Self-defeat, cute tricks, and the futility of syllogisms
Huemer recently presented an attempted, refutation of determinism.
It was apparently rejected for publication fifteen times, apparently because critics could cite objections. I’m with Huemer on that, at least: the fact that one can raise objections isn’t, itself, a good reason to reject an article for publication. Of course, I haven’t read either the reviews or the editors’ decisions; maybe there were good reasons to reject the article in the end. Without access to those details, I can’t be sure.
In any case, Huemer did eventually get it published, and has opted to make it available as a post on his blog.
Huemer begins with an intuitive idea that appears to serve as the foundation of his argument. This intuition is motivated by remarks from Epicurus and J. R. Lucas. First, a quote from Epicurus:
The man who says that all things come to pass by necessity cannot criticize one who denies that all things come to pass by necessity: for he admits that this too happens of necessity.
I dislike this use of “cannot” here. Cannot has a perfectly good descriptive reading, yet here it only makes sense if it is used in a normative sense, in which case, why not just use normative language, and say “would not be justified in criticizing one who denies…” This is not necessarily a critique of Epicurus. I don’t know what the original remark looked like, and we’re clearly dealing with a translation. Yet this only makes matters worse: we’ve had over two thousand years of hindsight for philosophers to translate ancient words of wisdom in a more perspicuous way that avoids misunderstandings. Of course, Huemer probably isn’t responsible for the translation, but my objections aren’t only against Huemer’s argument, but the background in which such arguments emerge. Analytic philosophy is supposed to be clear, precise, and rigorous. In practice, I typically find that it isn’t.
Analytic philosophy is supposed to be about clarity and precision. Descriptive/normative distinctions are important, so why not be clear about them?
In any case, I don’t share the intuition that if you’re a determinist, you can’t criticize those who deny determinism, nor do I have any sympathy for the notion. It appears to convey the idea that if you think determinism is true, you shouldn’t be criticizing people who reject determinism, since they are incapable of not rejecting determinism.
My inclination to criticize others isn’t predicated on whether their actions aren’t determined. So I simply reject whatever presuppositions would prompt one to think that you “cannot” (why not shouldn’t? Why the lack of clarity?) criticize someone for rejecting determinism. I can (and I will) criticize whomever I want.
Philosophy isn’t some magical force that’s going to stop me, as though it were some kind of gravitational pull that yanks us towards only saying, thinking, and doing things consistent with some analytic philosopher’s conception of what we’re “justified” in doing. If a philosopher thinks that I should not criticize a particular position given my currently held beliefs, then that is another matter altogether. In such a case, the philosopher will need to convince me that I should not, but “cannot” is presumptuous and dialectically unproductive.
Huemer provides a remark from J.R. Lucas apparently conveying a similar notion to the Epicurus quote above:
Determinism … cannot be true, because if it was, we should not take the determinists’ arguments as being really arguments, but as being only conditioned reflexes. Their statements should not be regarded as really claiming to be true, but only as seeking to cause us to respond in some way desired by them.
Why shouldn’t we take a determinist’s argument to “really” be an argument? Even if an argument were the product of a “conditioned reflex,” why would that make it not an argument? Is it a precondition of arguments that they be presented in a way that isn’t determined? Suppose I ask a computer program to present an argument, and it presents the following:
(P1) All dachshunds are dogs. (P2) Pitanga is a dachshund. (C) Therefore, Pitanga is a dog.
In case you’re wondering, this is Pitanga:
As you can see, Pitanga is a dachshund. Is Pitanga a dog? Well, I’m not sure. Is the above an argument? Apparently not, since it was written by a computer, and not by me. But if I suddenly say the following:
“All dachshunds are dogs. Pitanga is a dachshund. Therefore, Pitanga is a dog.”
…is this an argument? Are premises and conclusions only arguments when they’re asserted by someone? Wait: I endorse determinism, and don’t believe in free will. So apparently this isn’t an argument, either. Only people who believe in free will can present arguments. We should go back through the historical record of past philosophical work and identify all those “arguments” presented by determinists, and reject them as non-arguments. That will really clear things up.
Since text doesn’t convey tone, that was sarcasm. I don’t think there’s anything intuitive, or even plausible about the suggestion that a “determined” argument isn’t actually an argument. I’m not even sure it’s sufficiently clear to evaluate what such a proposal is supposed to be claiming. I’m loathe to appeal to intuitions, but the notion that if determinism is true that we therefore couldn’t present “arguments” seems obviously false, even asinine. I’ve never heard anyone claim that a set of premises and a conclusion are only an “argument” if they’re articulated by someone who doesn’t endorse determinism.
According to Huemer:
The intuitive idea is that determinism is self-defeating when you apply it to beliefs about the subject of free will and determinism itself.
A classic self-defeating argument, of course. I despise self-defeat arguments, typically because they purport to show that some argument defeats itself only given presumptions or commitments that the person presenting the argument isn’t required (and probably doesn’t) have to endorse. That is, self-defeat arguments typically work only when the critic presenting the self-defeat argument presumes some conditions or standards must be met for an argument to be sound, but that the argument would require one to deny or negate those presuppositions. Yet the person presenting such an argument doesn’t have do endorse the necessity of the presuppositions. As a result, such arguments only tend to work by importing extraneous philosophical assumptions that the critic proposing the self-defeat argument would have to argue for independently, and which the proponent of the argument they’re charging with self-defeat could (and in my case, often does) deny.
In this case, Huemer provides the assumptions in question:
Per Epicurus, it implies that you can’t criticize anyone for believing in free will, nor (presumably) can you say that anyone should believe determinism. In its most common (physicalistic) forms, per Lucas, determinism implies that good reasons play no role in explaining why one believes determinism itself. So the determinist couldn’t hold that he himself knows determinism to be true.
It would only follow that I couldn’t criticize someone for rejecting determinism if I agreed that I needed good reasons to do so. But I don’t. What does Huemer mean by a good reason? I endorse antirealist conceptions of reasons, but not standard realist ones, e.g. the notion that there are irreducibly normative reasons. It’s unlikely I think that there are any such things as the sorts of “good reasons” Huemer has in mind here.
The view is only “self-defeating” if the argument required or appealed to the notion that one needed good reasons. But a person presenting the argument doesn’t have to think we need “good reasons” of the relevant kind to endorse determinism or criticize people for rejecting it. We can (and I do) simply deny there are any “good reasons” of that kind and I deny such reasons are required to present arguments for determinism or to criticize people for rejecting determinism.
Such arguments and criticisms would only be self-defeating if I were compelled to accept what my interlocutors already accept, and if they are compelling me that far, then why not the whole way? Why not simply compel me to agree with them outright, why the song and dance about what I supposedly “cannot” do? It's only self-defeating if these extraneous standards were necessary and were required for something to be an argument/criticism. It’s, at best, a kind of conditional self-defeat, conditional on the critic being correct about background assumptions that I already rejected prior to them presenting the self-defeat argument.
Even if they were correct about those background assumptions, why the hell would I endorse a position that conflicted with them? Of course if I shared Huemer’s metaphilosophical and philosophical presuppositions - the kinds that would entail that Huemer was correct about determinism or correct about the premises of the argument presented here, then I’d endorse the conclusion. But that’s trivial. Huemer might as well argue that if we agree with his presuppositions, then we should agree with the entailments of those presuppositions. Well, sure, okay, perhaps I would. But you’d have to convince me of your presuppositions first.
According to Huemer:
My idea was related to these. It was that in thinking about any issue, one always presupposes certain norms governing belief. E.g., that you should avoid falsehood, or that you should base beliefs on evidence. But any such norm, I think, is incompatible with the truth of determinism. So if you think determinism is true, you’re in an inherently self-defeating position: You’re committed to rejecting norms that you are implicitly presupposing.
This is interesting. It’s likely that I endorse the norms Huemer mentions but I don’t think they’re incompatible with the truth of determinism. Let’s see.
The cute trick in question is supposed to be a refutation of determinism. Here it is:
We should believe only the truth. (premise)
If S should do A, then S can do A. (premise)
If determinism is true, then if S can do A, S does A. (premise)
So if determinism is true, then if S should do A, S does A. (from 2, 3)
So if determinism is true, then we believe only the truth. (from 1, 4)
I believe I have free will. (empirical premise)
So if determinism is true, then it is true that I have free will. (from 5, 6)
So determinism is false. (from 7)
As is typical of such arguments, the premises are too underspecified to evaluate. I don’t know what Huemer means by “should” or “truth.” I don’t endorse irreducibly normative or realist conceptions of “should,” nor do I endorse the correspondence theory of truth, so it’s likely I would reject this if the terms in question presuppose one or both of these notions. As such, P1 strikes me as difficult to evaluate without knowing precisely what Huemer means by the terms “should” and “true.” It’s likely that I’d regard the premise as unintelligible or false.
At face value though, I don’t have the intuition that we should only believe what’s true in a truth correspondence sense; that is, even if I put on my analytic philosopher’s hat and consider whether I think we ought to believe only what’s true, my inclination is to deny this. For example, I’d rather believe useful fictions than “truths” that undermine my ability to achieve my goals. Not only this, but the correspondence theory of truth presupposes a mind-independent world and a world-independent mind, neither of which are possible. As William James says in his classic Pragmatism, truth can be described as correspondence with reality, but it depends on what is meant by either term (cf. Lecture 6).
As such, it’s likely that, insofar as I could make sense of the first premise, I’d reject it.
The same holds for P2: if the notion of “should” implies irreducibly normative reasons, then I’d reject it. On my antirealist conception of normative claims, the best way to intelligibly interpret such claims is to treat them as consistency relations, so we might say something like:
If it is consistent with S’s goals to do A, then S should do A.
In which case, I’d reject P2 as well. I don’t think that if it would be consistent with someone’s goals to do something then they are therefore capable of doing that thing. I’d probably reject other conceptions of normative claims that don’t reduce to descriptive claims, so I’d either reject this premise because it presupposes characterizations of the terms used that I reject or because, to the extent that the terms are used in ways I would accept, the premise would be straightforwardly false.
P3 doesn’t seem like a problem.
Now consider C2:
So if determinism is true, then we believe only the truth. (from 1, 4)
This seems like a reductio of the argument itself, and not an entailment of (or reductio of) determinism. If your argument somehow leads you to characterizing determinism as entailing that if determinism is true, then we can only believe the truth, maybe you should revisit the argument’s premises, because that’s a ridiculous conclusion.
Not only is it ridiculous, but it equivocates on what the truth is under a standard correspondence notion: a relation between the output or content of a mind (usually sentences or ideas) and reality itself, but such a construal of determinism shifts away from that to an identity notion of truth, i.e. that “truth” is somehow identical with reality. But determinism is a thesis about causality, not truth, and truth is a relation with the world, not the world itself.
Nothing about determinism seems to prohibit us from sometimes forming false beliefs. Nothing about an event being causally necessitated by antecedent events seems to render this impossible (logically or otherwise). Quite the contrary, antecedent causes can necessitate that we sometimes misperceive, or misconstrue, or misunderstand something, which, so long as we were disposed to form beliefs in accordance with the cognitive processes in question, would necessitate that we formed false beliefs.
I suspect the problem here is the very beginning of Huemer’s argument: P1 and P2 may seem superficially appealing, but I think their appeal derives largely from their underspecificity.
I am generally in favor of believing what’s true, mostly because I think the consequences of having true beliefs tend to be good. Also, I don’t think I could readily opt to deploy my epistemic practices in ways that led me to form false beliefs. I can’t choose to believe I’m a millionaire or that I am an immortal, even if I’d like to.
But I don’t know what Huemer means by the claim that “We should believe only the truth.” I suspect this claim is underwritten by precisely the kind of dubious analytic conceptions of normativity and truth that I reject, and that I think others should reject, because I think they’re the result of conceptual and linguistic confusions. Those confusions result in such premises appearing to seem plausible but, when we dig a bit deeper, we must delve into the dark underbelly of analytic philosophy, where premises that seem copacetic conceal unstated presuppositions. For comparison, suppose I began an argument with the premise:
(P1) We should be kind to others.
Who would deny that? But suppose by kind I meant “demonstrating compassion in accordance with the will of God.” Now an explicit articulation of this premise would actually be:
(P1*) We should demonstrate compassion to others in accordance with the will of God.
Many people would be reluctant to reject P1, but would be perfectly happy to reject P1*. Just the same, many people would be reluctant to reject P1 and P2 in Huemer’s arguments, but may be more reluctant to do so if the philosophical presuppositions implicit behind these arguments were made explicit, just as they are here.
Unfortunately, I believe Huemer’s argument is a cute trick, unironically. I think such cute tricks are endemic to philosophy, and are one of the central culprits behind the contemporary analytic philosophy constituting a largely empty and useless enterprise of linguistic smoke and mirrors, where one puts forward premises using underspecified and superficially appealing terms and statements which are intended to invite affirmation and to discourage denial—precisely by presenting themselves in an ambiguous and underspecified form that is designed to be superficially appealing to everyone and to invite incredulity or scoffing were one to reject them.
Compare to P1 and P1*, here. Imagine you’re in a public debate, up on the podium, and your rival hits you with:
“Do you think we should be kind to others? Yes or no?”
What would be the consequences of saying “no”? Probably not very good. You’d look like a jerk. And if we swapped out the premise for an obvious descriptive claim rather than a normative claim, e.g., “Do you think the sun exists?” and you said “no,” you would look like an idiot rather than a jerk. Either way, to deny such obvious claims is to reveal to one’s audience that one is stupid, a jerk, or both (technically speaking, you’d be a stupid jerk).
You would almost certainly endorse these claims if you knew that behind them was a construal of the notions of kindness to others and the existence of the sun that you think is true. But suppose you were patently aware, since the other person made it very clear to you prior to any public discussion, that by P1 they meant P1*. That is, what they mean by “we should be kind to others,” is just that “we should demonstrate compassion to others in accordance with the will of God.” Suppose they told you this, and told you that they’d ask you whether we should be kind to others in advance. Once again, you’re asked:
“Do you think we should be kind to others? Yes or no?”
Assuming you reject P1*, what are you supposed to do? This is obviously a trap. If you say “no,” you look like a jerk, but if you say “yes,” they can say:
“Aha! So you do believe in carrying out the will of God!”
The problem with passing arguments through the sieve of simplified syllogistic formalism is that they conceal the hidden presumptions behind the premises. Instead, they present us with a sentence that could be interpreted in multiple possible ways, some of which we’d accept, and some of which we wouldn’t.
At the same time, the premises in such arguments may be intended merely to convey specific, formal philosophical positions. Yet they are conveyed in ordinary language (or in language that superficially resembles ordinary language), but ordinary language doesn’t work the way formal arguments are intended to function.
The truth or falsehood of the premises presented in a formal argument aren’t supposed to be judged on the basis of pragmatic implications that are not logically entailed by the claims in question. For instance, suppose a moral error theorist is asked, “is it objectively wrong to torture babies for fun?” If the error theorist says, “no,” it does not logically follow that the error theorist is indifferent to baby torture. Yet such a remark may be taken to pragmatically imply that they are, and may imply that they aren’t personally opposed to baby torture, that they don’t disapprove of it, that they wouldn’t intervene were it occurring, and so on.
Yet such implications aren’t entailed by moral realism. Nevertheless, someone assessing the premises in a formal argument may endorse premises an error theorist would reject because the denial carries such implications, implications that are not strictly entailed or intended by the person presenting the argument. Formal arguments are supposed to hold such pragmatic implications abeyance. Yet in practice, those presenting formal arguments often exclude any explicit articulation of their presuppositions, or omit technical terminology that would render the premises they do include more explicit. The result is that formal arguments often include premises the affirmation or denial of which would be reliably interpreted by people to carry the pragmatic implications in question, even though such implications ought to be irrelevant to assessing the truth or falsehood of the premise in question.
As a result, formal arguments often exploit an ambiguity between the formal, restricted, “purely semantic” meaning of the premises in question, and the pragmatic implications one would expect others to associate with the premises in question in accordance with the conventions of ordinary language, which is unrestricted and isn’t “purely semantic.”
In short, formal arguments often derive their persuasive force from a subtle and insidious form of equivocation, and as a result often function more as rhetorical devices than as clear and cleaned up articulations of an argument. Quite the contrary: they exploit the superficial appearance of the latter to bolster their persuasive appeal. In other words, formal arguments often function as covert rhetoric that avoids the blowback that would be associated with more explicit rhetorical appeals. Ironically, logic and formalism, which were devised with the intent to avoid rhetoric, have been conscripted into its service.
As always, arguments conveyed in contemporary analytic philosophy rely on a kind of vacillation between ordinary (or pseudo-ordinary) language and philosophical language, where the affirmation or denial of the latter leeches appeal from covert pragmatic implications associated with the former.
This is the key to the cute trick Huemer has performed. It’s a cute trick after all, but not for the reasons I suspect Huemer takes it to be. It is a cute trick in the sense of a stage magician’s cute trick, only the prestidigitation is performed with words rather than a hat and a wand.
To deny that “we should believe only the truth,” may suggest to others that one is unconcerned with or doesn’t value the truth (or at least not as much as those who’d affirm the premise), that one doesn’t think the purpose of exercising one’s epistemic practices is the pursuit of knowledge, or perhaps that one is disposed toward dishonesty, or self-deception, or is less serious about making good arguments, and…you get the picture. One could probably fill this in with a variety of negative associations, all of which amount to making one’s interlocutor appear to be a stupid jerk. Or to put it less technically, to deny the premise is to reveal that one fails to love wisdom, a betrayal of the eponymous aspirations of philosophy at their root.
To deny that we should only do what we can do likewise carries unpalatable implications. I deny the ought-implies-can principle because I deny conceptions of normativity that would render it plausible. As I mentioned above, I don’t endorse non-reductive conceptions of “should.” Instead, I hold terms like “ought” and “should,” as intelligible and reasonable insofar as they express reductive descriptive claims about consistency relations between means and ends. And on this conception, the premise seems obviously false to me, and no longer carries any undesirable pragmatic implications. Consider this premise:
P2* If it would be consistent with someone’s goals to X, then they can X.
This is obviously false, and rejecting it wouldn’t make you look like an incompetent jerk. Imagine telling someone that if it were consistent with their goal of not being eaten by a tiger to fly away, that therefore they could fly away. This is absurd on a descriptive account, and would be ridiculous even on an account that appealed to irreducibly normative facts. Of course, a realist might just pivot in the opposite direction, and insist that “Of course they shouldn’t fly away, because they can’t.” But why should we suppose that irreducibly normative facts must be sensitive to whether we can comply with them?
Yet many would interpret the rejection of the claim that “If S should do A, then S can do A,” to carry particular implications. Specifically, if you reject this principle, then your evaluation of other people’s actions is no longer contingent on what they were capable of doing. And once you abandon this precept, you could blame anyone for anything, even if there’s nothing they could have done to change the outcome. A physician arrives on the scene to find a person had been decapitated? Well, they’re blameworthy, because they didn’t reattach their head and revive the person. After all, the fact that they couldn’t do so is irrelevant. To deny P2 is to suggest that one’s attributions of praise and blame aren’t driven by reasonable evaluations of people’s capabilities.
Yet a determinist can attribute praise and blame to others based on considerations other than whether a person “could” have done otherwise, where “could” is understood in a sense suspiciously indicative of libertarian conceptions of free will (there are compatibilist construals of “could have done otherwise,” and the like, but I won’t get into those here, and, in any case, I don’t endorse compatibilism).
I have no interest in threading the needle on certain turns of phrase, rendering the notion of whether someone “can” do something consistent with some notion of free will. Such efforts strike me as, for the most part, merely verbal, and largely motivated by a desire to construct sentences which one feels comfortable explicitly affirming that both (a) are consistent with one’s metaphysical commitments and (b) avoid the undesirable pragmatic implications associated with flat denials of this or that principle or claim.
Huemer also anticipates and responds to some possible objections, titled “Confused objections.”
The first confusion is:
“Premise 1 begs the question, because the determinist would just reject all ‘should’ statements. No one ever ‘should’ do anything, because no one ever has any alternatives available!”
A determinist doesn’t have to insist the argument begs any questions to reject P1 (or P2, for that matter). They can reject P1 (or P2) for a variety of reasons without insisting that the argument is question begging (or at least question begging in some way they object to).
However, Huemer says something interesting here:
Now I’d like you to notice that that implies that all valid arguments are bad. In a valid argument, the conjunction of the premises, P, entails the conclusion, C. Therefore, it is also true that the denial of the conclusion, ~C, entails the denial of the premises, ~P. That, again, is true of all valid arguments by definition. So if you use your rejection of an argument’s conclusion to evaluate the premises, it would always be true that you would reject at least one of the premises. As far as I can tell, that’s all the objection is pointing to.
I don’t know what Huemer means by “bad.” But it does imply something about all valid arguments: we can always, upon seeing a conclusion we don’t accept, reason backwards that there must be at least one premise we reject, identify one or more premises, and reject them. Such a practice may result from motivated reasoning or a questionable attempt to evade a conclusion one doesn't want to accept, but reasoning in this way doesn’t have to be a product of questionable reasoning practices. Reasoning in this way is a normal product of attempting to reconcile conflicting beliefs and commitments, a standard instance of reflective equilibrium. When confronted with conflicting beliefs (endorsing premises that entail a particular conclusion, but rejecting the conclusion) one isn’t necessarily obligated to resolve the conflict by endorsing the conclusion: you can come to reject one of the premises. As the saying goes, one philosopher’s modus ponens is another’s modus tollens (Williamson, 2022).
A proponent of determinism could, on encountering Huemer’s argument, reject P1, P2, or both (they may have something to say about P3 as well).
Next is Confusion #2:
“The argument has an equivocation: (1) uses the epistemic ‘should’ but (2) uses the moral ‘should’.”
Huemer’s reply is interesting:
I don’t really think there are multiple senses of “should”, but even if there are, there is no reason to think the argument uses different senses in 1 and 2. The “ought implies can” principle should apply to all “should”s. There is no sense of the word “should” on which it’s correct to say that you should do the impossible. So premises 1 and 2 are both true using the epistemic “should”.
The first remark is interesting: Huemer doesn’t think there are multiple senses of “should”? What does Huemer mean by this? Ordinary use of the term “should” likely both (a) isn’t used in a rigid and uniform way and (b) probably differs from philosophers who (c) themselves endorse a variety of different conceptions of should. Is there a single correct account? If so, what is a correct account of? Hopefully not a descriptive account of actual usage: it’s an empirical question what people mean when they use the term “should,” and I’d be willing to bet it isn’t used in a single, uniform way. If that’s not what Huemer means, it’s likely I have different views about language and meaning. I don’t think “should” or any other term “has” a sense per se, so much as that I think words like “should” derive their meaning within specific linguistic contexts and that their meaning varies as a function of the communicative intent of the people employing those terms; some terms may be highly stable and may tend to be used in very similar ways across contexts and speakers, and there may be words that typically have (roughly) a single general sense, but I don’t know why we should suppose “should” is one of those terms.
This is all moot though, I don’t have any strong objections to Huemer’s point here. I don’t think the ought-implies-can principle is restricted to moral norms.
Conclusion
There’s nothing especially bad about Huemer’s argument. Which is to say there's nothing distinctive about it that makes it an especially poor instance of a philosophical argument, compared to some exemplar of a good argument. The problems I have with arguments like Huemer’s are more fundamental: they’re all trivial. Formalization of an argument can be helpful in organizing one’s thoughts and making it clear why one endorses a particular conclusion. But formal arguments have no power to compel. One can always reject one or more of the premises. And carefully crafting premises to make them appear unobjectionable isn’t the same thing as presenting an unassailable position the rest of us must, on consideration, accept as true. We can always question the meaning or truth of the premises.
The formalization of arguments can serve to clarify what we think and why we think it. Yet philosophers often present formalizations of arguments as a kind of argumentative weapon that has force beyond its organizational form. Formalizations of arguments don’t have any such force. Yet when they’re presented as though they do, they end up functioning as a kind of “cute trick.”
Unfortunately, my objections to Huemer’s argument aren’t distinctive to the argument itself: similar objections could be raised towards many other arguments presented in contemporary analytic philosophy. Much of the discipline seems to center on devising clever formalizations that derive their persuasive force from ambiguity and implication. Wittgenstein noted that “Philosophy is a battle against the bewitchment of our intelligence by means of our language.” I'm not so sure that’s true.
Many philosophers don’t strike me as battling against the bewitchment, but to battle alongside it, actively engaging in the bewitchment. Perhaps it would be more apt to say that philosophy is a battle against our intelligence by means of the bewitchment of our language.
References
Williamson, J. (2022). One philosopher's modus ponens is another's modus tollens: Pantomemes and nisowir. Metaphilosophy, 53(2-3), 284-304.
Note
Thanks to an anonymous friend who offered helpful feedback and contributed to the content of the article.