I – II – III – IV – V – VI – VII – VIII – IX – X – XI – XII – XIII – XIV – XV – XVI – XVII – XVIII – XIX – XX
[Single-page view]

So all right, we’ve established that the function of consequentialist/utilitarian morality is, as Alexander summarizes it, to “try to make the world a better place according to the values and wishes of the people in it” – and we’ve established that the way to tell whether an act meets that standard is to evaluate it in the broadest possible terms, making sure to account for all the various side effects and long-term implications.

The next issue we have to deal with, then, is the fact that trying to “make the world a better place according to the values and wishes of the people in it” depends on what those people’s values and wishes actually are. If those people happen to have values and wishes that are irrational or destructive or otherwise terrible (as a lot of people certainly do), then isn’t that a problem for our system? What if people hold desires that are sadistic or bigoted – or even more problematic, what if people hold desires that don’t even maximize their own utility? How do we deal with that?

Well, the first part of the question – what if people hold desires that are sadistic or bigoted – is one that we’ve somewhat addressed already. It’s true that the simplest version of utilitarianism counts the satisfaction of these harmful desires just as positively as the satisfaction of more benign ones. If someone feels revulsion whenever they encounter an interracial couple, for instance, then that person’s discomfort is in fact real, and does count as a reduction in their utility. And if they could get a law passed to ban interracial marriage, then the alleviation of their discomfort would in fact register as an increase in their utility (just as the axe murderer from before would get a utility boost from murdering his victim). But of course, acknowledging that one person might derive some positive utility from a particular act doesn’t mean that the act is an overall positive in the broadest sense. As much utility as our hypothetical racist might derive from banning interracial marriage, it would be far more of a decrease in utility for all the interracial couples who would no longer be able to get married – so in global terms, it would still be a major utility reduction. In other words, you wouldn’t be able to call it morally good; at most, all you could say is that it would be good for the racist who no longer had to experience as much discomfort.

That being said, though, we can make this example more difficult. Let’s imagine, for instance, that we aren’t just talking about just one racist individual, but a whole society. Say there’s an entire country that forbids interracial marriage, because its populace mistakenly fears that legalizing it would destroy their society. What’s more, let’s imagine that there’s no one within this country who actually wants to marry outside their race – so no one’s utility is being reduced by the ban (at least not directly). So far, then, it seems like maintaining the ban would meet our definition of being morally good, right? Everyone’s preferences are being met and no one’s utility is being reduced. But is it really right to say that encouraging racism would be good, even in a society where everyone approves of it? After all, as much benefit as they might derive from banning interracial marriage, maybe they’d be even better off if they learned not to feel revulsion towards people of different races in the first place. Their conscious, explicit preference might still be to maintain their racism, but wouldn’t they actually gain more utility from not having such toxic feelings at all? Wouldn’t the real utilitarian thing to do here be to go against the action that the people have ascribed the most subjective value to?

I think this is a valid question. So let’s imagine that one day, the leader of our hypothetical country (who let’s assume has been legitimately democratically elected and entrusted with making these kinds of judgments for the good of the populace) decides to go against the ostensible will of the people and launch a campaign pushing for the legalization of interracial marriage – which ultimately succeeds, and to everyone’s surprise (except the leader’s), doesn’t end up destroying their society at all, but actually improves it significantly. Lonely single people are suddenly able to find spouses, extended families are enriched by the opportunity to discover new perspectives from their new in-laws, and so on. The thing that everyone had regarded as bad for them actually ended up providing a significant increase in their utility. In other words, they were wrong about what was best for them.

(If you want to consider a different example of this, you might instead imagine a misguided populace wanting to start a disastrous war with a neighboring country or something (in a way that would be totally self-destructive), with their leaders being the only ones who realize this and want to avoid it.)

How, then, does that fit into our system? This is the second part of our question from before – what if people hold desires that don’t actually maximize their own utility? Utilitarianism is supposed to be all about satisfying people’s preferences, right? So what if their preferences are flawed, and the amount of goodness they ascribe to a potential outcome doesn’t actually correspond with the amount of goodness that that outcome would produce for them? On what basis could our hypothetical political leader justify legalizing interracial marriage as a moral act if nobody actually considered it a good thing? Sure, we could say that it was a good thing after the fact, once everybody came around and started ascribing a positive value to it rather than a negative one – but what about before then? How could we say that it would be good to take a certain action if doing so would go against everyone’s subjective valuations? Is goodness defined by those subjective valuations, or isn’t it?

This isn’t just a one-off problem, either, nor one exclusive to national governance; it can apply to all kinds of different situations. If you see someone about to cross the street, oblivious to the bus that’s heading straight for them, would it be good to stop them from crossing even though their explicit preference is to cross? If you see someone about to drink a beverage, oblivious to the fact that it’s actually poison, would it be good to stop them from drinking it even though they’ve ascribed more goodness to drinking it than to remaining thirsty? If goodness is nothing but a subjective valuation that people ascribe to things, then how can going against those subjective valuations be considered good for them?

As you might have already figured out, this question isn’t actually as problematic as it seems. Yes, it’s true that goodness is defined by people’s subjective valuations; an act can’t be good unless someone judges it to be good. And it’s true that people’s judgments of what’s good – their conscious preferences – don’t always match up with what they’d actually prefer if they knew better. But the thing is, the explicit object-level preferences that people have toward their immediate situations aren’t the only preferences that people hold. They also have meta-preferences – i.e. preferences about their preferences – and in situations like the ones above, those higher-order meta-preferences can supersede their object-level preferences. So for instance, if you were about to cross the street, your object-level preference might be to cross freely without anyone stopping you; but at the same time, you’d have a meta-preference that if you were wrong in your assumption that crossing freely would be the most utility-maximizing outcome for you (like if there were an oncoming bus about to run you over), you’d actually prefer for someone to step in and stop you from crossing. Similarly, if you were about to drink a beverage, you might have an object-level preference to drink it, but you’d also have a higher-order preference that if your object-level preference would actually be harmful in some way you didn’t anticipate (like if the beverage were poisoned), then someone should intervene and stop you. In other words, whatever your object-level preference is toward the specific situation you’re in, your meta-preference will always be that if there’s alternative outcome that would maximize your utility more than your object-level preference would, then your misguided object-level preference should be disregarded; and if there’s someone around who has a better understanding of the situation and is in a position where they can overrule your misguided preference, then your meta-preference will be to defer to them and let them act on your behalf. The outcome of having your misguided preferences overruled, even if it frustrates you at the object level, is the one that you actually ascribe the greatest value to overall. So in this way, your meta-preference is a kind of Master Preference – a preference over all other preferences – which simply says that no matter what, you will always prefer outcomes that maximize your utility, even if you aren’t explicitly aware beforehand that they will do so.

(If this all seems trivially obvious to you, well, I’m glad you think so, because this concept will do quite a bit of heavy lifting for us later on.)

One of the crucial elements of this Master Preference (which, I should note, has been proposed in a few slightly different forms before – e.g. John Harsanyi’s concept of “manifest” preferences vs. “true” preferences, Eliezer Yudkowsky’s concept of coherent extrapolated volition, etc.) is that, like many other preferences, you can hold it without ever explicitly affirming that you hold it – or, for that matter, even consciously realizing that you hold it. At no point is it necessary for you to acknowledge (even to yourself) that you’d prefer an outcome like “being deprived of a beverage against your will” to one like “dying because you didn’t realize your beverage was poisoned;” all that’s necessary for such a preference to exist is that you actually would get more value out of the first outcome happening than the second outcome happening. The mere fact that you’d derive more value from the first situation than the second one is what would constitute the first one being preferable to you in the first place. And in that sense, simply the fact that you have any preferences at all (whether conscious or unconscious) means that you must necessarily hold this Master Preference, because all other preferences – merely by virtue of being preferences – are automatically subsumed by it. Even if you were the kind of stubborn contrarian who would explicitly deny having the Master Preference, claiming that you’d never want any of your object-level preferences overruled for any reason, you still wouldn’t be able to get away from it – because the principle of never wanting your object-level preferences overruled is itself a preference that might turn out to be misguided; and if there were ever a situation in which having that preference overruled would be preferable to you, then the Master Preference would supersede it. (Granted, you might never encounter such a situation – maybe by sheer luck, your valuations really would always be perfect – but that wouldn’t change the fact that the Master Preference would still remain in effect for you; it simply wouldn’t ever require anyone to act on it.) To deny that you’d ever want your object-level preferences overruled if doing so would give you greater utility would be an incoherent self-contradiction – it’d be like saying that what was preferable to you wasn’t preferable to you.

It’s for this same reason, by the way, that if someone’s judgment is being impaired somehow, utilitarianism doesn’t necessarily require that all their object-level preferences be met. That is, if someone is (say) in the grip of a crippling drug addiction, it isn’t automatically a good thing to give them more drugs – even if they say that’s what they want – because they’ll be far better off if they can break that addiction and come out of their impaired state of consciousness. That’s the outcome that would actually be preferable to them, even if they don’t realize it at the time. Similarly, if someone suffers brain damage and loses the ability to make good decisions for themselves, the best thing to do isn’t just to satisfy whatever misguided preference they might assert; it’s to do what actually maximizes their utility. This is also true for people who are born with certain cognitive disabilities, who never have the capacity for making fully informed decisions in the first place; merely the fact that they’re capable of considering some states of their existence better than others is all it takes for the Master Preference to apply to them, and to therefore make it morally good for them to be treated well (even if they never consciously understand that that’s what they want). And in fact, this is even true for animals, despite the fact that they lack the biological capacity for human-level judgment altogether; again, merely the fact that they’re capable of preferring some outcomes to others is enough to mean that the Master Preference applies to them, and that maximizing their utility is morally better than not doing so – regardless of whether they can ever fully understand what would be best for them or why. In other words, even if your family dog really wanted to run out into busy traffic – maybe because she saw a rabbit on the other side of the highway or something – it would be morally better not to let her, because getting hit by a car would be such a terrible outcome for her (and would certainly be an outcome she’d want to avoid if she were aware of it). Likewise, even if she didn’t understand the nutritional content of the food you were feeding her – and didn’t know that it would be in her best interest to prefer the healthy kibble over the kibble-flavored asbestos – it would still be morally better to feed her the healthy kibble than the asbestos. Merely her ability to instinctively want what’s best for herself is all it would take to reify that preference as a moral consideration.

I want to talk a little more about how these ideas can apply to non-humans; but before I do, I should clarify a few more conceptual points. First of all, if it wasn’t clear already, when we talk about people having preferences and wanting to maximize their utility, that doesn’t necessarily just mean wanting to selfishly do things that are only good for themselves. A lot of people derive immense satisfaction from things like helping others and abstaining from worldly pleasures – so for those people, the thing that would maximize their utility might not necessarily be a purely hedonistic lifestyle; it might be to spend their lives doing charity work or some other form of public service instead. Similarly, there are some instances in which a person’s preference isn’t to enhance their own well-being at all, but to sacrifice it (if necessary) for the sake of something they value more. (Think about a parent sacrificing their time or their health – or even their life – for their children, for example.) So when we talk about calculating net utilities and figuring out what would best meet people’s interests, we have to account for the fact that people don’t always prefer to just benefit themselves alone. Sometimes, the best way to help a particular person is actually to help others.

On that same note, because some people’s preferences end up producing greater benefits beyond themselves than other people’s preferences do, we have to make sure to incorporate those second-order effects into our utilitarian calculus when we weigh people’s preferences against each other. For example, if we were in some kind of scenario where we had to choose between releasing a sadistic serial killer from prison and releasing a compassionate aid worker who had also been imprisoned, then even though both of them might value their freedom equally highly, it’s clear that freeing the aid worker would produce more goodness overall due to all the positive utility they’d bring to others later on – so the utilitarian choice would be to free the aid worker. Or to take a slightly less obvious example, if we decided to donate a bunch of money to charity, it wouldn’t be right to assume (as a lot of people unfortunately seem to do) that our donation would be equally good regardless of which charity we actually donated to or how much utility the charity produced relative to other charities; the most utilitarian thing to do would be to figure out exactly which charities were doing the most good per dollar, and donate to one of those. As I’ve been emphasizing all along, the best actions within utilitarianism are those that produce the highest level of goodness within the most all-inclusive possible context – the full state of the universe and all the sentient beings within it.

With all these clarifications out of the way, then, let’s take a step back and return to our original question of whether it’s possible to say that things can be objectively good or bad – and let’s apply it to the whole global state of affairs. Is it possible to coherently talk about “moral progress” in the world? Can we really say that our species’ moral behavior has improved (or can improve) in any way over the course of history? A moral relativist might argue that it’s impossible to do so – that morality is entirely relative to its cultural context, and that what’s good in one era might be bad in another era. But within the utilitarian system we’ve set up here, we can say decisively that moral progress is possible. States of the world that have higher levels of utility are objectively better than states of the world that have lower levels of utility – regardless of era, and regardless of context; and to whatever extent we manage to bring our actions into alignment with what would bring about the highest level of global goodness, that’s the extent to which we’re making moral progress as a species. If we implement some new policy or program that makes everyone happier, that’s obviously a good thing. If we implement a policy that has some drawbacks and makes some people unhappy, but still does more good than harm overall, then that may not be ideal, but it can still be considered a positive. But even if we implement a policy that everyone seems to be against (at least at the object level), like legalizing interracial marriage in our hypothetical country from before, it can still be a good thing overall if it ultimately increases the people’s utility, because then it would still satisfy their Master Preference, which is the true measure of their well-being in the end.

Of course, I should caution here against abusing this idea of the Master Preference – because it really is easy to abuse if not understood properly. History is tragically rife with examples of people harming and oppressing others and claiming that it’s “for their own good.” But just because people have a meta-preference that their object-level preferences should be overruled if it would better increase their utility doesn’t mean that it would naturally be a good thing for us (or anyone else) to forcefully impose our preferred way of life on everyone else just because we think it would be better for them. The only way we could call such an intervention good would be if it actually were better for them (which forceful actions hardly ever are); otherwise, it would most definitely be a bad thing. (And I should point out that one of the biggest reasons of all why an intervention might not be morally acceptable is the simple fact that respecting people’s explicit wishes is such an important norm in itself – so even if a particular intervention seemed otherwise worthwhile, the mere fact that it would undermine this norm might still keep it from being justifiable.) In short, then, while it is in fact possible to make moral progress in the world, there’s a difference between actually making moral progress in the world, and causing things to become worse in the name of making moral progress. It’s critically important, when making any kind of major moral decision, to keep this distinction in mind – and to have enough humility to recognize that if you think somebody’s object-level preferences should be disregarded for the sake of upholding their Master Preference, it’s always possible that you may actually be the one whose preferences are misguided and ought to be overruled, both for your own sake and for the sake of those who will be affected by your actions.

Continued on next page →