A Critique of Stefan Molyneux's Ethical Theory of Universally Preferable Behaviour

So in case anyone (or everyone) has forgotten, in the early days of this blog I went through Stefan Molyneux's book, Universally Preferable Behaviour: A Rational Proof of Secular Ethics. My goal was to figure out whether there was really anything to all the fuss that had been made about Stefan and his site, Freedomain Radio, but also to win a bet that I had made with Stefan about the claims that he made about the book. After wrestling with the project for a while, I concluded that Stefan's ideas were troublesome for a rather large number of reasons, and tossed out a rather slipshod overview of some of the things that were wrong with the book. By that time, Stefan had lost interest in continuing to speak with me about the subject, and I was sort of bored with the whole thing anyway, so I was mostly glad to just be done with it.

Recently, however, and for reasons that are not known to me, Stefan's ideas have re-entered the limelight. Perhaps most significant in my world has been Alex "Brainpolice" Strekal's recent involvement in a recorded debate with Stefan that was subsequently plastered all over the internet. But in any case, I thought it might finally be time to actually finish what I started in the form of a systematic critique of the book. So here's that effort.

I. Introduction

Stefan's project is an audacious undertaking. He describes his task as a battle against the "beast" of folk morality -- a view that he defines as "...the superstition that, without the tirades of parents, the bullying of gods or the guns of government, we cannot be both rational and good" (7) -- and contends that "Of course, if I have failed, I have at least failed spectacularly, which itself can be both edifying and entertaining!" (10). In this critique, it will become clear that I believe that Stefan has failed. But I am somewhat inclined to agree that his failure was rather spectacular.

In his book, Stefan boldly attempted to address a wide swath of philosophical disciplines, ranging from metaphysics to the philosophy of science to ethics -- all in less than 150 pages and without a single footnote. So it should surprise no one that he may have left a little bit to be desired in his handling of each subject. But because Stefan's discussion is so wide ranging, and because I don't have the time, energy, or interest to dissect Stefan's metaphysical or scientific views, I will limit my critique to Stefan's handling of ethics. After all, this book is supposed to be a rational proof of secular ethics, and if it fails at that, then the other problems with the book will be irrelevant.

In this analysis, I will rely on a certain interpretation of Stefan's approach in the book which suggests that there are actually at least two separate kinds of arguments that are being used. I proceed this way because there are elements of Stefan's approach that don't necessarily work together and which can be isolated into coherent and separate arguments. I will defend this interpretation with extensive reference to the text of the book, but there will nevertheless remain the possibility that I have misunderstood Stefan's intentions at some point along the way. So keep in mind that while I do think that my accounts of Stefan's arguments are accurate and that my critiques effectively demonstrate what is wrong with Stefan's arguments, it is still true that I am only attacking one interpretation of the text, and that interpretation may be wrong.

This critique will involve three steps. First, I will show that Stefan's terminological choices don't work, and that the first line of argumentation (which is built on this terminology) is non-functional. Second, I will show how Stefan's second line of argument works. And finally, I will explain what's wrong with this second line of argument.

II. Preferable...For What?

The first strain of Stefan's argument is the one that gives the book its title, and deals with the idea that certain behavior can be "universally preferable." The meaning of this term is likely not immediately clear, so it will be necessary to spell out what Stefan means by it.

[Added later: Some people have expressed frustration at the following paragraphs, calling them "nitpicky" as a way of dismissing them as unimportant. But what could be more important to a theory of universally preferable behavior than the definition of universally preferable behavior? In what follows, you will see that I think that the examples which Stefan uses to illustrate his definition contain some errors. These errors may seem small -- indeed, I think they are small -- but nevertheless, they are errors in illustrating the definition of the concept upon which the entire theory is based. I think that I would be doing a disservice to the reader if I simply allowed them to stand without explaining why the examples that Stefan uses are not examples of universally preferable behavior. So I hope that you will bear with me through a tiny bit of nitpicking, since at the other end will hopefully be a clearer understanding of what Stefan's theory looks like than is conveyed by his own examples.]

On page 30, Stefan writes:

When I speak of a universal preference, I am really defining what is objectively required, or necessary, assuming a particular goal. If I want to live, I do not have to like jazz, but I must eat. "Eating" remains a preference - I do not have to eat, in the same way that I have to obey gravity - but "eating" is a universal, objective, and binding requirement for staying alive, since it relies on biological facts that cannot be wished away.

It should be immediately clear that "eating" is not the best example, because you could stay alive by drinking smoothies, injecting yourself with nutrients intravenously, or whatever. But it shouldn't be too difficult to understand what Stefan means; here's another statement from the same page:

If I say that it is preferable for human beings to exercise and eat well...What I am saying is that if you want to be healthy, you should exercise and eat well.

Again this is potentially a bad example, but that's beside the point. What we need to get from this is that the word "preferable" means "required in order to achieve some end." We also need to acknowledge that the claim that something is preferable demands that we identify the end for which the preferable act is required. It isn't enough to simply say "it is preferable to eat"; we need to say, "it is preferable to eat because you want to live." And Stefan acknowledges this, writing:

Naturally, preferential behaviour can only be binding if the goal is desired.

The term "universally preferable behavior," then, could coherently be used in two senses. The first sense, which I will call the "weak" sense, would be to identify something that is required for all people to achieve some end, without claiming that it's actually going to be preferable for everyone. So, for example, you could say that "not killing yourself" is universally preferable for "living," in the sense that anyone who wants to live must not kill herself. But you could allow the possibility that some people wouldn't want to live, and so it wouldn't be preferable for them to not kill themselves. The second sense, which I will call the "strong" sense, would be to identify something that is required for all people to achieve some end that they all actually want to achieve. So if it were true that everyone on Earth actually did want to live, you could say that it is universally preferable that they all not kill themselves, because that would be incompatible with their ends.

So let's see this in action. On page 34, Stefan says:

...if a man wants to cure an infection, he should take antibiotics rather than perform an Aztec rain dance. The preference for taking antibiotics rather than doing a rain dance is universal, since dancing cannot cure infections. Thus, although there is the occasional madman who will try to cure himself through dancing, it is still universally preferable that if a man wants to cure himself, he must take antibiotics.

Here (again in spite of the bad example -- there might be other ways to cure an infection) we can directly see the weak sense of the term being used. It's not necessarily true that everyone actually wants to cure their infections, but if they did, then Stefan is identifying what would be required for them to achieve their ends: taking antibiotics. I hope that this will be sufficient for establishing that this is how Stefan is using the term; if not, I can find some other examples.

But now it's time for the first bit of that spectacular and entertaining failure that I promised at the beginning. On page 33, Stefan writes:

When I say that some preferences may be objective, I do not mean that all people follow these preferences at all times. If I were to argue that breathing is an objective preference, I could be easily countered by the example of those who commit suicide by hanging themselves.

What do we make of this? Let's try running it through our two senses of the term "universally preferable." According to the strong sense, arguing that breathing is objectively preferable would mean that breathing is required for some end that everyone shares, and since he gives suicide as the example, we might presume that the end in question is something like "living without the aid of life-support machinery." But I would imagine that someone who commits suicide probably does not want to live (with or without life-support machinery), and so it simply wouldn't be true that breathing is universally preferable in the strong sense.

In the weak sense of the term, we can say that breathing is universally preferable for living without life-support, since anyone wanting to do that would need to breathe. And since we're only using the weak sense of the term, we acknowledge that if the suicidal person doesn't want to live, it won't be necessary for her to breathe: it's only preferable to breathe in order to achieve the end in question, and she doesn't want to achieve it. So this makes sense.

But in the next paragraph, Stefan writes:

Thus when I talk about universal preferences, I am talking about what people should prefer, not what they always do prefer.

Now if we were using the strong sense of the term "universally preferable," this would make sense: because everyone shares the end in question, we can say that everyone should do what is universally preferable. But in this example, we're forced to use the weak sense, since it's not the case that everyone shares the end in question. Accordingly, it simply isn't true that by this line of thinking, we would arrive at the conclusion that the suicidal person should breathe. In fact, the suicidal person's goal is to suffocate: for people with that end, it is actually universally preferable (in the weak sense) not to breathe.

Okay; back to business. Now that we have an understanding of what Stefan means by "universally preferable," we can turn to his theory of ethics. First, we need to get a bit of confusion out of the way. On page 30, Stefan writes:

Ethics as a discipline can be defined as any theory regarding preferable human behaviour that is universal, objective, consistent - and binding.

But later in the book, Stefan uses a much more restricted definition of the discipline of ethics. On page 48, for example, he writes:

In general, we will use the term aesthetics to refer to non-enforceable preferences - universal or personal - while ethics or morality will refer to enforceable preferences.

And he continues:

Ethics is the subset of UPB which deals with inflicted behaviour, or the use of violence.

Throughout the book, Stefan uses this second, more restricted sense, and not the first sense, to talk about morality. For example, on page 76 he writes, "...morality is defined as an enforceable subset of UPB...," and in Appendix A on page 125, he writes, "The subset of UPB that examines enforceable behaviour is called "morality." He explains this shift on page 48 by explaining that "Although we first focused on UPB in the realm of ethics, UPB can now be seen as an "umbrella term"..." Fair enough; it just seems like it should be explicitly acknowledged here.

So, then, how would we use the concept of universally preferable behavior to think about "inflicted" or violent behaviors? To use the strong sense of "universally preferable behavior," we'd need to think about a kind of situation where all people need to use violence in order to achieve their ends. But given the vast diversity of ends, it would seem sort of odd to suggest that everyone is going to be required to do a particular sort of violent action in order to achieve a particular end that they all share. I mean, maybe there's a way to make that work, but I'd be sort of surprised.

What about the weak sense of the term? It seems rather clear that there are at least some ends that would require the use of violence, no matter who was trying to achieve them, and in these cases we'd want to say that it was universally preferable to use violence in order to achieve those ends.

But notice that so far, we're not talking about morality at all. We're pretty much talking about prudence -- that is, what you should do in order to achieve your goals. And this makes sense, because as Stefan said, "Naturally, preferential behaviour can only be binding if the goal is desired."

So, then, it's interesting to see where Stefan goes next. He writes:

Any theory that justifies or denies the use of violence is a moral theory, and is subject to the requirements of logical consistency and empirical evidence.

But what does he mean by "justifies" and "denies"? The idea of "justification" usually has to do with the idea that someone ought to be able to provide certain defenses of the choices he or she makes. But if we're just talking about ways of achieving our goals, the only justification we would need would be to a) identify what our goals are, and b) show how our chosen behaviors are consistent with achieving our goals.

It's this feature of the meaning of "universally preferable" that proves to be the fatal flaw in this line of argumentation. As long as we understand "universal preferability" in the terms that Stefan outlined in the beginning of the book, we will never be able to use the term to establish anything beyond the fact that something is necessary in order to achieve certain ends. So, then, it cannot be on the basis of universal preferability that we condemn the rapist, the murderer, or the thief -- that is, unless we somehow want to argue that rape, murder, or theft are somehow incompatible with the desires of the rapist, the murderer, or the thief, and our condemnation is simply one of taking exception to their stupid and ignorant choices. But Stefan doesn't argue this way, and that's good, because if he did it would be disastrous.

To illustrate this point, let's run the example of rape through our two senses of the term "universally preferable." The strong sense of universal preferability relies on the idea that everyone shares a certain end, and that there is something that is universally required in order to achieve this end. But it should be uncontroversial that neither raping nor non-raping is necessary for any end that everyone shares (after all, there are rapists, and presumably it will help them achieve their goals if they rape). So it simply would not be true that in the strong sense of "universally preferable," it is universally preferable for people to either rape or not rape.

And in the weak sense of "universally preferable," we can surely imagine that there are some ends that would require rape to achieve. For example, we might imagine that someone wants to achieve the end of "Having sex with someone who doesn't want to have sex with them, and who won't be amenable to being convinced or bargained with." In order to achieve this end, it would seem that rape would be required, no matter who was trying to achieve the end. So in the weak sense, rape would be universally preferable for anyone trying to achieve this end. And the same sort of thinking would apply in the cases of murder and theft.

So the concept of "universally preferable" behavior is simply not going to get Stefan to where he wants to go. He's just defined it in a way that makes it incompatible with the kinds of things he wants to establish -- among them that rape, murder, and theft are (thankfully) not moral. Accordingly, we'll turn to the next kind of argument that Stefan offers in his book.

III. A Maxim-Based Approach

Stefan's second line of argument is also built on the vocabulary of "universally preferable behavior," but seems to use the term in a way that's very different from the way that it is defined at the beginning of the book. In this other sense, "universally preferable" no longer means "required in order to achieve certain ends," but instead takes on a role in what I will call a "maxim-based" approach to ethical reasoning.

Because Stefan's words are tied up in the flawed terminology of "universally preferable behaviors," it will be necessary to find another way to interpret what Stefan is talking about. As we saw in the previous section, Stefan tells us that moral theories identify inflicted behaviors that are either justified or unjustified. We saw that when we think about this in terms of a behavior's being "universally preferable," this doesn't make sense. But it seems to me that we can proceed if we just cut out the "universally preferable" terminology altogether and just use terms like "morally required" or "morally unjustified" in their stead.

So for example, when Stefan is talking about the Non-Aggression Principle (NAP) on page 54, he says:

The NAP is basically the proposition that "the initiation of the use of force is morally wrong." Or, to put it more in the terms of our conversation: "The non-initiation of force is universally preferable."

As we saw in the last section, calling the non-initiation of force "universally preferable" doesn't actually mean what Stefan seems to want it to mean. But if we just ignore the term "universally preferable" and substitute "morally required" in its place, then the statement is basically a tautology (which is what we want). It would now say:

The NAP is basically the proposition that "the initiation of the use of force is morally wrong." Or to put it more in the terms of our conversation: "The non-initiation of force is [morally required]."

Now that we have figured out a way to interpret Stefan's ideas without being hindered by his terminology, we can turn to the question of how he's using the term in order to build a moral framework. I said that the way he goes about doing this is through a "maxim-based" approach. So what is a maxim-based approach, and how does it work? The fundamental building block of such an approach is -- unsurprisingly -- a "maxim," which is a principle on which a person bases her decisions. The form of a maxim is as follows:

"In circumstances C, I will do X."

The role of C is to tell the actor when the rule is supposed to apply by describing the sorts of situations in which the action is to take place, and the role of the X is to describe the action that this person is supposed to be doing. So for example, we might offer the maxim, "When I am outside in the rain, I will cover myself with an umbrella." The C in this case would be any set of circumstances in which I am outside in the rain, and the X would be the action of covering myself with an umbrella. Clearly, when it is not raining or when I am inside, the maxim will have nothing to say; it only applies when the situation I face matches the circumstances set out in C. But we can say that I am always following this maxim if whenever I'm outside in the rain, I cover myself with an umbrella.

Stefan's approach to moral theorizing looks at maxims in order to determine whether they can be morally legitimate as guides for action. We can see the maxim-based approach most clearly on page 44, when Stefan is considering whether a soldier could ever be justified in killing. After first dismissing the possibility that murder is always morally required (a bold stance!), he introduces another possibility by saying:

If, to save the virtue of soldiers, I alter my theory to argue that it is moral for people to murder if someone else tells them to...

We can translate this proposed maxim as saying, "In circumstances where I am told to murder, I will do so." Stefan goes on to dismiss this maxim, and for reasons I will discuss in the next paragraph. But for right now, the thing to understand is how Stefan's approach can be understood as being based on evaluating maxims. The question of whether the soldier's killing is morally justified is answered by first figuring out what maxim the soldier would be following in deciding to kill, and then considering whether or not that maxim is an acceptable guide for the soldier's action.

So how does Stefan determine whether or not a maxim is acceptable? The first step that he takes is to suggest that we ought to consider what would happen if everyone followed the maxim in question. Stefan uses this approach to dismiss the maxim we just discussed:

If Politician A can order a soldier to murder an Iraqi, then the Iraqi must also be able to order the soldier to murder Politician A, and the soldier can also order Politician A to murder the Iraqi. The application of this theory results in a general and amoral paralysis, and is thus proven invalid.

If we generalize this idea, we can arrive at the following rule, which I think is foundational to Stefan's argument:

Claim 1: If it would be impossible or unreasonable for everyone to follow a particular maxim all the time, then that maxim is not morally legitimate as a guide for action.

How does this fit in? It seems clear that the problem that Stefan is identifying with the maxim in question -- "In circumstances where I am told to murder, I will do so" -- is that it would be impossible or unreasonable for everyone to follow it as a "rule to live by." It is on this basis that Stefan is claiming that it is defective as a guide for action.

But why should we accept Claim 1? Stefan offers the following explanation on page 43:

...the first test of any scientific theory is universality. Just as a theory of physics must apply to all matter, a moral theory that claims to describe the preferable actions of mankind must apply to all mankind. No moral theory can be valid if it argues that a certain action is right in Syria, but wrong in San Francisco. It cannot say that Person A must do X, but that Person B must never do X. It cannot say that what was wrong yesterday is right today - or vice versa. If it does, it is false and must be refined or discarded.

When I first read this, I was a little confused, since this isn't obviously how all theories in physics actually work. That is, the theory of gravity may apply to all matter, but other theories (e.g., theories about black holes) clearly don't apply to all matter (though they presumably apply to all black holes). What physical theories do follow is something like a principle that things that are identical in all relevant properties must be treated in the same way by a theory -- so a theory of matter would apply to anything that is matter, and a theory of black holes would apply to anything that is a black hole.

But are people identical in the senses that would be relevant for ethical theorizing? I think that's a sort of complicated question, and an overly simplistic attempt to answer it sets up for the next episode of "Stefan's spectacular and entertaining failure" (this time from page 44):

Since all human beings share common physical properties and requirements, proposing one rule for one person and the opposite rule for another is invalid - it is like proposing a physics theory that says that some rocks fall down, while others fall up. Not only is it illogical, it contradicts an observable fact of reality, which is that human beings as a species share common characteristics, and so cannot be subjected to opposing rules.

Really, Stefan? So do you pee sitting down, or does your wife pee standing up? Or do you both pee in your diapers (holler back, baby Molyneux)? *Rimshot!* But I don't actually want to object to Claim 1, so we can move on while leaving it intact. I personally think it's sort of reasonable (though perhaps not as solid as Stefan seems to think), and there are much bigger fish to fry. [Added later: For those interested in thinking more about the moral equality of humans, a classic piece is Bernard Williams' essay, "The Idea of Equality." Taking the discussion a step further to ask how humans can be morally equal and morally unequal from animals is Bonnie Steinbock's excellent piece, "Speciesism and the Idea of Equality."]

The next step in Stefan's ethical theory is a bit more troubling, but ultimately can be repaired to make sense. On page 66, Stefan writes:

The opposite of "virtue" must be "vice" - the opposite of "good" must be "evil." If I propose the moral rule "thou shalt not steal," then stealing must be evil, and not stealing must be good.

It should immediately be recognized that "not stealing" is not the opposite of stealing, but is rather the negation of stealing. To illustrate this point: Most people would say that the opposite of "black" is "white." But "not black" is not the same thing as "white." "Green" is "not black," but it isn't "white."

Unfortunately, if we try to repair Stefan's point by changing the word "opposite" to "negation," then the point becomes false. The negation of "virtue" isn't "vice," and nor is the negation of "good" "evil." A person can lack virtue without displaying vice, and a person can be not good without being evil.

But Stefan does not recognize this problem, and so adopts something like the following:

Claim 2: If adopting a particular maxim would be virtuous or morally good, then failing to adopt that maxim would be evil. Likewise, if adopting a particular maxim would be evil, then failing to adopt that maxim would be virtuous or morally good.

This is embodied rather clearly on page 66 when Stefan writes:

If "rape" is a moral good, then "not raping" must be a moral evil - thus it is impossible for two men in a single room to both be moral at the same time, since only one of them can be a rapist at any given moment - and he can only be a rapist if the other man becomes his victim.

It should be uncontroversial that rape is not actually a moral good, but for our purposes this is beside the point. I use this example only to demonstrate how Claim 2 is being used in Stefan's theory. Presumably, the maxim that Stefan is discussing here is something along the lines of "In all circumstances, I will rape," and the question appears to be what would follow if we said that adopting the maxim is morally good. Applying Claim 2, Stefan arrives at the conclusion that failing to adopt the maxim would be morally evil.

But where Stefan has introduced Claim 2 as if it were simply a logical truth, the disconnection between the concept of an "opposite" and a "negation" eliminates this defense of the claim. Accordingly, Claim 2 is a rule in search of a justification, and as far as I can tell, Stefan doesn't offer another argument for it.

So what do we think of Claim 2? Personally, I think it depends on what we mean by "morally good" or "virtuous" When we normally talk about morally good or virtuous actions, we mean things that speak highly to a person's character, but which it would not necessarily be evil for me not to do. Charity is a great example of something which most people find to be morally good, and helping my sister with her homework is seemingly virtuous. But I wouldn't obviously be evil if I didn't do those things. If this is how we mean "morally good," then Claim 2 is clearly going to be false. We can only save Claim 2 if by "morally good," we mean something like "morally required." Rephrased this way, we get:

Claim 2': If adopting a particular maxim would be morally required, then failing to adopt that maxim would be evil. Likewise, if adopting a particular maxim would be evil, then failing to adopt that maxim would be morally required.

And Claim 2' seems obviously true. If we interpret Stefan's example with this altered terminology, we can see that things make a little more sense: If rape were morally required, then not raping would be evil, and Stefan's concerns would seem to apply.

So to summarize, we thus far have two more-or-less plausible claims underpinning Stefan's maxim-based approach:

Claim 1: If it would be impossible or unreasonable for everyone to follow a particular maxim all the time, then that maxim is not morally legitimate as a guide for action.
Claim 2': If adopting a particular maxim would be morally required, then failing to adopt that maxim would be evil. Likewise, if adopting a particular maxim would be evil, then failing to adopt that maxim would be morally required.

The third step in Stefan's theory is even more troubling, and will ultimately lead to the objections that we will see in the next section. But let's not get ahead of ourselves. Recall that on page 48, Stefan defined "ethics" as "the subset of UPB which deals with inflicted behaviour, or the use of violence," and said that "Any theory that justifies or denies the use of violence is a moral theory, and is subject to the requirements of logical consistency and empirical evidence."

This realm of "ethics" is distinguished by the presence of violence, and is contrasted by Stefan against the realm of aesthetics, which involves behaviors that are not "inflicted" on others (for our purposes, we can ignore the potential difficulties of distinguishing between behaviors that are "inflicted" and those that are not). The distinction is illustrated on page 50:

If I accept your invitation to a dinner party, but find the conversation highly offensive, I can decide to get up and leave - and I can also choose to never accept another invitation from you. This capacity for escape and/or avoidance is an essential characteristic differentiating aesthetics from ethics.

If, however, when I decide to leave your dinner party, you leap up and chain me to my chair, clearly I no longer have the free choice to leave. This is the moment at which your rudeness becomes overt aggression, and crosses the line from aesthetics to ethics.

We see, then, that when we're trying to determine whether something is a moral question or not, we need only establish that there is violence involved, or that something is being inflicted on someone else.

So then let's jump forward to page 69, where Stefan is trying to determine whether we can think of rape in terms of personal preference. He writes:

If we propose the moral rule: "personal preferences must be violently inflicted upon other people," how does that stand up to the framework of UPB? (Note that I cannot propose that personal preferences may be violently inflicted upon other people, since that is a violation of UPB, which states that moral rules must be absolute and universal - if they are not, they fall into APA [aesthetically positive action] territory, and so cannot be inflicted on others.)

There's a lot wrong with this statement, particularly since Stefan seems to be confused about his own distinction between ethics and aesthetics -- he now seems to think that it would be wrong to impose aesthetic matters on people, but any imposed behavior has already been defined as falling within the realm of ethics, and not aesthetics.

However, what I want to focus on for the moment is another idea captured in this statement: that moral rules can't be optional. What Stefan seems to be saying is that we must accept either that "personal preferences must be violently inflicted upon other people" or that "personal preferences must not be violently inflicted upon other people." Otherwise, we would need to say that "personal preferences may be violently inflicted upon other people, or they may not be," and Stefan has ruled out this possibility. To more explicitly phrase this issue in terms of the maxim-based language, Stefan is posing a choice between "In all circumstances, I will violently inflict my personal preferences upon other people" and "In all circumstances, I will not violently inflict my personal preferences upon other people."

If this is how Stefan wants to structure his theory, then we get a third claim:

Claim 3: For any maxim that involves an inflicted behavior, it is either the case that everyone is morally required to follow that maxim all the time, or that everyone is morally prohibited from ever following that maxim.

And we can see this claim at work, for example, on pages 79-81 where Stefan uses an illustration involving the theft of a lighter in order to argue that stealing must be universally prohibited, since it would be ridiculous to think that everyone could possibly be morally required to steal all the time.

At this point I think we have a pretty clear understanding of the substance of Stefan's second line of argument. The argument is based on a maxim-based approach, and involves at least three central claims:

Claim 1: If it would be impossible or unreasonable for everyone to follow a particular maxim all the time, then that maxim is not morally legitimate as a guide for action.
Claim 2': If adopting a particular maxim would be morally required, then failing to adopt that maxim would be evil. Likewise, if adopting a particular maxim would be evil, then failing to adopt that maxim would be morally required.
Claim 3: For any maxim that involves an inflicted behavior, it is either the case that everyone is morally required to follow that maxim all the time, or that everyone is morally prohibited from ever following that maxim.

Now there's a fourth claim that is implied in Stefan's argument that I want to eliminate right away for being too dumb to even consider. The claim is this:

Claim 4: For any maxim that involves an inflicted behavior, the set of circumstances to which the maxim applies must be "where the action in question will be inflicted upon a human being."

Stefan introduces this idea on page 73 when he considers the maxim "I can shoot a man anytime I want" (or to use the more precise form we've been using, "In circumstances where my shooting will be inflicted upon a person who is asleep, and I want to engage in the shooting, I will shoot the person"):

The problem here is not only the sleep that Bob will lose based on his universal premise, but also the logical impossibility of reversing moral propositions based on the differences in the states of sleeping and waking. Biologically speaking, a man does not become the opposite of a man when he falls asleep, any more than gravity reverses when he blinks.

Since a man remains a man when he falls asleep, it cannot be the case that opposite moral rules apply to him in this state. Thus to say that it is immoral to murder a man when he is awake, but it is moral to murder a him [sic] when he is asleep, is to create a logical contradiction unsupported by any objective biological facts. A physicist may say that a rock falls down, but a helium balloon rises up - but that is because a rock and a helium balloon have fundamentally different properties. No credible physicist can say that one rock falls down, but that another rock with almost exactly the same qualities falls up. The same is for moral theories - no credible philosopher can say that morality reverses itself when a man is asleep, since a man's nature does not fundamentally alter when he naps.

The number of things wrong with this argument is utterly astonishing; I won't even begin to address them all, since to do so would require an analysis all its own. But what's important to pay attention to is the last sentence, where Stefan says that we can't draw moral distinctions that make reference to the fact that a man is asleep because a man's nature does not fundamentally alter when he naps. To see why this is dumb, consider the following: In the game of hockey, a goal is scored when a puck goes into the net during a legal play. A goal is not scored when a puck does not go into the net under such circumstances. What Stefan is suggesting basically amounts to saying that we logically can't differentiate a goal from a non-goal because a puck's nature does not reverse itself when it goes into the net. This is dumb, and I don't even want to try to think about what kind of nonsensical ethical theory we would get from accepting it.

[Added later: It occurs to me that the hockey example is a rhetorically weak one, regardless of whether or not it captures the conceptual point well. To see why this sort of claim is a problem for Stefan's theory, consider the case of self-defense. When we are put in a position where we must defend ourselves, most of us think that we are justified in taking measures against our attackers which would be unjustified under other sorts of circumstances. So, for example, it would be okay for me to punch Albert in the face if he were assaulting me, even though it would normally not be okay to punch Albert in the face. However, notice that Albert's nature as a man does not fundamentally change when he attacks me. Thus if we accept what Stefan says here, then I should not be able to justify treating Albert differently when he is assaulting me than when he is leaving me alone. This seems silly to me. Of course it would not take a fundamental change in Albert's nature as a man to justify treating him in ways that are normally unacceptable. We could justify different treatment for far less.]

Accordingly, I will do my best go on as if Stefan never said this. It will come up again a little later on, but I'll do my best to minimize the importance of this error. So setting aside Claim 4, we can now proceed to the critique.

IV. What's Wrong With This Structure?

I want to argue two things about the structure I discussed in the previous section (ignoring Claim 4). First, I will show that even if we accept the maxim-based approach and the three claims, we will not be able to generate satisfying conclusions for moral problems. And second, I will contend that it leaves out an important component of ethical thinking that makes the resulting structure very awkward to work with.

My argument that Stefan's theory will not generate satisfying conclusions focuses mainly on Claim 3. The first thing to notice about the Claim is that Stefan applies it in such broad strokes that it seems to be rather implausible; the example on pages 79-81 (where Stefan attempts to show that theft must always be evil because it couldn't possibly be morally obligatory to steal all the time) illustrates this point well.

In order to see why this is an implausible way to apply the Claim, we must remember that Stefan's approach is based on evaluating maxims, and not actions directly. Claim 1 was "If it would be impossible or unreasonable for everyone to follow a particular maxim all the time, then that maxim is not morally legitimate as a guide for action." Accordingly, the fact that "In all circumstances, I will steal" is unacceptable as a maxim does not demonstrate that stealing is wrong, as Stefan seems to think it does. It only demonstrates that the moral rule, "In all circumstances, I will steal," is unacceptable.

In order to use Claim 3 coherently, we would need to focus on the C term in the maxim-based framework (remember, the form of a maxim is "In circumstances C, I will do X"). Stefan's discussion focuses on maxims where the C term covers all possible circumstances, so that the maxim is prescribing a particular action all the time. But if we changed the C term to say, for example, "When I will die if I don't steal an object, and I can be reasonably certain that the victim will not miss what I took very much, and I want to live" we would seemingly be able to avoid the problems that Stefan presents. Phrased this way, the maxim would be, "In circumstances where I will die if I don't steal an object, and I can be reasonably certain that the victim will not miss what I took very much, and I want to live, I will steal the object." And it seems rather clear to me that everyone could follow this rule all the time without any horrible problems arising (even though it would only prescribe that the follower take a particular action in a very limited set of circumstances). Accordingly, this maxim would be able to get past Claim 1 intact.

Stefan seems to think that this is not acceptable because of Claim 4. But remember, we're not going to pay any attention to Claim 4, because it's dumb.

It is important, though, to see how this works with Claim 1, which we are paying attention to. Remember, Claim 1 was that "If it would be impossible or unreasonable for everyone to follow a particular maxim all the time, then that maxim is not morally legitimate as a guide for action." This claim was not built on the idea that maxims must prescribe actions that must be carried out by all people at all times; it was built on the idea that it must be possible for all people to follow the maxims at all times. And in fact, he actually provided an example in the same place where he introduced the universality requirement with such a limitation: "In circumstances where I am told to murder, I will do so." Clearly, this maxim would not instruct someone to murder in circumstances where no one told them to murder, but Stefan doesn't raise any objections to this feature of the maxim in his discussion. Remember, we got Claim 1 from his objection that it would be ridiculous for everyone to follow this maxim all the time, not from any claim that it would be ridiculous for everyone to murder all the time -- that was an objection raised to a different maxim.

What implications does this have for the theory's capacity to generate satisfying conclusions? Remember, Claim 3 says, "For any maxim that involves an inflicted behavior, it is either the case that everyone is morally required to follow that maxim all the time, or that everyone is morally prohibited from ever following that maxim." So Stefan's approach rules out maxims, and not actually the action prescribed by the maxim (remember, the X component). Accordingly, we can see that what Stefan's system requires us to do is to figure out ways to justify our inflicted behaviors so that the maxims we follow could be morally obligatory for all people to follow all the time (this is not true if we accept Claim 4, which essentially rules out adjustment for circumstances, but remember, Claim 4 is dumb).

This would still mean that "In all circumstances, I will steal," would be problematic (because of Claim 1), but we might be able to make something out of "In circumstances where I will die if I don't steal an object, and I can be reasonably certain that the victim will not miss what I took very much, and I want to live, I will steal the object." It will at least be true that if we want to object to this latter maxim, we won't be able to effectively do so on the basis of the three Claims that I offered without begging the question. Again, they were:

Claim 1: If it would be impossible or unreasonable for everyone to follow a particular maxim all the time, then that maxim is not morally legitimate as a guide for action.
Claim 2': If adopting a particular maxim would be morally required, then failing to adopt that maxim would be evil. Likewise, if adopting a particular maxim would be evil, then failing to adopt that maxim would be morally required.
Claim 3: For any maxim that involves an inflicted behavior, it is either the case that everyone is morally required to follow that maxim all the time, or that everyone is morally prohibited from ever following that maxim.

Subjecting our maxim to Claim 1, it doesn't seem like we would be able to mount any conclusive attack on the basis of it being impossible or unreasonable for everyone to follow it all the time. Claim 2' wouldn't be a problem for our maxim, since it's just sort of a formal requirement for how we think of morally required behaviors and evil behaviors. And Claim 3 would just say that either people are morally required to follow the maxim all the time (since it involves an inflicted behavior), or they're prohibited from ever following it. But there's nothing about the maxim itself that makes it clearly true that either one of these would be true and the other false; to rule in favor of one side over the other would simply beg the question. This, I think, is very unsatisfying.

The second thing I wanted to establish about this system is that it leaves out an important component of ethical thinking, and that this makes the whole theory a bit awkward. The component I have in mind is the idea of a behavior that is morally permissible, but not morally required. In most moral theories, there are three kinds of actions (even ones that involve violence): things you must do, things you may do if you want, and things that you may not do. Stefan rules out the second category with Claim 3, which claims that moral behaviors must either be obligatory or prohibited. But this would seem to mean that you could never have a choice about whether to act violently where you wouldn't be evil for choosing one of the alternatives. Surely this would be implausible.

In order to save the theory from being dismissed out of hand, the personal choice has to be snuck back in through the C component of the maxim. So for our maxim, "In circumstances where I will die if I don't steal an object, and I can be reasonably certain that the victim will not miss what I took very much, and I want to live, I will steal the object," we should immediately notice that "and I want to live" is a critical part of the idea. Without it, we would need to be saying that it would be evil for the person in question to choose to simply die rather than steal. And surely someone who thinks that stealing in such situations would be okay would not want to imply that not stealing in that situation would be evil if the person didn't want to do it. This, I think, makes the theory very awkward to work with. You end up saying that people must do things in circumstances where they want to do them (and, presumably, that they must not do things in circumstances where they don't).

V. Conclusion

In this critique, I have attempted to establish three things. First, I have argued that Stefan's choice of the term "universally preferable behavior" simply doesn't work with the rest of his theory; the way he defines the term makes it completely incompatible with the sorts of things he wants to argue for. Second, I have attempted to build a coherent way of evaluating moral propositions based on a maxim-based approach to ethical reasoning that involves three central claims about maxims. And third, I have tried to show that (without the ridiculous and implausible Claim 4) the resulting theory is incapable of generating satisfying answers to moral questions, and is awkward to work with. The system is unsatisfying because it does not offer us a way to determine definitively whether certain actions are morally required or not as long as the maxims being followed are sufficiently specific. And it is awkward because it requires us to twist ourselves into weird shapes in order to accommodate the idea of morally permissible behaviors.

If these arguments hold, it would mean that Stefan has not only failed to "slay the beast" and deliver us with the salvation that he promises in the beginning of the book, but that he has also failed to even provide us with a coherent ethical theory. This is an important "if," I will admit: the amount of conceptual reconstruction that I've had to do in order to put together this critique almost ensures that the resulting structure is about as much Stefan's as it is my own. And if Stefan rejects that his argument is based on maxims, or that the three claims I've offered are actually indicative of his arguments, then I may have attacked a straw man.

It should be noted that I justified each step of my interpretation with direct evidence from Stefan's book, and so in order to reject my interpretation, Stefan would need to explain how my interpretation doesn't actually capture what's being said in the passages I cited. But I don't dismiss the possibility that this can be done.

With this, I think we can finally consider The Molyneux Project to be a success, though it will not be truly complete until all objectors are satisfied. I hope that this has been instructive for any of you who have been keeping up, and that those who are interested in Stefan's ideas will take seriously the need to respond to what's been said here.

-----------------------

P.S. Yes, I know that my critique can be answered by insisting that Claim 4 is true. But Claim 4 is so obviously ridiculous that this would be a pretty troubling defense to offer. Advance it at your own risk.

P.P.S. I've added a few clarifications. The first is at the beginning of section II where I discuss the definition of universally preferable behavior. Some people had focused their responses to this critique on that discussion, and so I added a note giving it a little bit more context. Hopefully that will address the criticisms. The second clarification is in my discussion of Claim 1 in section III, pointing to some further reading on the issues raised there. I added the note because I sort of made light of Stefan's discussion of Claim 1, but then just moved on without really offering any other explanation for why it makes sense. While I don't think that hurt the critique, I do think it was a bit unhelpful for people who might have been reading through it and who might have wanted more. The third clarification was in my discussion of Claim 4 in section III. I think that it adds an important element that was absent before, so hopefully others find it a worthwhile addition.