belief at all without affecting behavior).

Perhaps I should adopt a more Plantingan approach about this.  Think not of us but alien creatures on some alien world where N&E&SE is true with respect to them.  In this world, we may assume that the semantic properties of beliefs supervene on syntax (the neurophysiological properties), but the supervenience laws SL regarding syntax and semantics are for all we know wildly different from what human naturalists believe about themselves.  Since we don’t know how friendly SL is for true beliefs (it’s certainly conceivable for a supervenience relation to yield mostly false beliefs), and that the semantic beliefs could be anything and it wouldn’t matter, it seems that Pr(RA|N&E&SE) is low or at best inscrutable.  Similarly, the likelihood of R given just N&E&SE appears low or at best inscrutable.

Furthermore, it seems that under such a situation RA would be defeated, and would remain undefeated even if we knew that RA’ was true and that the aliens themselves believe there is overwhelming evidence for RA.  But if RA would be defeated here, then it seems that R would be defeated in our case as well.

And if monkeys were flying out of my butt then “Pr(I’d stage an amazing Wizard of Oz production) is high” would be true.  But without any support for the antecedent we needn’t accept the consequent.  There’s nothing in N&E&SE itself that makes R high, and indeed given just N&E&SE, R appears to be low (or at best, inscrutable).  To better see this perhaps it would help if I adopted a more Plantingan approach as I did with afunugsamongus.

Think not of us but of alien creatures on some alien world where N&E&SE is true with respect to them.  We know that natural selection selects for the sort of syntax (i.e. neurophysiological properties) that produces advantageous behavior, allowing the creature to better adapt to the environment.  Yet due to semantic epiphenomenalism we also know that the semantics could be literally anything and it wouldn’t matter.  For example, whether they believed that 2 + 2 equals 3, 4, 5, 7, or 906 wouldn’t affect behavior; similarly so if their semantic beliefs were unrelated to the external world, as in dreams.  If semantics supervenes on adaptive syntax, the supervenience relation that determines semantics could yield any semantic belief at all and it would make no difference.  With the semantic beliefs left to epistemic chance then, the odds that the substantial bulk of their beliefs would be true appears low, or else inscrutable.  Similarly, the likelihood of R for us given just N&E&SE appears low, or else inscrutable.

Incidentally, your summary is a pretty good one, though I’d lose the ugly numbers.  Maybe something like this?

2.5a) If (N&E&SE entails A), then (Pr(R|N&E&SE) is low or inscrutable)

2.5b) N&E&SE entails A.

2.5 Therefore, Pr(R|N&E&SE) is low or inscrutable

I don’t think you understood that passage correctly.  Plantinga was suggesting that indicators in and of themselves have nothing to do with (semantic) beliefs.

I never said nor implied that the rule of disjunctive addition gave warrant to “Pr(R|N&E&SE) is inscrutable.”  Fortunately, since my disjunctive addition approach seems to have confused you (despite my repeated and apparently failed attempts at clearing up said confusion), I’m adopting a different method of justifying 2.5.

Why? It doesn't matter what the semantic beliefs turn out to be, nor does it matter what the supervenience relation turns out to be.  Both semantic beliefs and the supervenience relation are invisible to natural selection. What do you mean by “encoding the environment”?  If you’re referring to semantic beliefs, natural selection cannot select for “different NCS that encode the environment better than others” unless by fantastic luck the supervenience relation between syntax and semantics is such that adaptive syntax is associated with the great majority of the beliefs being true.  But there’s nothing on N&E&SE alone that says the supervenience relation will be that way; the supervenience relation could yield any belief at all (e.g. whether they believe that 2 + 2 equals 3, 4, 5, 7, or 906) and it wouldn’t matter.  Left up to pure chance, it seems very unlikely that these aliens will have the substantial bulk of their beliefs be true.  Their semantic beliefs are, after all, a useless byproduct of their biological processes, and it makes no difference what they turn out to be.

Let me introduce a slightly different argument:

(2.5a) If (N&E&SE entails A), then (Pr(RA|N&E&SE) is low or inscrutable)

(2.5b) If Pr(RA|N&E&SE) is low or inscrutable, then Pr(R|N&E&SE) is low or inscrutable.

(2.5c) N&E&SE entails A.

(2.5) Therefore, Pr(R|N&E&SE) is low or inscrutable

To help guard against bias towards our own species, I’d first like to argue for (2.5a).  In the case of the aliens it seems we don’t know what the semantic beliefs are like because not even what the supervenience relation is like is deducible from N&E&SE alone.  That said, if the supervenience relation could produce any semantic beliefs at all and it wouldn’t matter, even to the extent of the semantic beliefs being unrelated to the external world (as in dreams), then left up to chance Pr(RA|N&E&SE) appears low, or else inscrutable.

It depends what you mean by beliefs being subjective experience of neural cognitive representation. If you're referring to semantic beliefs supervening on syntax, I'm fine with assuming that. Indeed, I've more or less been assuming that under naturalism. But this supervenience isn't enough to prevent Pr(RA|N&E&SE) from being something other than low/inscrutable, as I explained earlier.

If semantics supervenes on syntax, the semantics can be indirectly selected for, but at the end of the day it doesn't matter what those semantics turn out to be. Again, the supervenience relation that yields semantic beliefs could be anything for these aliens and it wouldn't matter.

If this is talking about semantics supervening on syntax, great, but as I indicated earlier this does nothing to prevent Pr(RA|N&E&SE) from being low/inscrutable.

No, I'm doing nothing of the sort, because the type of disjunction I'm using is an inclusive-or. So if I know that one disjunct is true, I can add any other disjunct I want, even one that is completely unwarranted. For example, the following is a valid use of the rule:

1. "2 + 2 = 4"
2. "2 + 2 = 4" or "there exists a married bachelor"

The second statement is true because at least one of the disjuncts is true, even though the second disjunct is known to be false.

Again, what do you mean by semantic beliefs being “encoded by neural correlates”?  If you’re talking about semantics supervening on syntax, I’m fine with that.  But then that wouldn’t address my objection here.  The supervenience relation could yield any semantic belief at all (e.g. whether they believe that 2 + 2 equals 3, 4, 5, 7, or 906) and it wouldn’t matter.  There’s nothing on N&E&SE alone that says the supervenience relation of these aliens are such that it would yield mostly true beliefs.

An experience that, thanks to SE, doesn’t need to be there.  Heck, not even SA1 needs to be there.  Going back to the alien scenario, suppose the senses of the aliens are such that they see a three dimensional landscape.  Their beliefs about their sense experience could again be anything and it wouldn’t matter.  The alien neurophysiology that brings about the experience of SA1 could even bring about “garbage” beliefs unrelated to their sense experience, as in dreams.  There’s simply no reason on N&E&SE alone that the beliefs being generated by the alien neurophysiology will be mostly true.  Given that the beliefs could be just about anything, Pr(RA|N&E&SE) appears low, or else inscrutable.

I don’t need to do that at all.  I can just point out that there’s no reason to suppose the hypothesis to be true on N&E&SE alone, and that he supervenience relation could yield any semantic beliefs at all without affecting the decisions of natural selection.  It’s not that RA is impossible on N&E&SE, it’s that RA is unlikely (or else, that’s likelihood is inscrutable, though my money is on “unlikely”).

In any case, let’s go back to H1 and H2:

You never did quite clearly elucidate what this meant.  At first blush, H1 and H2 seem to be referring to semantics supervening on syntax, which is very compatible with my argument.  But if that’s not what they’re talking about, what are they talking about?

Yes, I did add “Pr(R|N&E&SE) is inscrutable” via disjunctive addition, though the disjunctive addition uses an inclusive-or.

First, all this is a red herring.  I was talking about the aliens having garbage beliefs, not us.  Note the paragraph you quoted was speaking of Pr(RA|N&E&SE) being low/inscrutable.  So do you have an objection against what I actually said?

Second, you have again kind of ignored the question I asked: what do you mean by semantic beliefs being “encoded by neural correlates”?  Are you talking about semantics supervening on syntax?

They don’t?  You might want to look up the term supervenience.  Nomic supervenience is the type of supervenience I had in mind.

The warrant for the premise (“Pr(R|N&E&SE) is low or Pr(R|NE&&SE) is inscrutable) is via disjunctive addition, but that doesn’t mean there’s any warrant for “Pr(R|N&E&SE) is inscrutable.”

I honestly do not; if you know of a good objection though, please point to a specific premise of the argument and explain why you believe it is wrong.

What part of evolution does Plantinga get wrong?

You seem to have forgotten my argument for 2.5.  I’ll reproduce it here for your convenience.

Let A represent the following:

A: the semantic properties of the beliefs could literally be anything at all and it wouldn’t matter (e.g. if semantics supervenes on syntax, this supervenience relation could yield any semantic belief and it wouldn’t matter) with regard to how we behave.

(2.5a) If (N&E&SE entails A), then (Pr(RA|N&E&SE) is low or inscrutable)

(2.5b) If Pr(RA|N&E&SE) is low or inscrutable, then Pr(R|N&E&SE) is low or inscrutable.

(2.5c) N&E&SE entails A.

(2.5) Therefore, Pr(R|N&E&SE) is low or inscrutable

I was arguing for 2.5a.  Perhaps you disagree with 2.5b?

I guess I should define some terms.  When A-properties supervene on B-properties, that is to say that B-properties determine (i.e. always yield the same) A-properties.  Nomic supervenience refers to the supervenience relation obtaining by nomic necessity.  This type of supervenience seems to fit what you had in mind for “bridge laws.”

H3 is a combination of claims: (a) nomic supervenience must be real (though we are ignorant of the specific nomic laws); and (b) R is true (via the type of nomic supervenience we have).  I’m willing to grant the naturalist (a), but assuming (b) under naturalism would be a bit question begging in EAAN discussions.  You can’t just assume the defeatee to defeat the defeater.

Or (c), that the naturalist has insufficient warrant for H3 thanks to EAAN.  This is a de jure objection rather than a de facto one.

You’re forgetting something: the bridge laws could yield any semantic belief at all without affecting behavior; hence, A is still true here and Pr(RA|N&E&SE) still appears low.

First, I deny that one can replace 2.5 with a different true premise and get a conclusion that is logically contradictory to 2.6, i.e. get a conclusion 2.6’ such that it’s logically impossible for both 2.6 and 2.6’ to be true.  For example, suppose you replace the premises with different true premises such that 2.6’ is “Pr(R|N&E) is low or Pr(R|N&E) is high.”  Is 2.6’ logically inconsistent with 2.6?  No, one logically possible way both could be true is if “Pr(R|N&E) is low” is true.

Second, a sound argument is by definition one that is both valid and has all true premises.  So if an argument is both valid and has all true premises, the argument is sound, period.  As long as those two conditions are met, it matters not whether e.g. a different conclusion can be made with different premises.