Does Negative Utilitarianism Override Individual Preferences?

By Brian Tomasik

First published: . Last nontrivial update: .

Summary

I this article, I defend negative utilitarianism (NU) against the charge that it overrides individual preferences. I suggest that the contrast between negative and non-negative flavors of utilitarianism can be seen as boiling down to the issue of interpersonal comparisons of utility. Since there's no non-arbitrary way to interpersonally compare utility, there's also no non-arbitrary basis for claiming that non-NU is more respectful to individuals' preferences than is NU.

Contents

The allegation

A common charge against NU is that it may contradict the preferences of the individuals whose interests it aims to consider. For example, suppose a person currently feels that his life is worth living and is glad to have been born. NU might, depending on the extent of suffering in this person's life, say that this person's life has net negative value overall. Shouldn't we try to respect the wishes of agents, rather than forcing how we feel onto them?

We can problematize this allegation in a number of ways. For example, we might ask how to non-arbitrarily identify a person's true, idealized preferences at a given moment if idealized preferences differ from that person's stated or revealed preferences. We might ask what to do in situations where the parts of a person's brain that control speech favor an action that other, less powerful parts of the person's brain disfavor (Tomasik "What ..."). However, in this piece I'll focus on one particular problem for the allegation that I think is most severe.

Interpersonal comparisons of utility

Imagine that the entire history of the universe contained only a single individual for a brief period. If that individual prefers X over Y, then it seems plausible that it's morally better for X to be the case than Y.

However, things become tricky as soon as we have more than one individual, if there are conflicting preferences. If Alice prefers X to Y while Bob prefers Y to X, is it morally better for X or Y to be the case? There's no obvious answer. In order to choose between X and Y, we as moral arbiters have to decide if we care more about Alice's preference or Bob's preference. This introduces inevitable arbitrariness into our moral views.

Suppose we think Alice's preference is more weighty than Bob's and favor outcome X. Have we thereby disrespected Bob's preferences about what he prefers? Sort of, but there was nothing we could have done not to disrespect someone's preferences in this case.

Application to NU

Suppose the universe contains only Alice and Bob. Bob is glad to exist and considers his life to have net positive welfare. Alice despises her life and wishes she hadn't been born. Would it be better if neither Alice nor Bob existed? Or is it better that they exist?

As we saw above, there's no "right way" to make this judgment call. If we favor nonexistence, we violate Bob's preference (or, at least, what would have been Bob's preference had he existed). If we favor existence, we violate Alice's preference.

A negative utilitarian can be seen as someone who, when making these arbitrary interpersonal comparisons of utility, tends to take preferences not to exist or to avoid harm very seriously. Meanwhile, a more positive utilitarian can be seen as someone who takes preferences to exist or to have pleasurable experiences very seriously.

Loosely, we could say that negative utilitarians "give more weight" to anti-suffering preferences in interpersonal comparisons than non-negative utilitarians do. But such a statement shouldn't be interpreted as implying that there's any non-arbitrary baseline of how much moral weight different preferences should get (see Knutsson 2016). We could just as well say that non-negative utilitarians "give more weight" to pro-happiness preferences.

Inter-person-moment comparisons of utility

The classic problem of interpersonal comparisons of utility assumes that each individual has a single utility function. But in fact, people's preferences change over the years, and even from minute to minute. So there's also a problem of intrapersonal comparisons of utility across time. If we think of each moment of an individual as a different "person-moment", then we could call this the problem of inter-person-moment comparisons of utility.

Bob may affirm his own existence at time 1 but, due to unbearable suffering, wish he didn't exist at time 2. If we're faced with the choice of either creating Bob's existence or preventing it, which is the right choice? There's no clear answer. Either way, we'll contravene the wishes of Bob-at-time-1 or Bob-at-time-2. (Actually, if Bob never exists, we never contravene his wishes because he doesn't have any wishes, so we could also argue for nonexistence on those grounds.)

What kind of NU is this?

The arbitrariness of interpersonal and inter-person-moment comparisons of utility is perhaps my favorite way to think about NU. It's not a moral tragedy if someone experiences a pinprick but still, during that moment, is overall happy to be alive. The moral force of NU comes from instances of extreme suffering that, to the person afflicted, feel so bad that the person wishes the world had never existed. These are the kinds of intense preferences where it feels genuinely morally unclear how to weigh these preferences against the pro-existence preferences of other people or other person-moments.

An approach to interpersonal/inter-person-moment comparison that always respects nonexistence preferences over existence preferences, and always respects preferences not to have some collection of experiences over preferences to have some collection of experiences, is what I've elsewhere called consent-based NU. For example:

In cases where both available options would cause some amount of non-consensual suffering, consent-based NU favors the action that causes fewer person-moments to non-consensually suffer, causes less intense non-consensual suffering, and so on.

We can also imagine a weaker form of NU than full consent-based NU by generally giving precedence to anti-existence and anti-suffering preferences while sometimes letting pro-existence and pro-happiness preferences win out if the number and intensity of "pro" preferences is large enough compared with the "anti" ones.

How much does preference-respecting NU differ from hedonistic NU?

I'm accustomed to thinking in terms of hedonistic NU, rather than preference NU, because I personally care more about hedonic experiences than non-hedonic preferences. However, I also feel the force of the argument against hedonistic NU in cases where hedonistic NU seems to override an individual choice in which all person-moments agree about the choice. (Note that hedonistic symmetric utilitarianism can also sometimes override individual choice.) This led me to wonder how much of a difference there is in practice between hedonistic NU and a version of NU closer to what I've been discussing in this article.

Ord (2013) presents a "worse-for-everyone argument" against hedonistic NU:

suppose that there was a situation in which all individuals want to accept 5 wellbeing units of suffering in order to gain 10 wellbeing units of happiness. This would be in everyone's interest. However, [...] NU would say that it was impermissible, and that it is instead obligatory to prevent it [...].

Let's assume that while enduring those 5 units of suffering, everyone still agrees that the tradeoff was a good idea. Otherwise, the trade wouldn't actually make everyone better off—it would make the person-moments who reject it during the period of suffering worse off. (There's also the issue that some suffering subroutines within people's brains might reject the trade, even if everyone's higher-level verbal thoughts always accept the trade. However, I won't worry about that problem here.)

Vinding (2022) offers several caveats and critiques regarding Ord's argument. For example, in many cases where people seem to accept suffering in exchange for happiness, they're really accepting suffering in order to reduce other suffering. Ord mentions going to the gym as an example suffering-for-happiness trade, but in my experience, exercising is mainly useful for preventing me from feeling agitated, stressed, etc. (Also, I happen to find exercise fun, not painful, though that's because I don't do it really intensely.) Other people might go to the gym to reduce the risk of painful health conditions or to ameliorate distressing feelings of being ugly or lazy. So with Ord's example, we should make sure to imagine a situation where the 10 happiness units really are just added happiness, not avoiding some other suffering. Note that cravings for pleasure that one is missing out on are themselves a form of suffering, so we have to assume that the people in the thought experiment wouldn't feel bad if they didn't get the happiness—or at least that their feeling bad would be less than 5 units of suffering (Vinding 2022).

However, assuming we satisfy all of those conditions, then it does seem to me that it's ok and maybe even net good if people experience 5 units of suffering for 10 units of happiness as long as they endorse the trade throughout. We can see this as an extension of the principle that it's ok for one person to choose to suffer for the benefit of someone else. In the case of Ord's example, the suffering person-moments are volunteering to suffer for the sake of the later happy person-moments.

One of the most important areas where this issue arises is deciding whether it's good or bad to allow new beings to come into existence. Creating a new being is a "package deal" that involves some suffering and some happiness.

Imagine (unrealistically) a person who actively endorses having been born at all moments during her life, even during times of suffering, because she thinks her life has net positive wellbeing. A sufficiently suffering-focused hedonistic NU would say that she experiences some suffering, so her life is net bad, and her suffering contributes to the world overall being net bad. However, a consent-based NU would say that this person's life is fine, because she didn't object to it overall at any moment. So we shouldn't count her suffering as saying that the world is made worse by her birth (ignoring the side effects she has on other beings), because her suffering is able to be offset by her happiness.

By the way, note that this case is consistent with Vinding (2022)'s reminder that the happiness in the tradeoff should be just happiness and not something that offsets other suffering, because if this person didn't exist, she wouldn't suffer at all. So her wanting to exist isn't because doing so will reduce her suffering in some other way. (That said, the person might think she would feel sad if she didn't get to exist, even though she actually wouldn't feel anything if she didn't exist. This kind of error in imagination may be fairly common, and it can infect people's judgments about whether they're actually net glad to exist.)

A case where a person endorses having been born during every moment of life seems very unlikely if not impossible. Most people have moments, even if brief, during which they wish they didn't exist. These moments might be due to depression, anxiety, pain, frustration, or other emotions. According to strict consent-based NU, this renders these lives net bad. However, it's plausible to me that we should be a bit less strict than this, especially if the "I wish I didn't exist" moments are rare, brief, and not the result of extreme pain but rather some less severe mood swing. A weak consent-based NU could look at the ratio of "I wish I didn't exist" moments compared with "I'm glad I do exist" moments, where each moment is counted based on its intensity, and if this ratio is small enough, the life isn't net bad.

I tend to picture this in terms of arrows. We can imagine arrows pointing up as endorsing one's existence, and arrows pointing down as rejecting one's existence. The length of each arrow represents how strongly this opinion is felt. Moments of unbearable agony would be extremely long downward arrows. Strict consent-based NU says that any downward arrow renders the life net bad. Weak consent-based NU says that downward arrows count a lot more than upward ones, but if the cumulative length of upward arrows is big enough compared with the cumulative length of downward ones, then existence can be ok overall.

Carl Shulman echoes Ord's argument about hedonistic NU overriding people's choices using a more somber example: the Holocaust. Shulman says that for concentration-camp victims who wanted to live, killing them (or, more relevant to my discussion, preventing their existence in the first place) "if they would otherwise escape with a delay would not be helping them for their own sakes, but choosing to be their enemy by only selectively attending to their concerns." We could compare this to Ord's example of enduring 5 units of suffering (in the concentration camp) for 10 units of later happiness (after the prisoners escaped), although the motivations of real humans are much more complex than just avoiding suffering and seeking happiness. If during all moments of life, a concentration-camp victim still was glad to exist on balance, then consent-based NU would agree with Shulman that this person should be allowed to exist. Meanwhile, if a moment of that person's life actively rejected having been born, then it's no longer fair to say that NU is being an "enemy" to this person's wishes, because the person has conflicting wishes at different points in time. Consent-based NU takes the side of the anti-existence wishes in the dispute. If it seems too extreme to reject existence based on a single fleeting moment of wishing one hadn't been born, we could adopt a weaker consent-based NU that requires a higher number of or more intense anti-existence preferences before saying that a life shouldn't have existed on balance. My personal intuition on this topic is that a few brief suicidal-type feelings during a period of erratic emotions aren't enough to make an otherwise nice life net bad. However, if the anti-existence sentiments come from more than a fleeting moment of unbearable physical pain, such as due to torture or horrific injury, I think it's unfair for the other parts of a person's life to impose that suffering on those moments of agony who desperately wish their life had not happened. Those tormented person-moments might address their happy counterparts using Greta Thunberg's famous phrase: "How dare you?" How dare the happy parts of the person's life subject the tormented parts to unspeakable horrors? (Of course, the happy parts of the person's life could say the same thing to the anti-existence parts: How dare you prevent us from being born?)

Ord (2013) notes regarding his "worse-for-everyone argument" that:

Lexical Threshold NU could avoid [the situation of NU overriding personal choice] in cases below its threshold if it said that happiness and suffering were equally important below that level, but this would create a particularly odd kind of discontinuity [...].

In consent-based NU, if suffering is below the threshold of unbearableness, such that the person still endorses his existence overall, then happiness can outweigh that suffering. Meanwhile, if some person-moment judges his suffering to be so bad that he wishes he hadn't been born, happiness can't outweigh that suffering. This approach implies a lexical threshold at the point where suffering becomes so intense that it leads to rejecting existence. This threshold isn't "particularly odd" at all but falls out naturally from paying attention to what the various person-moments of the life are saying about their own endorsement or rejection of having been born.

In weak consent-based NU, the point where a person switches from endorsing to rejecting existence wouldn't create a fully lexical threshold, but there would still be a threshold at which the seriousness of the suffering jumps considerably. Downward-pointing "anti-existence" arrows weigh much more heavily than upward-pointing "pro-existence" ones do.

Consent-based NU could differ from lexical threshold NU if a person's endorsement or rejection of existence is based on something other than an assessment of the net hedonic balance of his life. For example, suppose someone is quite miserable and expects to suffer more than he's happy, but he still prefers to have been born. If he endorses his existence throughout life, then consent-based NU would be ok with his existence, while hedonistic lexical threshold NU (and even hedonistic symmetric utilitarianism) would have favored his nonexistence.

I find it difficult to decide on the right answer in cases like these. In an extreme case, we could imagine a world filled with creatures who endure unending agony but still prefer to have existed. One possible reply to such cases is that if we look at the brains of these agents, we'd probably see a subcomponent that speaks strongly against existence (namely, the part of the brain creating the agony sensation), while other parts of the brain are forcibly overriding that in order to cause the overall preference for existence. So we could see the brain as containing a sub-agent who doesn't consent to existence, which means we should actually reject existence in this case. But it's unclear how far to extend that reasoning, since all of our brains presumably contain some subagents who are upset at any given moment.

Another tricky moral question is how to handle cases where it's unclear if the agent endorses or rejects existence. Most of the time, people aren't actively reflecting on Hamlet's question of whether it's better to be or not to be. If a person is feeling moderate pain without reflecting on whether it's worth enduring that pain for the good things in life, does his brain count as implicitly rejecting existence in that moment? Or should we say that people who generally endorse their existences implicitly carry around that preference even when not actively thinking about it? Maybe one solution to this problem in the case of verbal humans is to hypothetically ask the person-moment during the suffering whether that suffering is worth enduring for his overall existence. However, this solution isn't available for babies or non-human animals.

In the case of babies and non-human animals, another consideration is that they may not have the cognitive capacity to reflect on whether their lives are net good anyway. Their experiences are presumably somewhat more moment-to-moment, reacting strongly in favor of or against what's happening right now. Whereas an adult human might think "This pain sucks, but I'm still overall glad I exist", a fish might merely think "This pain sucks." If the moment of suffering just feels like "Ouch! Make this stop!" without any further qualifications, that would seem to be a downward arrow against existence? (I think an academic philosopher has made roughly this point in the case of animals: that if their lives are fairly moment-to-moment, then each moment of their existence counts as a separate individual, so the happiness of some moments can't outweigh the suffering of other moments. I may try to find this paper at some point.)

If we do think that endorsing existence while suffering is uncommon in the case of non-human animals, then hedonistic NU and preference-respecting NU would be mostly in agreement there, since most hedonic suffering would be nonexistence-preferring suffering. And since most of the sentient experience on Earth at the moment is in non-human animals (or other even simpler minds), hedonistic and preference-respecting NU wouldn't differ very much in total.

As far as risks of astronomical future suffering, I think a decent portion of that expected suffering is from scenarios in which agents are created who endure fairly constant and severe torment during most or all of their lives. Presumably most of that suffering would be in moments when they reject their existence. Hence, most of that hedonic suffering would also be preference-respecting anti-existence suffering.

It's plausible that in most cases, when pain gets bad enough, people reject their existence. It's common for people to think "I want to die" when their suffering is too intense to bear. So at the very least, most unbearably intense suffering is also anti-existence suffering. Thus, hedonistic NU and preference-respecting NU should generally agree in cases of the very worst suffering. I already feel like the very worst cases of suffering are vastly more important than mild suffering, and if we adopt that perspective, the gap between hedonistic and preference-respecting NU may not be very large.

Note that there are many cases of non-extreme suffering in which a person still rejects having been born, such as if a person is depressed or just has a generally miserable life. In my case, I'm not depressed, and my life is hedonically ok, but I still wish not to have existed anyway (from a selfish rather than altruistic perspective), because I expect to suffer more severely at the end of my life, and there's a small chance of experiencing really bad suffering. Plus, I don't think the joys of existing are that important, though I think they do have a little bit of positive value.

Acknowledgments

I first made an argument along these lines in a 2013 online discussion. Simon Knutsson makes essentially the same point, spelled out in more detail, in "What Is the Difference Between Weak Negative and Non-Negative Ethical Views?"