Why I’m Not a Negative Utilitarian
A Review of Toby Ord’s Essay
B.Contestabile First version 2016 Last version 2020
Table of Contents
2.1 Absolute and Lexical NU
2.2 Weak NU
2.3 Moderate NU
2.4 Lexical Threshold NU
2.5 Theoretical versus Practical Priority
Appendix: Terminology Summary
In his essay Why I’m Not a Negative Utilitarian, Toby Ord shows himself surprised to see that some of his friends and acquaintances in the effective altruism community identify as Negative Utilitarians, although negative utilitarianism is discarded in mainstream philosophical circles.
Type of problem
Is negative utilitarianism (NU) an implausible theory?
Versions of NU which overrule compensation within the same person (like Absolute NU, Lexical NU and Weak NU) could be called implausible theories, but not versions of NU which question compensation across different persons (like the moderate NU).
The moderate NU is a metric within hedonistic utilitarianism, which is functionally equivalent to prioritarianism.
- Prioritarianism does not exclude negative totals, but is usually associated with positive totals.
- The moderate NU, conversely, does not exclude positive totals, but NU is usually associated with negative totals.
The rationality of a negative total is investigated in The Denial of the World from an Impartial View.
From a strictly hedonistic perspective a world without humans is preferable to a world with negative total well-being. But the non-existence of humans is no realistic option. We only have a choice between more or less suffering populations. If there is a life-friendly way to turn the total positive (e.g. transhumanism) then NU is not hostile to the existence of life.
The totalitarian potential is not a peculiarity of NU, but a characteristic of consequentialism in general. Adherents of NU, who recognize totalitarianism as a problem, amend their theory with human rights. Such an anti-totalitarian form of NU corresponds to a political party or movement within a democratic system.
From a NU point of view the utilitarian life evaluations have to be reformed. Almost all OECD countries now use a life evaluation on a 0 to 10 rating scale [Helliwell, 15], which is interpreted as a linear point scale. Examples: OECD Better Life Index, Satisfaction with Life Index, Where-to-be-born Index, World Happiness Report. Linear scales don’t account for the asymmetry between happiness and suffering.
For more information about NU indices see Negative Utilitarianism and Justice.
In his essay Why I’m Not a Negative Utilitarian, Toby Ord shows himself surprised to see that some of his friends and acquaintances in the effective altruism community identify as Negative Utilitarians, although negative utilitarianism (NU) is discarded in mainstream philosophical circles.
Type of problem
Is negative utilitarianism an implausible theory?
Negative utilitarianism (NU) is an umbrella term for all types of utilitarianism which model the asymmetry between suffering and happiness [Fricke, 14]. For more information about this definition see Negative Utilitarianism and Justice.
Toby Ord defines these versions as follows:
Only suffering counts.
The Absolute NU corresponds to the original NU as defined by R.N.Smart; see Negative Utilitarianism.
Since happiness has no moral value, no state of affairs can be better than non-existence.
The term Absolute NU can be misleading. It does not mean “absolute priority of the minimization of suffering over the maximization of happiness”. Since happiness has no moral value, there is no maximization of happiness and accordingly no question of priorities.
Suffering and happiness both count, but no amount of happiness (regardless of how great)
can outweigh any amount of suffering (no matter how small).
The minimization of suffering has an absolute priority over the maximization of happiness.
First comes the eradication of suffering, then the promotion of happiness.
In both Absolute and Lexical NU the The Pinprick Argument applies. It is counter-intuitive, however, to deny the possibility of compensations/trade-offs in cases like the birth of a child, cosmetic surgery etc. [Fricke, 16] and it is absurd in the case of a pinprick. Intra-personal compensation is reason enough to refute these versions of NU. If we consider the compensation across different persons the picture doesn’t get better.
Toby Ord dedicates a considerable part of his paper to the refutation of the Absolute and Lexical NU. But these versions of NU are simply misinterpretations of Popper’s notes on ethics; see Negative Utilitarianism and Justice.
Toby Ord’s definition of the Weak NU goes as follows:
Suffering and happiness both count, but suffering counts more. There is an exchange rate between suffering and happiness or perhaps some nonlinear function which shows how much happiness would be required to outweigh any given amount of suffering [Ord].
The term suffering counts more distinguishes the Weak NU from classical utilitarianism. Toby Ord applies his definition to intra-personal compensations.
This diagram was taken from Why I’m Not a Negative Utilitarian
But what is the horizontal scale supposed to represent? There is no obvious natural unit of suffering or happiness to use. It might be possible to have a consistent scale in the happiness direction and a separate consistent scale in the suffering direction, but it is very unclear how they are both supposed to be on the same scale. This is what would be needed for Weak NU to be a coherent theory and for the diagram to make any sense [Ord].
The incoherence argument
Psychometrics confirms that positive and negative affect carry different information and need to be separately measured and analyzed [Diener]. Toby Ord speaks of an incoherence argument against the Weak NU. But “separately measured” does not mean that the above horizontal axis becomes meaningless:
- Happiness and suffering are comparable; otherwise compensations/trade-offs would be impossible. If happiness and suffering are comparable, then they can be arranged on the same scale.
- The scale of happiness and the scale of suffering touch each other, at the point where suffering and happiness are zero.
Obviously there is no incoherence or discontinuity in a mathematical sense. The horizontal axis is not a peculiarity of NU, it is borrowed from classical utilitarianism. The problem is of a different nature:
- Classical utilitarianism assumes that suffering and happiness have intrinsic moral value. But given the same event, these values can be different for each individual. Also the aggregation function, which says if (and how) suffering is compensated by happiness, can be different for each individual (Furthermore the function is dynamic, but we don’t need to elaborate on that for the purpose of this paper).
- The Weak NU assigns moral value by the two (green) linear utility functions. The values of suffering and happiness have to be modified accordingly, before they enter the aggregation function. This modification, however, has unpleasant consequences. Following an example:
The worse-for-everyone argument
A person is watching a film and feels hungry. Should he/she leave the cinema and appease hunger or watch the end of the film? The person comes to the conclusion, that net well-being is positive, if he/she stays in the cinema. However, the Weak NU would multiply the values of happiness and suffering with a weighting factor and come to a different conclusion:
For example, in some cases the Weak NU will say that it is immoral to watch the end of the film while you are really hungry, even if this tradeoff increases your wellbeing, because the suffering counts more morally [Ord].
In other words:
- the personal evaluation asks for staying in the cinema because this option increases net well-being
- the Weak NU overrides the personal evaluation and asks for leaving the cinema.
Toby Ord speaks of a worse-for-everyone argument against all versions of NU. That is not entirely correct.
- In the Absolute NU watching the end of the film is worse for everyone, because happiness (watching the film) is completely devaluated. Only suffering (being hungry) counts.
- In the Weak NU, however, the weighting function could be such that the net well-being is positive for some persons (those who suffer less from being hungry, and who feel more pleasure watching the film).
Decisive is the fact that the Weak NU overrides individual valuations in completely implausible cases. This is reason enough to refute this version. A similar refutation can be found in Fabian Fricke’s article [Fricke, 14-16]. The Weak NU is theoretically weak: nomen est omen.
Risk-aversion versus risk-neutrality
So far we assumed that suffering and happiness are known in taking a decision. The general case, however, is a comparison of opportunities and risks. Could the Weak NU be defended by introducing the concept of risk-aversion? Following an example with probabilities:
Suppose you live in New York City and are offered two jobs at the same time. One is a tedious and badly paid job in New York City itself, while the other is a very interesting and well-paid job in Chicago. But the catch is that, if you wanted the Chicago job, you would have to take a plane from New York to Chicago (e.g. because this job would have to be taken up the very next day). Therefore there would be a very small but positive probability that you might be killed in a plane accident) [Angner, 5]
Let us assume the various outcomes can be evaluated with the following numbers [Angner, 25]:
Take job in New York
Take job in Chicago
Now, the rational course of action depends on the probabilities assigned to the two relevant events. If the agent has determinate probabilities over the two events, it is easy to confirm that she should take the job in Chicago so long as the probability of a plane crash is less than one percent, and stay in New York if it greater than one percent; if the probability is exactly one percent, she is indifferent, and is rationally permitted to take either job [Angner, 13].
not crash 99%
Take job in New York
1 x 0.01 = 0.01
1 x 0.99 = 0.99
0.01 + 0.99 = +1.0
Take job in Chicago
-890 x 0.01 = -8.9
10 x 0.99 = 9.9
-8.9 + 9.9 = +1.0
Harsanyi uses a similar example to show that high risk-aversion is irrational in everyday situations [Harsanyi]:
▪ The majority assigns e.g. -89 to the plane crash (instead of -890) and therefore takes the job in Chicago. Harsanyi associates this behavior with risk-neutrality, utility maximization and rationality.
▪ A minority assigns e.g. -8900 to the plane crash (instead of -890) and therefore – with the same assumptions about probabilities – denies the job in Chicago. Harsanyi associates this behavior with high risk-aversion and irrationality.
But possibly the minority is just more sensitive for suffering. From the external perspective the minority’s behavior seems to be risk-averse, but from the inner perspective it is risk-neutral. The minority feels suffering stronger and therefore acts as rational as the majority. To override individual attitudes towards risk is as questionable as overriding the individual evaluation in the cinema example. The Weak NU is an implausible version of NU.
The better-for-everyone argument
We have excluded Absolute NU, Lexical NU and the Weak NU because they override intra-personal compensations/trade-offs.
There is a reverse case, however, where classical utilitarianism overrides intra-personal evaluation. The worse-for-everyone argument against NU has a mirror image in the better-for-everyone argument against classical utilitarianism. The following diagram refers to the QALY axiology, which is used in hospitals (Quality-adjusted life year, Wikipedia):
The vertical axis does not contain negative values. Death has value zero, so that it is better for everyone to be alive. Voluntary Euthanasia is necessarily immoral, because it destroys positive value instead of eliminating negative value. NU, in contrast, asks for an axiology which allows negative values. It maintains that overriding the personal evaluation in the hospital is as questionable as overriding the personal evaluation in the cinema example above. If a person feels that his/her suffering cannot be compensated with happiness [Fricke, 18], then NU assigns a negative well-being to this person and not – like most QALY models – a low positive value. The situation is then characterized by a choice between the two evils suffering and death. For the purpose of NU a scale which does not know negative numbers (e.g. a point scale from 1 to 10) has to be converted into a signed scale. In the following we presuppose a scale with positive and negative values.
So far we assumed that everyone strives to maximize his/her individual utility. But maximizing individual utility may not be the best strategy for maximizing the societies’ total utility. In above example it may e.g. be better for the person to take the job in Chicago, but worse for the global climate, because of the environmental impact of aviation. Morally right – in a strict consequentialist sense – is only the action which maximizes the societies’ total utility. For that reason we shift the focus from the individual level to the public policy level.
The moderate NU is a metric within hedonistic utilitarianism, which assigns a higher weight to the avoidance of suffering than to the promotion of happiness. The moral weight of suffering can be increased by using a "compassionate” metric, so that the result is the same as in prioritarianism.
Hedonistic utilitarianism is associated with the work of John Broome in this paper:
- The term moderate means that happiness has (some) moral value.
- The hedonistic scale contains negative values [Broome 2004, 213].
For a comparison of Broome’s utilitarianism with prioritarianism see Antinatalism and the Minimization of Suffering.
If the moderate NU is “only a metric” within hedonistic utilitarianism: What is the reason for using a special term for this metric?
The reason is that utilitarianism – similar to prioritarianism – is associated with positive totals:
▪ In classical utilitarianism the hedonistic scale is linear and symmetric. The theory does not exclude negative totals, but in practice most utilitarians assume that – given the current state of affairs – total well-being is positive.
▪ In the moderate NU the hedonistic scale is non-linear and asymmetric. The theory does not exclude positive totals, but it considers that – given the current state of affairs – total well-being might be negative. The intuition that “global suffering cannot be compensated by happiness” turns global well-being negative, so that the maximization of happiness turns into a minimization of suffering.
metric 1 metric 2
classical utilitarian negative utilitarian
Intuitions with regard to global well-being are controversial:
▪ The majority – in particular prioritarians and classical utilitarians – think that the happy majority can outweigh the suffering minority (Fig.1 metric 1)
▪ Buddhists, Gnostics, Schopenhauer, and antinatalists do not share this intuition (Fig.1 metric 2)
The majority considers the latter people to be highly risk-averse or even irrational, but possibly the majorities’ perception of risk is distorted.
The rationality of the negative utilitarian intuition is investigated in The Denial of the World from an Impartial View.
There is a well-known moral intuition that we should prioritise helping the worse off and this is much more widely accepted than NU (…). Such an intuition can be accounted for in the theories of Prioritarianism, Egalitarianism, and Sufficientarianism [Ord].
The moderate NU is functionally equivalent with Prioritarianism, so that Ord’s objection does not apply.
For more information about the moderate NU see Negative Utilitarianism and Justice.
(1) Suffering and happiness both count, but suffering can only be compensated up to a certain threshold.
The problem with this version is that we cannot define a threshold (quantity times quality of suffering), which is valid for everyone. The capability to compensate suffering depends on the constitution, the environment and the biography of each individual.
The idea of a threshold makes more sense in the context of inter-personal compensation (compensation between different persons). In this context the term suffering means uncompensated suffering [Fricke, 18] respectively negative welfare. In an ideal case the data come from surveys and focus groups [Broome 2004, 261], but in cases of severe suffering this kind of data collection is often impossible. We don’t know under what circumstances the welfare of a specific person turns negative, but we can declare a certain magnitude of negative welfare as a threshold for inter-personal compensation. Following the example used by Toby Ord:
(2) The Lexical Threshold NU stems from a deep sense of compassion at the sheer scale and intensity of suffering in the world. No amount of happiness or fun enjoyed by some organisms can notionally justify (outweigh) the indescribable horrors of Auschwitz [Pearce].
The term compassion in (2) clarifies that this definition refers to the compensation across different persons. The horrors of Auschwitz consist of a large number of people with negative welfare. The example also illustrates that in cases of extreme suffering the negative welfare has to be estimated and cannot be collected by means of surveys.
Comparison with the moderate NU
The moderate NU agrees with (2) that the compensation of suffering by the happiness of other people is limited. But instead of defining a threshold it maintains that – depending on the kind of suffering – the happiness needed in order to compensate suffering may be immense. At some point the compensation becomes virtually impossible so that the result is the same as in the Lexical Threshold NU. The solution without a threshold has the advantage that Ord’s continuity argument [Ord] does not apply.
In Fig.3 the notion of a maximum refers to the point scales of surveys [World Happiness Report].
- metric 3 is designed in such a way that a majority of happy persons cannot outweigh a minority of suffering persons.
- metric 4 is designed in such a way that a vast majority of maximally happy persons cannot outweigh a single maximally suffering person.
For a numerical example see Negative Utilitarianism and Justice, chapter Implementation.
metric 3 metric 4
White squares = happy persons
Shaded squares = suffering persons
The metrics 1, 2, 3 and 4 (Fig.1 and Fig.3) illustrate an increasing reluctance to compensate suffering by other people’s happiness. In the moderate NU it is possible to continuously downgrade happiness and upgrade suffering, in analogy to prioritarianism. In particular, if the kind of suffering matches the horrors of Auschwitz, then the metric can be designed in such a way, that it satisfies Pearce’s Lexical Threshold intuition (i.e. that compensation becomes impossible). Since the metrics are flexible enough to satisfy the Lexical Threshold intuition in all practical cases, the moderate NU solves Tomasik’s problem with the three inconsistent intuitions.
The rationality of compensation
Pearce’s intuition about the horrors of Auschwitz is challenged by Ord as follows:
I don't agree about whether these quantities of suffering, vast though they are, can be outweighed by the positive side of human experience [Ord].
Who is closer to a realistic view? For an answer see The Denial of the World from an Impartial View.
Toby Ord uses two definitions for the practical priority of (the avoidance of) suffering:
Weak Practically-Negative Utilitarianism = Classical Utilitarianism with the empirical belief that it in many common cases it is more effective to focus on alleviating suffering than on promoting happiness [Ord].
Does it make sense to use the term negative utilitarianism for an empirical belief or a cost-benefit analysis within the classical utilitarian axiology? It probably adds to clarity, if we reserve the term negative utilitarianism for the application of an asymmetric scale, as in the moderate NU (chapter 2.3). An asymmetric scale constitutes a theoretical priority of (the avoidance of) suffering [Fricke, 14].
Strong Practically-Negative Utilitarianism = Classical Utilitarianism with the empirical belief that suffering outweighs happiness in all or most human lives [Ord].
The term “in …human lives” (as opposed to “across …human lives”) makes clear that this definition addresses the compensation of suffering and happiness within the same person. Again, it is rather confusing to use the term negative utilitarianism for an empirical belief within the classical utilitarian axiology. The Weak NU (as described in chapter 2.2) systematically increases the weight of suffering, and decreases the value of happiness within an intra-personal compensation and is therefore clearly different from classical utilitarianism.
Another potential cause for confusion is the interrelation between NU and population ethics. You may think that a clear case for NU is comparing the alleviation of present suffering with the creation of new happy lives. However, this case brings in changing populations, which is the topic of population ethics. The intuition might not be to do with suffering and happiness per se, but to do with the lack of value in creating new lives in comparison to improving existing ones. If this is what guides you, then you should consider views such as Presentism or Critical Level Utilitarianism, which deny that adding a happy life is good, while agreeing that adding happiness to existing people is good and even that preserving the life of existing people is good [Ord].
The idea that adding a life with positive welfare does not make a population better (also called Prior existence utilitarianism or Asymmetry) has a theoretical deficiency [Arrhenius, 137] and is not supported by the moderate NU.
The negative utilitarian view – according to which the present global suffering cannot be compensated by happiness – has the following consequences:
1. From a strictly hedonistic perspective a world without humans is preferable to a world with negative total well-being. However, as David Pearce emphasizes, the non-existence of humans is no realistic option. We only have a choice between more or less suffering populations. If there is a life-friendly way out of suffering (e.g. transhumanism), then the moderate NU is not hostile to the existence of life. In the current situation, however, a reduction of the population probably reduces global suffering; see Antinatalism and the Minimization of Suffering.
2. From a NU point of view the utilitarian life evaluations have to be reformed. Almost all OECD countries now use a life evaluation on a 0 to 10 rating scale [Helliwell, 15], which is interpreted as a linear point scale. Examples: OECD Better Life Index, Satisfaction with Life Index, Where-to-be-born Index, World Happiness Report. Linear scales don’t account for the asymmetry between happiness and suffering.
For more information about NU indices see Negative Utilitarianism and Justice.
Many advocates of NU claim that on average human lives have net negative intrinsic moral value (…) This implies that much healthcare and lifesaving is of enormous negative value. It says that the best healthcare system is typically the one that saves as few lives as possible, eliminating all the suffering at once. [Ord]
- There is an empirical argument against this scenario. A health care system with above characteristics would provoke an immense distress and eventually produce more suffering than it pretends to avoid. One could argue that empirical arguments against moral killing are not convincing under all circumstances, but similar deficiencies can be found in classical utilitarianism and prioritarianism, see the section about Human rights and the higher purpose in Negative Utilitarianism and Justice.
- Hospitals often use an axiology which does not know negative well-being, the so-called qalys (see chapter 2.3). In this case we are confronted with a reversed kind of theoretical deficiency. A horribly suffering patient cannot escape from his/her situation by voluntary euthanasia, because non-existence is defined as worst case.
Adherents of consequentialism, who recognize the totalitarian potential as a problem, amend their theory with human rights. An anti-totalitarian health care system (in the spirit of Karl Popper) respects the individual’s will to extend lifetime, as well as the right to palliative care and voluntary euthanasia in well-defined and controlled situations which prevent abuse.
There is also a type of example that is phrased in terms of whether it would be right or wrong to create a utopia if the very foundation of that utopia required the forced suffering of the innocent during its construction [Ord].
Again, this kind of conflict can be found in all consequentialist ethics. Strictly speaking one would have to distinguish between anti-totalitarian and totalitarian negative utilitarians, prioritarians, classical utilitarians etc., where the anti-totalitarian respect human rights. However, if we talk – like Toby Ord – about negative utilitarians within the community of effective altruism, then we do not need this specification, because we know that this community respects human rights.
I would like to thank Brian Tomasik and Simon Knutsson for their helpful comments and suggestions during the review of this paper.
1. Angner Erik (2002), Revisiting Rawls: A Theory of Justice in the light of Levi’s theory of decision, Final version (2004), Theoria 70(1), pp.3-21, Wiley-Blackwell, New Jersey
2. Arrhenius Gustav (2000), Future Generations, A Challenge for Moral Theory, FD-Diss., Uppsala University, Dept. of Philosopy, Uppsala: University Printers
3. Broome John (1991), Weighing Goods, Oxford: Basil Blackwell, Paperback version 1995
4. Broome John (2004), Weighing Lives, Oxford University Press, Paperback version 2006
5. Chao Roger (2012), Negative Average Preference Utilitarianism, Journal of Philosophy of Life, Vol.2, No.1, p.55-66
6. Diener, Ed, and Robert A. Emmons. 1984. “The Independence of Positive and Negative Affect.” Journal of Personality and Social Psychology 47: 1105–1117.
7. Fricke Fabian (2002), Verschiedene Versionen des negativen Utilitarismus, Kriterion Nr.15, pp.13-27
8. Harsanyi John C. (1975), Can the Maximin Principle Serve as a Basis for Morality? A Critique of John Rawls’ Theory, American Political Science Review 69: 594-606.
10. Helliwell, John, Richard Layard, and Jeffrey Sachs, eds. (2015), World Happiness Report. New Nork, The Earth Institute, Columbia University Press.
9. Hirata Johannes (2004), Happiness and Economics, Discussion Papers of the Institute for Business Ethics No.99, University of St.Gallen, Switzerland
10. Ord Toby (2013), Why I’m Not a Negative Utilitarian
11. Pearce David, Negative Utilitarianism: Why be negative?
12. Popper Karl R.(1945) The Open Society and its Enemies, Volume I / II, Routledge Classics, London, UK, 2003
13. World Happiness Report (2013), Edited by John Helliwell, Richard Layard and Jeffrey Sachs, The Earth Institute, Columbia University Press, New York
The paper Why I’m Not a Negative Utilitarian does not discuss all versions of negative utilitarianism.
In our review and in the following table we adopt Toby Ord’s terminology as far as possible.
Priority of suffering
ordered by decreasing moral value of happiness
Public policy level 5)
Relative priority of suffering
Happiness has moral value, but less than suffering
Conditional priority of suffering
Happiness has moral value, if the suffering beyond a certain threshold is eradicated
Lexical Threshold NU 1)
with intra-personal threshold
Lexical Threshold NU 2)
with inter-personal threshold
Absolute priority of suffering
Happiness has moral value, if suffering is eradicated
Negative Total Utilitarianism 2)
with happiness as second moral priority
Non-existence as best possible state of affairs
Happiness has no moral value
Original or Absolute NU
Negative Total Utilitarianism or
Classical NU 3)
Non-existence as best possible state of affairs
Preference-satisfaction has no moral value
Negative Preference Utilitarianism4)
with life satisfaction as
highest order preference
1) An intra-personal threshold is the kind of suffering, which makes the individual compensation by happiness impossible. Negative welfare is equal to uncompensated suffering by definition [Fricke, 18].
2) Negative total utilitarianism (with happiness as second priority) can be seen as Lexical Threshold NU, where the inter-personal threshold equals negative welfare. In this case a single person with negative welfare is sufficient to make the compensation with other people’s happiness impossible. A different possible threshold is mentioned in chapter 2.4.
3) The term “classical NU” emphasizes the hedonistic view, as opposed to the preference-based view [Chao, 58]. The term “negative total utilitarianism” emphasizes the public policy view and the difference to average negative utilitarianism.
4) In Negative Preference Utilitarianism perfect life satisfaction has zero moral value; imperfect life satisfaction has negative moral value;
in Negative Total Utilitarianism positive welfare has zero moral value, negative welfare has negative moral value.
5) Note that the terms happiness and suffering have a different meaning on the individual and public policy level. On the public policy level the term happiness refers to life satisfaction, quality of life, well-being or positive welfare. For information about NU on the public policy level see Negative Utilitarianism and Justice.