This post is a reply to Johan Gustafsson’s draft paper “Against Negative Utilitarianism”. Gustafsson acknowledges that for many common objections raised against negative utilitarianism (NU), there are corresponding objections that can be raised against classical utilitarianism (CU) (see e.g. Knutsson, 2021a). Hence, as he writes, “these objections have little force when we assess the relative merits of Classical and Negative Utilitarianism” (Gustafsson, 2022, p. 1).
The aim of Gustafsson’s paper is to present novel counterexamples against NU that have no analogues in the case of CU. My aim in this post is to show that CU does face such analogous counterexamples, and that these counterexamples are worse than those facing NU. I also argue that views that give overriding importance to the reduction of extreme suffering seem uniquely plausible in light of the counterexamples reviewed here.
Gustafsson’s main counterexample: Bliss versus Torture
The main counterexample Gustafsson raises against NU is the following (p. 3):
Bliss versus Torture
You have a choice between the following outcomes, where the same people live for the same duration:
A Everyone gets a century of pure bliss followed by a pinprick.1
B Someone gets a century of torture, and everyone else gets a century of no pleasure and no pain.
Given a large enough population, you ought to choose B over A according to Negative Utilitarianism.
Gustafsson continues (pp. 3-4):
Yet A seems more choice-worthy than B on, basically, any plausible moral metric: A is overwhelmingly in everyone’s subjective interest (given, as seems plausible, that everyone strongly prefers ending up in A to ending up in B). A is more equal than B. The worse-off are better off in A than in B. There is less torture in A than in B. And so on. Moreover, it seems that Bliss versus Torture lacks an analogue for Classical Utilitarianism.
Favoring the most plausible choice by accident?
I will present analogous counterexamples against CU in the next section. But first, I think it is worth clarifying that the counterexample raised above does not apply to all versions of NU. In particular, many lexical versions of NU would maintain that B is worse than A, because these views hold that the suffering caused by the torture in B is worse than any amount of pinpricks. (Gustafsson also provides a counterexample against such lexical versions of NU, which I will reply to below.)
In other words, Gustafsson’s counterexample only applies to versions of NU that make certain (highly non-trivial) assumptions about aggregation, which many proponents of NU — and similar suffering-focused views — would firmly reject (see e.g. Gurney, 1887, ch. 4; Mendola, 1990; Ryder, 2001; Leighton, 2011).
Indeed, I think Gustafsson’s counterexample derives its force primarily from the repugnance of choosing a century of torture in order to avoid many minor pains — an implication that is also shared by CU when those bads are pitted against each other in isolation (cf. Gustafsson, 2022, p. 2).
A proponent of views that give lexical priority to extreme suffering could thus argue that CU only happens to give the most plausible answer by accident in Gustaffson’s example, because the addition of bliss in A happens to align the CU choice with these lexical views. And a proponent of such views could then further argue that if we remove the bliss in the example above — such that A involves that “everyone gets a century of no pleasure and no pain, followed by a minor pain” — we see that non-lexical versions of both CU and NU are less plausible than views that assign lexical disvalue to extreme suffering (because the non-lexical views would choose the century of torture, i.e. B, in that case).2
At any rate, it is not correct when Gustafsson writes that his counterexample “cannot, plausibly, be blocked by clinging to the intuition that evil and suffering have greater moral import than goodness and happiness” (Gustafsson, 2022, p. 3). Those who endorse versions of NU that give overriding priority to extreme suffering can indeed hold on to that intuition while avoiding the choice of torturous suffering over many minor pains.
Counterexamples to classical utilitarianism
I think there is not only one, but many counterexamples to CU that are similar to Gustafsson’s Bliss versus Torture example. Below are two quite similar ones.
Counterexample I: Torture for Micro Pleasures
Torture for Micro Pleasures
Assume that a micro pleasure has the same intensity as a micro pain, such that a classical utilitarian would say that these respective states cancel each other out. You have a choice between the following outcomes, where the same people live for the same duration:
A1 Everyone gets a century of hedonic neutrality followed by a micro pain.
B1 Someone gets a century of torture, and everyone else gets 50 years of micro pains plus 50 years of micro pleasures, followed by two micro pleasures.
Given a large enough population, you ought to choose B1 over A1 according to classical utilitarianism.
One could here make the same claims that Gustafsson makes regarding his counterexample, about how A1 seems more choice-worthy than B1 on any plausible moral metric. In particular, it is trivial to see that A1 is more equal than B1, the worse-off are better off in A1 than in B1, and there is less torture in A1 than in B1. And one could argue that A1 is in everyone’s interest, and that most people probably would prefer to live in A1 rather than B1.
Justification for this latter claim can be found in descriptive research on people’s preferences regarding tradeoffs between suffering and happiness (or pain and pleasure). Such research finds that people tend to endorse a significant asymmetry between happiness and suffering, especially as far as lives in potential worlds are concerned (Caviola et al., 2022; Contestabile, 2022, sec. 4).
Indeed, one can argue that this example against CU is even stronger than Gustafsson’s example against NU, since we are here ultimately allowing torture for the sake of micro pleasures. That is, while allowing torture for the sake of reducing micro pains seems implausible and repugnant (at least to many people), it is arguably more implausible and more repugnant to allow torture for the sake of creating micro pleasures (cf. Vinding, 2020, ch. 3; 2021).
Brief reply from Gustafsson
In a footnote, Gustafsson brings up essentially the same thought experiment as the one I raised above, and attempts to briefly argue that it is not analogous to his counterexample against NU.
Before proceeding to Gustafsson’s argument, I should say that I find it odd that he only devotes a single footnote to the discussion of potentially analogous counterexamples against CU, seeing that his principal claim is that the counterexample he presents against NU has no analogue in the case of CU. Such a claim would seem to warrant elaborate discussion of potential analogues, to convincingly show that they are in fact not genuine analogues. As I will argue below, his cursory remarks in his current draft do not succeed in showing this.
Here is what Gustafsson writes in response to the counterexample (p. 4):
But, if the micro pleasures really have the corresponding intensity to the pinpricks, they should outweigh the pinpricks in [B1] (otherwise they wouldn’t have the corresponding intensity according to Classical Utilitarianism). If you don’t find that intuitive, you may be imagining micro pleasures of too low intensity. Once the intensity is imagined correctly, [B1] should then be in most people’s subjective interest (or, at least, [A1] would not be overwhelmingly in their interest).
This reply is unsatisfactory, since it simply begs the question in favor of CU. After all, a negative utilitarian could similarly argue that a micro pain cannot be outweighed by a micro pleasure (because otherwise the micro pain wouldn’t be a genuine micro pain according to NU). That micro pleasures can outweigh micro pains is a key premise that needs to be established, and hence it cannot simply be asserted like this.
More glaring, however, is that Gustafsson’s reply wholly ignores the most problematic aspect of this counterexample against CU, namely that it allows micro pleasures to outweigh a century of torturous suffering, which seems uniquely repugnant, and seems to render this counterexample worse than the one he raises against NU (cf. Vinding, 2020, ch. 3; 2021).
Counterexample II: Torture for Everyone for Micro Pleasures
The following counterexample against CU is even more serious, and hence arguably stronger still compared to the counterexample raised by Gustafsson against NU.
Torture for Everyone for Micro Pleasures
You have a choice between the following outcomes, where the same people live for the same duration:
A2 Everyone gets a century of hedonic neutrality.
B2 Someone gets a century of torture and everyone else gets 50 years of torture (i.e. intense suffering) and 50 years of “correspondingly” intense pleasure, followed by one micro pleasure.
Given a large enough population, you ought to choose B2 over A2 according to classical utilitarianism.3
Again, one can make all the same claims that Gustafsson makes in relation to his counterexample against NU: A2 is more equal than B2, the worse-off are better off in A2 than in B2, and there is less torture in A2 than in B2 — in fact, there is far less torture, since everyone gets tortured for at least 50 years in B2, whereas nobody even suffers in A2. Thus, in terms of the “least torture” criterion that Gustafsson mentions in his paper (a highly important criterion, in my view), the difference is much greater in this case than in the counterexample that Gustafsson presents against NU.
It likewise seems plausible to claim that A2 is in everyone’s subjective interest, and that most people would strongly prefer to end up in A2 rather than B2. After all, not only do people generally endorse a significant evaluative asymmetry for “equally intense” pleasure and pain, but this asymmetry seems even more pronounced in the case of intense suffering versus intense pleasure (Caviola et al., 2022, p. 6). Indeed, in one small informal survey (n=99), roughly 45 percent said that they would not endure just one minute of intense suffering for any number of happy years added to their life (Tomasik, 2015).
In everyone’s subjective interest?
Suppose that a proponent of CU wanted to argue that these counterexamples against CU are not fully analogous to, or not as devastating as, the counterexample that Gustafsson provides against NU. How could a proponent of CU do this?
One strategy might be to focus on Gustafsson’s claim that, in the counterexample he provides, “A is overwhelmingly in everyone’s subjective interest (given, as seems plausible, that everyone strongly prefers ending up in A to ending up in B)” (Gustafsson, 2022, p. 3).
After all, if we take the case of Torture for Everyone for Micro Pleasures, one could argue that it is not clear that everyone would strongly prefer to end up in A2 rather than B2. It seems likely that at least some people would choose B2 (e.g. if they were asked in a survey).4
Yet the same objection can be raised against Gustafsson’s original claim. Indeed, an important class of axiological views would say that the individuals in B who experience absolutely no pain in their entire lives are better off than all the individuals in A, who each experience both bliss and (one) pain. Thus, Epicureans and many Buddhists would hold that the pain-free lives are in fact better, and that the absence of pain (and other problematic states) is in some sense the highest bliss (see also Schopenhauer, 1819; 1851; Fehige, 1998; Geinster, 1998; 2012; Gloor, 2017; Ajantaival, 2021/2022; Knutsson, 2022).
In other words, Gustafsson’s claim that “A is overwhelmingly in everyone’s subjective interest” is rejected by (what many consider) reasonable axiological views, and adherents of these views would likewise deny that “everyone strongly prefers ending up in A”. In particular, proponents of these axiological views could argue that A is not in the subjective interest of the experience-moments that undergo the pain of a pinprick, and then further maintain that these dispreferred experience-moments are not outweighed by pleasure that occurs in other experience-moments. (And those who endorse these views might in any case say that they personally would prefer a hedonically neutral life over any life that contains even the smallest amount of pain or suffering.)
A proponent of CU might instead argue for a weaker claim, namely that most people would strongly prefer to end up in A over B (in Gustafsson’s Bliss versus Torture example). And one could then further argue that even if most people would also prefer to end up in A2 over B2 (in Torture for Everyone for Micro Pleasures), it still seems likely that the majority that favors A over B is significantly larger than the majority that favors A2 over B2.
This claim is, of course, quite uncertain. After all, 50 years of torture is quite a lot to sign up for, and we have reason to expect that the vast majority of people would prefer to not endure 50 years of intense suffering in order to gain 50 years of (“similarly”) intense pleasure plus a micro pleasure (cf. Tomasik, 2015; Caviola et al., 2022; and this informal survey).
Yet even if we grant that the “greater majority” claim above is true, what would follow? This “greater majority in one case than the other” criterion seems quite strange and ad hoc, and it is not clear why we should consider it particularly relevant. After all, most people could be wrong about what is good for them, or about what the experiences they choose to undergo will in fact end up feeling like.
Moreover, we can reasonably ask why this (seemingly ad hoc) “greater majority” criterion should be given greater importance than other criteria, including the “less torture” criterion mentioned by Gustafsson, on which the difference between A2 and B2 is vastly greater compared to the difference between A and B.
All in all, the second counterexample raised against CU above does not appear to be weaker than, nor meaningfully disanalogous from, the counterexample Gustafsson raises against NU. On the contrary, one can argue that it is considerably stronger overall, not least since it forces CU to accept torture for everyone for the sake of micro pleasures, while Gustafsson’s counterexample against NU “only” implies torture for one individual, and only for the sake of reducing (supposedly) greater suffering (in the form of micro pains) — not for the sake of creating pleasure whose absence causes no problem (cf. Vinding, 2020, ch. 3).
Objection against “critical-level lexical NU”
Gustafsson concedes that his counterexample against NU is not applicable to views that grant lexical priority to the reduction of intense suffering. Yet against such views, he presents another counterexample (p. 6):
Bliss and Severe Pain versus Almost-Severe Pains
You have a choice between the following outcomes, where the same people live for the same duration:
C Everyone gets a century of pure bliss followed, for someone, by the briefest, least severe pain that counts as severe.
D Everyone gets a century of pain that is just slightly less severe than the critical level in severity.
On the lexical view, we ought to choose D, no matter how large the population is.
There are basically two kinds of replies to this counterexample, which are sketched out below.
Reply from the perspective of abrupt lexical views
Someone who endorses an abrupt threshold at which lexicality kicks in would argue that it is not, in fact, implausible to accept the conclusion presented by Gustafsson above, as long as we are careful to clarify what the scenarios above entail.
In particular, one could argue that no number of mild discomforts could ever be worse than even one instance of genuine pain, and maintain that a sharp threshold exists between these respective states (Klocksiem, 2016). So if we adopt this perspective, the choice above can be equivalently rephrased in the following way:
C Everyone gets a century of pure bliss followed, for someone, by the briefest, least severe pain that counts as a genuine pain.
D Everyone gets a century of mild discomfort that is just slightly less severe than the least severe pain that counts as a genuine pain.
A proponent of lexicality who maintains that C is worse than D would have a number of things to say in response to the following claims made by Gustafsson:
There is a lot more suffering in D than in C. While the pain in C is worse in severity, it is only slightly worse in severity than the pains in D which last much longer and afflict an arbitrarily large number of people. The difference in severity between the pain in C and the pains in D is, we can assume, barely perceptible.
The claim that “there is a lot more suffering in D than in C” is somewhat vague, and adherents of abrupt lexical views may disagree with it in a number of ways. First, they might disagree that D contains any genuine suffering at all, since they might hold that the bad states found in D fall short of qualifying as suffering proper (i.e. they may prefer to reserve this term for states that are worse than mere discomfort).
Second, even if a proponent of this lexical view did consider the bad states in D to amount to suffering, they would still reject the claim that there is more suffering in D than in C in the most relevant sense. That is, while there is a greater duration of suffering in D than in C (if we grant that there is suffering in both), there is nevertheless more suffering in C than in D in terms of the disvalue of the suffering, according to the abrupt lexical view.
Gustafsson’s other claim — about how the bad state in C is just barely worse than those in D — can be met with the reply that virtually all views entail some form of lexicality between one state and a barely worse state. For instance, a classical utilitarian would hold that a barely perceptible pain carries greater evaluative significance than any number of hedonically neutral states, including hedonically neutral states and lives that contain features that other axiological views consider intrinsically bad, such as extra-experiential preference frustration, bad motives, bad acts, etc.5
In other words, virtually all standard views in value theory entail what we may call abrupt but gradual lexical thresholds, whereby the tiniest change implies value lexicality, and hence this feature is not unique to views that imply lexicality between different bads. (One might object that the generic form of lexicality entailed by most views between neutral and non-neutral states is not comparable to lexicality between different bads; a brief reply to that objection is found here.)
Reply from the perspective of non-abrupt lexical views
The second and perhaps more important reply is to point out that Gustafsson appears to overlook non-sharp lexical views, which his objection does not apply to. That is, one can hold that a single instance of extreme suffering is worse than arbitrarily many mild states of suffering, while also maintaining that there is no sharp lexical threshold between them (see e.g. Knutsson, 2016a; 2016b; 2021b; Vinding, 2022).
Gustafsson has thus not established that it is implausible to endorse views that give lexical priority to extreme suffering. Indeed, such views arguably stand as the most plausible ones in light of the various counterexamples reviewed above.6
Appendix: Reply to the rest of Gustafsson’s footnote
In his earlier-mentioned footnote, Gustafsson replies to another potential counterexample against CU. I do not find this counterexample nearly as strong as the ones I have provided above, which makes it a bit of a distraction relative to the stronger counterexamples, especially Torture for Everyone for Micro Pleasures. But it might nevertheless be worth replying to Gustafsson’s objections to this weaker counterexample. (The following counterexample against CU is one of a number of counterexamples I have suggested to Gustafsson.)
Here is what Gustafsson writes (p. 4):
Magnus Vinding suggests a possible analogue, where, in A’, everyone gets a century of hedonic neutrality filled with a very large amount of non-hedonic goods followed by a pinprick and, in B’, someone gets a century of torture and everyone else gets a century of hedonic neutrality followed by two pinpricks and four micro pleasures (pleasures corresponding in intensity to a pinprick). Given a sufficiently large population, you ought to choose B’ according to Classical Utilitarianism.
But this counter-example is disanalogous. It introduces, in addition to pleasures and pains, a third element, namely non-hedonic goods. And, if those goods are good for people, it would merely motivate a switch to a version of Classical Utilitarianism where these non-hedonic goods also contribute to well-being rather than a switch to Negative Utilitarianism.
First, even if this thought experiment introduces an element that goes beyond pleasures and pains, it still implies a highly implausible choice for CU (in its classical form focused on pleasure and pain), and it still fulfills all the same criteria that Gustafsson listed under his counterexample against NU. So the counterexample still appears to speak strongly against (traditional) CU, and it thus seems to merit a response from proponents of (traditional) CU, degree of analogy notwithstanding.
Second, as Anthony DiGiovanni notes, one could argue that there is an important parallel in terms of how CU construes the value of (putative) non-hedonic goods and how NU construes the value of bliss. That is, CU can acknowledge that non-hedonic goods — such as knowledge, relationships, autonomy, etc. — are highly valuable in practice, due to their effects on hedonic states. Yet in abstract thought experiments, CU requires us to ignore these secondary effects, and to only count hedonic states. This theory-practice distinction is why it can seem so counterintuitive that CU would prefer to forgo the creation of arbitrarily many insights, relationships, freedoms, etc. in order to create a single micro pleasure (or so a proponent of CU may argue). Similarly, NU can acknowledge that bliss may be valuable in practice, due to its various positive roles, while nevertheless denying that bliss has any intrinsic positive value. And this, a proponent of NU could argue, is why it can seem counterintuitive to forgo the creation of any amount of bliss for the sake of avoiding the tiniest of pain.
Third, we could simply remove the non-hedonic value entities in the counterexample above, whereby we would still get an analogous counterexample against CU (though one that is still far weaker than Torture for Everyone for Micro Pleasures, which I think should be the primary focus of the discussion). In particular, we would have a counterexample in which CU allows torture for the sake of micro pleasures, and in which minimalist axiologies would say that everyone is better off in A’ compared to B’.
Finally, it is not the case that any presumed significance of non-hedonic value entities necessarily motivates a move to other forms of CU. Such non-hedonic value entities could just as well motivate a move to other forms of NU — i.e. broader harm-focused versions of NU — according to which the absence of certain non-hedonic value entities is bad (even if it causes no pain), and where the presence of these value entities amounts to a less bad state (cf. Fehige, 1998; Benatar, 2006, ch. 2; Knutsson, 2016a).
Acknowledgments
For helpful feedback, I am grateful to Teo Ajantaival, Tobias Baumann, Anthony DiGiovanni, Simon Knutsson, and Winston Oswald-Drummond. I also wish to thank Johan Gustafsson for engaging in a dialogue about his paper.
References
Ajantaival, T. (2021/2022). Minimalist axiologies. Ungated
Anonymous. (2015). Negative Utilitarianism FAQ. Ungated
Benatar, D. (2006). Better Never to Have Been: The Harm of Coming into Existence. Oxford University Press.
Breyer, D. (2015). The Cessation of Suffering and Buddhist Axiology. Journal of Buddhist Ethics, 22, pp. 533-560. Ungated
Caviola, L. et al. (2022). Population ethical intuitions. Cognition, 218, 104941. Ungated
Contestabile, B. (2022). Is There a Prevalence of Suffering? An Empirical Study on the Human Condition. Ungated
DiGiovanni, A. (2021a). Tranquilism Respects Individual Desires. Ungated
DiGiovanni, A. (2021b). A longtermist critique of “The expected value of extinction risk reduction is positive”. Ungated
Fehige, C. (1998). A pareto principle for possible people. In Fehige, C. & Wessels U. (eds.), Preferences. Walter de Gruyter. Ungated
Geinster, D. (1998). Negative Utilitarianism – A Manifesto. Ungated
Geinster, D. (2012). The Amoral Logic of Anti-Hurt (Modified Negative Utilitarianism). Ungated
Gloor, L. (2017). Tranquilism. Ungated
Gurney, E. (1887). Tertium quid: Chapters on various disputed questions. Kegan Paul, Trench, & Co. Ungated
Gustafsson, J. (2022). Against Negative Utilitarianism. Ungated
Klocksiem, J. (2016). How to accept the transitivity of better than. Philosophical Studies, 173(5), pp. 1309-1334.
Knutsson, S. (2016a). Thoughts on Ord’s “Why I’m Not a Negative Utilitarian”. Ungated
Knutsson, S. (2016b). Value lexicality. Ungated
Knutsson, S. (2019). Epicurean ideas about pleasure, pain, good and bad. Ungated
Knutsson, S. (2021a). The World Destruction Argument. Inquiry, 64(10), pp. 1004-1023. Ungated
Knutsson, S. (2021b). Many-valued Logic and Sequence Arguments in Value Theory. Synthese, 199, pp. 10793-10825. Ungated
Knutsson, S. (2022). Undisturbedness as the hedonic ceiling. Ungated
Leighton, J. (2011). The Battle for Compassion: Ethics in an Apathetic Universe. Algora.
MacAskill, W. et al. (2020). Moral Uncertainty. Oxford University Press. Ungated
Mendola, J. (1990). An Ordinal Modification of Classical Utilitarianism. Erkenntnis, 33, pp. 73-88.
Ryder, R. (2001). Painism: A Modern Morality. Open Gate Press.
Schopenhauer, A. (1819/1909). The World as Will and Representation. Kegan Paul, Trench, Trübner & Co.
Schopenhauer, A. (1851/1973). Essays and Aphorisms. Penguin.
Tomasik, B. (2015). A Small Mechanical Turk Survey on Ethics and Animal Welfare. Ungated
Vinding, M. (2020). Suffering-Focused Ethics: Defense and Implications. Ratio Ethica. Ungated
Vinding, M. (2021). Comparing repugnant conclusions. Ungated
Vinding, M (2022). Lexicality between mild discomfort and unbearable suffering: A variety of possible views. Ungated
- The pinprick in this thought experiment should ideally be replaced with something like a “micro pain”, cf. the various pitfalls of focusing on stimuli rather than experiential states.[↩]
- A proponent of CU may opt for a lexical version of CU instead, but such a view is hardly more plausible if it still allows torment to be created for the sake of bliss.[↩]
- One might object to the name of this thought experiment, since CU is not merely trading torture for micro pleasures in this case, but also for intense pleasure. However, it is still the case that the micro pleasures are what ends up tipping the scale such that CU favors torturing everyone. Besides, this is hardly a fundamental point of contention by the lights of CU (at least in its non-lexical forms), since one could in any case reframe the thought experiment such that CU would torture everyone purely for the sake of micro pleasures — by replacing the intense pleasure with micro pleasures and by shortening the torture by a sufficient duration that is in turn replaced by micro pleasures.[↩]
- Note, however, that the informal survey mentioned above tentatively suggests that only a small minority of people would choose B2, perhaps around 1 percent or less, Tomasik, 2015.[↩]
- Such putative non-hedonic bads would presumably all be granted some degree of “expected disvalue” or “choice-relevance” by someone who favors CU within a framework of moral uncertainty (cf. MacAskill et al., 2020), which means that there is a significant parallel between views that integrate CU within a moral uncertainty framework and views that entail lexicality between different bads at the object-level. That is, while CU would entail a lexical value difference between the mildest hedonic state and any non-hedonic state, where the latter has absolutely zero value, this zero-to-one difference between hedonic and non-hedonic states would not persist within an all-things-considered choice-worthiness framework if one assigns non-zero credence to non-hedonistic views. Hence, if one prefers to maintain value lexicality between hedonic and non-hedonic bads (as opposed to accepting tradeoff ratios between them or the like), one would also end up with lexicality between value entities that each have non-zero “expected” disvalue. And if lexicality between different bads is a critical problem for views that entail such lexicality directly at the object-level, it would presumably also be a problem for views that arrive at the same conclusion within a moral uncertainty framework.[↩]
- Gustafsson also raises objections against what he calls Weak NU (Gustafsson, 2022, pp. 7-8). These objections are similar to the ones I have addressed here in my reply to Toby Ord’s “Why I’m Not a Negative Utilitarian”.[↩]