The Post

Suppose a mad scientist creates an artificial neurophysiological (ANP) apparatus, a many-tentacled device implanted near Smith’s brainstem that controls both his thoughts and behavior.  The mad scientist can remotely control the ANP device’s electrochemical processes to vary Smith’s beliefs and behavior in innumerable and diverse ways.  For example, Smith is dehydrated, and the mad scientist, wanting his victim to be in good health, uses the ANP device to force Smith to drink some water while simultaneously making him believe “I am thirsty and this water will quench my thirst.”  The second time Smith is dehydrated, the mad scientist uses a different electrochemical setting to make Smith believe “drinking this water will grant me superpowers in the afterlife” while producing the same drinking behavior.  Here, the electrochemical process that produces fitness-enhancing behavior also produces a false belief.  The ANP device can even produce “garbage” semantic beliefs that have little to do with the forced behavior, such as making Bill believe that “grass is air” or that “1+1=3” at the same time it causes Smith to drink the water.  The third time Smith is dehydrated the mad scientist does just that; causing Smith to drink the water while also causing him to believe “1 + 1 = 3.”  Indeed, the mad scientist can associate just about any belief with the same drinking behavior.  Such an artificial neurophysiological device is not only metaphysically possible, but it also seems to be physically possible (given that beliefs and behavior can be brought about by electrochemical means).

The mad scientist scenario shows that false beliefs can be associated with fitness-enhancing behavior, even to the point where the false beliefs are wildly unrelated to the external environment, as in dreams.  But if the scenario’s artificial neurophysiology is physically possible, then it’s at least metaphysically possible for an evolved creature’s natural neurophysiology to have the same “disconnect” between semantics and behavior.  As with semantic epiphenomenalism, the semantic content produced by the fitness-enhancing neurophysiological process doesn’t need to have anything to do with the environment.  So even if we accept that syntax and semantics are the same thing, in a purely physical view of the mind the spirit of semantic epiphenomenalism remains: for any given behavior B, there are innumerably many semantic contents C—even C’s wildly unrelated to the external environment—that could be associated with B.  One could argue that the relation between semantic content and behavior is in this way functionally equivalent to SE in spite of the falsity of SE.  Call this view semantic pseudo-epiphenomenalism.

The mad scientist scenario suggests that given naturalism, if semantic epiphenomenalism is not true, then semantic pseudo-epiphenomenalism is, with both choices giving us SE-like behavior.  Both semantic epiphenomenalism and semantic pseudo-epiphenomenalism permit a great divorce between beliefs and behavior (think of the third and final case in the mad scientist scenario).   One reason for the SE-like behavior is this: if semantics and syntax were (perhaps per impossibile) the same property, it would be possible for an agent’s “semantics” to causally influence behavior in a very different way from how a rational agent would behave.  It is of course quite natural for an adherent of the “semantics causes behavior” view to believe that it does so in a way that a rational agent would behave.  Yet while semantics influencing behavior in a rational manner are possible, e.g. “I am hungry” causing the person to eat food, semantics causing behavior in an irrational manner is also possible.  To illustrate, think of an alien frog whose neurophysiology is very different from an earth frog.  In light of the mad scientist scenario, we can conceive of a neurophysiology that causes the alien frog to snap out its tongue to capture the alien fly while also causing the frog believe that “eating this fly will kill me and I don't want to die, therefore I should eat it.”  We might say the alien frog is being irrational in its behavior, but natural selection does not select for rationality, it selects for advantageous behavior.  We can even conceive of the alien frog having a semantic content that has nothing to do with the external environment, as in the third case of the mad scientist scenario, and yet the neurophysiology still causes the alien frog to eat the alien fly.

In response one could put forth the following rebuttal.  Even though naturalism unavoidably entails an SE-type problem--whether via semantic epiphenomenalism or semantic pseudo-epiphenomenalism--the fitness-enhancing neurophysiological properties that are most likely to be selected by natural selection (say that a certain neurophysiology is selectable just in case it’s likely to be selected by natural selection) happen to be those that are truth-conducive.  The mad scientist scenario is contrived and produces certain belief-behavior pairs that are unlikely to obtain in real human physiology.  The most selectable and efficient way for neurophysiology to produce advantageous behavior also produces true beliefs.  Thus, even though the SE-type situation exists for semantics and behavior, luckily for us the physiological relation between semantics and behavior is such that true beliefs usually obtain.

All that may be true, but as an objection against the Probability Thesis it falls short.  A major problem is that a favorable physiological relation between beliefs and behavior does not appear to be knowable from N&E alone.  To illustrate, consider a slightly modified form of the alien scenario, where the neurophysiology of these creatures is quite literally alien to us and radically different from our own (though we don’t know much more about it).  Given this, the mad scientist scenario, and the SE-like situation for beliefs and behavior, for all we know the most selectable and efficient fitness-enhancing alien neurophysiology available to natural selection has a physiological relation between beliefs and behavior that is wildly different from what human naturalists believes about themselves.  So there are possible worlds where the most selectable alien neurophysiology is such that the fitness-enhancing neurophysiology produces mostly false beliefs as in the mad scientist scenario.  Of course, there are also possible worlds where the most selectable alien neurophysiology produces mostly true beliefs.  But there’s no way to determine on N&E alone which sort of neurophysiology is more selectable, in part because the alien neurophysiology is too mysterious and far too different from our own.

Moreover, if we were calculating the likelihood of RA on just N&E and thus without any (further?) background knowledge about what sorts of physiological relations between beliefs and behavior obtain in actual N&E worlds, we would have no reason to suppose Pr(RA|N&E) is high regardless of whether we assume semantic epiphenomenalism or semantic pseudo-epiphenomenalism.  Indeed, in light of the mad scientist scenario the semantic beliefs of the aliens could (at least in the epistemic sense) be just about anything, and thus Pr(RA|N&E) is low or at best inscrutable.  Similarly, Pr(R|N&E) is also low/inscrutable.

One could concede that the Pr(R|N&E) is low but also claim we know some proposition P (perhaps that the physiological relation between beliefs and behavior happens to be benevolent for our species) such that Pr(R|N&E&P) is high.  On that note, consider first the case of drug XX.  Drug XX is such that if one were to take it, there is a 95% chance that it will make one’s cognitive faculties unreliable within thirty minutes of taking it, though those so affected will be unable to detect their own reliability.  The other 5% are immune to the drug’s effects thanks to a gene that blocks the effects of the drug.   Suppose I mistakenly take the drug and neither you nor I have any beliefs as to whether I have the gene.  It would appear that both of us have a defeater for my belief that my cognitive faculties are reliable.  Suppose I later come to believe that a doctor called me saying that I have the gene that blocks the effects of the drug.  Would you have a defeater-defeater for the belief that my cognitive faculties are reliable?  It would appear not.  The belief that the doctor did indeed call me relies on the very cognitive faculties (my own) that were defeated in the first place, and one cannot sensibly assume the truth of the defeatee to defeat the defeater.

In case you missed it, the claim that "evolved creatures having something appropriately called beliefs implies that the creatures have to possess favorable belief-behavioral relationship such that RA would be obtained" was attacked via the semantic epiphenomenalism/pseudo-epiphenomenalism disjunction, which was itself justified via the mad scientist scenario.  You didn't address this sort of objection your latest post to me.

The Post

The Post

The Post