Let us derive the moral theory that includes every valid thing (principle, law, value) in normative ethics.
Welfare as the only valid intrinsic value
Consider ecocentric values, such as biodiversity, integrity, stability or naturalness of an ecosystem. The ecosystem itself is not capable of experiencing the loss of those values, and does not have a subjective preference for those values. Ecosystems do not care about their biodiversity or integrity. Therefore, those ecocentric values are not valid intrinsic values. The same goes for aesthetic values such as beauty (a work of art is not able to care about beauty) and communal and communitarian values (a group of individuals is not an entity that is able to value something).
Consider on the other hand a pathocentric value, namely the welfare of a sentient being. The welfare of a sentient being at a specific moment (or shortest time interval that the sentient being can experience) in a specific situation measures how strongly that being subjectively prefers that situation at that moment, i.e. how it feels about the situation overall. Welfare measures how good a situation is for an individual, according to that individual. That overall preference is the composition of everything that the individual consciously or unconsciously prefers at that moment. Subjective preferences are preferences that can be experienced consciously, where consciousness refers to ‘what it’s like’. A conscious or sentient being knows what it’s like to have a preference. Hence, only sentient beings are capable of experiencing and valuing their own welfare. This welfare is a valid intrinsic value, because there is always someone, namely the sentient being itself, who values its welfare. By definition of a sentient being and the condition that a valid intrinsic value is always valued by the value-holder, welfare is the only valid intrinsic value.
The welfare of a sentient being in a situation is positive when it prefers that situation over a situation where that sentient being does not exist and all other sentient beings keep their welfare levels (i.e. all else equal). It is negative when the sentient being prefers the situation where it does not exist and all else is equal.
Interpersonally comparing welfare
No objective method is known yet to compare the welfare levels of different individuals. For the time being, the best procedure to compare welfare, goes as follows. Take a representative group of moral agents who are able to estimate (using science and empathy) the welfare of others. Those moral agents consider the welfare of a momentaneous mind of a sentient being in a given situation (i.e. its preference in the shortest time interval that can be experienced by that individual) as well as the highest welfare that any momentaneous mind could possibly experience in any of the possible situations that one could choose (i.e. the strongest preference experienced by a momentaneous mind in its most preferred situation). The relative welfare of a sentient being at that moment in time in that situation, as estimated by one of the moral agents, is the best estimate of the actual welfare of that individual according to that moral agent, divided by the best estimate of the maximum achievable welfare of any individual according to that moral agent. As it is a ratio of welfare values, the units of relative welfare are dimensionless, which makes relative welfare interpersonally comparable. Henceforth, the momentaneous mind of a sentient being shall be called an individual for short, and ‘everyone’ will refer to all momentaneous minds of all sentient beings that exist in the past, present and future of the situation.
As there are many moral agents in the representative group, to avoid unwanted arbitrariness, one can take the average of their relative welfare estimates, averaged over all moral agents in the group. Hence, the interpersonally comparable welfare of an individual is the average of the relative welfare levels of that individual as estimated by the moral agents in that representative group. This procedure to derive the interpersonally comparable welfare is the most valid one, because it avoids all avoidable unwanted arbitrariness: no moral agents are arbitrarily excluded from that representative group, all sentient beings can consistently want that procedure, and any further unwanted arbitrariness is unavoidable. Therefore, this interpersonally comparable welfare is the most objective and valid measure of the welfare of an individual.
For more on the interpersonal comparison of welfare, see “Interpersonal welfare comparisons and the veil of ignorance.”
Aggregating welfare
The moral theory of everything uses an aggregate of the intrinsic values (the welfare of individuals) to find the best situation. The best situation is the one that has the highest value of this aggregate welfare. Three conditions determine the validity if the chosen aggregate. First, one should avoid all avoidable unwanted arbitrariness in choosing the aggregate function of welfare. Second, the aggregate should be sensitive to Pareto improvements in someone’s welfare (at least when this condition does not lead to inconsistencies): one situation is always better than another situation that has the same population, when in the first situation at least someone has a higher welfare and no-one has a lower welfare than in the second situation. Third, aggregate welfare should be as closely linked to the individually experienced welfare of the sentient beings.
Due to the interpersonal comparability of welfare (on a common cardinal scale), it is possible to aggregate the interpersonal comparable welfare (henceforth ‘welfare’ for short). To avoid avoidable unwanted arbitrariness, the welfare of each individual, should be equally weighted in the aggregate.
One possibility is taking the product of welfare over all individuals, but as soon as one individual has a zero welfare, the product becomes zero as well. As in this case the aggregate welfare is no longer sensitive to the welfare values and welfare improvements of the other sentient beings, the Pareto condition will be violated.
Another possibility is taking an expression that includes the sum of a function of everyone’s welfare. According to prioritarian welfarism that prefers a more equal distribution of welfare by giving priority to the worst-off, one takes the sum of a concave function of welfare. But a sentient being directly experiences its welfare and hence does not experience that concave transformation of its welfare. As the aggregate welfare should be as closely linked to the individually experienced welfare of the sentient beings in order to be valid, taking such function of welfare in the aggregate welfare is not allowed. The only permissible function is the identity function.
The same goes for average welfarism, where the aggregate is the average of the welfare of each individual. This average equals the sum of all welfare values divided by the total number of individuals who exist, existed in the past and will exist in the future of the given situation. But no sentient being experiences its welfare being divided by the number of individuals.
And the same goes for critical level welfarism, where the aggregate is the sum of everyone’s relative welfare, where relative welfare is the welfare minus a constant critical level. The aggregate becomes the sum of everyone’s welfare minus a term (the number of individuals times the critical level). That extra term is an intrinsic (dis)value, but it is not valid because no-one experiences and values that extra term.
The only remaining valid aggregate is simply the sum of everyone’s welfare (where everyone means all momentaneous minds of all sentient beings).
Demandingness
Maximizing the sum of everyone’s welfare, as recommended by the moral theory of total welfarism (classical utilitarianism), is highly demanding. It requires large sacrifices of individuals, because individuals have to consider the welfare of everyone else. If all individuals would experience the total welfare (i.e. the sum of their welfare), the individuals would simply prefer to maximize that total welfare. But no-one experiences the sum of everyone’s welfare. An individual’s welfare is a valid intrinsic value, but as no-one experiences total welfare, total welfare is not a valid intrinsic value itself. The only reason to assign moral value to the total welfare, was to avoid unwanted arbitrariness. It would be arbitrary to pick one individual and only consider its welfare, and such arbitrariness is unwanted by the other individuals whose welfare does not count.
As total welfare is not a valid intrinsic value, we are allowed to put constraints on it, or modify it, in a way that does not introduce unwanted arbitrariness. Consider a case of personal sacrifice: one individual can save the lives of several others, at a personal cost (e.g. death). As total welfare would increase, total welfarism dictates that one should sacrifice oneself. A sacrifice means that the individual has to do or undergo something against its will. Of course, the individual can only be sacrificed if that individual exists. If the individual did not exist, the other people could not be saved by that individual. Perhaps there are other individuals who could do the sacrifice and save the people. Consider the group of all individuals who can sacrifice themselves. Then the existence of that group is necessary in order for the people to be saved.
To avoid the sacrificial duties demanded by total welfarism, we can put a constraint on the maximization of total welfare: we should choose the situation that maximizes total welfare, whereby the group of individuals who do or could increase the total welfare (for example by saving many more lives than their own group) by decreasing the total welfare of their own group (i.e. sacrificing at least someone of the group), does not have to exist.
This constraint on total welfarism corresponds with a libertarian deontological ethic that assumes everyone has a deontological right not to be used as a means for someone else’s ends against one’s will. An individual is (1) used as a means (2) against its will, when (1) its existence is necessary to obtain the goal and (2) its welfare decreases.
This right can easily be granted to everything in the universe, including non-sentient objects: those objects do not have a subjective will and hence their deontological right cannot be violated. Only for sentient beings does this right become non-trivial: their deontological right can be violated.
With this deontological right or mere-means principle, many other principles in deontological ethics can be derived: doing versus allowing, negative versus positive duties, partiality in imperfect duties,… (see here for an academic paper on the unification of deontological principles).
The deontological right does not have to be absolute or infinite: it may be possible that when a huge amount of welfare is at stake, we are allowed to use sentient beings as a means against their will. The deontological right may be finite in strength. That strength has an inverse relationship to the demandingness of the moral theory: the stronger the deontological right, the less demanding the moral theory becomes.
With a non-absolute or finite deontological right, we can formulate the constraint as: “whereby the group of individuals who do or could increase the total welfare at most a bit further by decreasing the total welfare of their own group, do not have to exist.” If the “a bit” is small, the deontological right becomes weak. In the limit of “at most nothing”, there is no deontological right and we arrive at maximally demanding pure total welfarism. In the other limit of “at most an infinite amount”, the deontological right becomes infinitely strong and we arrive at a minimally demanding pure libertarian deontological ethic.
For more on avoiding the demandingness of total welfarism, see “Mild welfarism: avoiding the demandingness of total(itarian) welfarism.”
Person-affecting condition
Another problem of total welfarism, which relates to the idea that no-one experiences the total welfare, is that one situation can be worse than another, even if it is not worse for anyone who exists in that situation. An extreme example is the very repugnant conclusion in population ethics: a situation where a group of people are extremely happy is worse (according to total welfarism) than a situation where those people are extremely miserable and a huge number of extra people exist who have lives barely worth living. The latter situation has a higher total welfare, due to the huge number of extra individuals with a slight positive welfare.
To make the moral theory a bit more in line with what individuals prefer and experience, we can introduce a person-affecting restriction: a situation can be improved only if the improvement benefits an actual person who exists in that situation. With such a restriction, the very repugnant conclusion is avoided.
One version of a welfarist theory that respects the person-affecting restriction, is actualism: when choosing the best situation, only the welfare of the individuals who actually exist, existed and will exist in the chosen situation matters. However, in its naïve version, this theory entails some contradictions or paradoxes. It may be possible that, when choosing between situations A, B and C, situation B has the highest welfare of all the individuals who actually exist in A, and C has the lowest welfare of all the individuals who actually exist in A. So B is preferred over A. But once B is chosen, the group of individuals who actually exist might change. Situation B might have a different population than situation A. As a result, it might be possible that situation C is better than B according to the individuals who exist in B. In other words: once the best situation (B) is chosen, another situation (C) becomes better and should be chosen. And this third situation C might be worse than A according to the individuals that exist in A. This is a so called dynamic inconsistency: choosing the best option turns that option in an inferior option.
We can refine the actualist theory to make it consistent. In case of an inconsistency, one is not allowed to switch to a situation that has a higher total welfare by adding extra individuals. Bringing extra people into existence is not allowed as a means to increase total welfare, when this results in the choice of a situation that is worse for the people who exist or will exist in the initial situation. Or in other words, we add a restriction to total welfarism: we should choose the situation that maximizes total welfare, whereby those individuals who could increase the total welfare further by decreasing the total welfare of the group of individuals who do exist, do not exist.
As with the deontological right, this restriction does not have to be absolute. We can say that if total welfare increases a lot, it is permissible to switch to that welfare maximizing situation, even if it is at the cost of a small decrease of total welfare of the group of individuals who exist in the initial, suboptimal situation. Hence, instead of ‘further’, we add ‘at most a bit further’.
For more on population ethical theories such as actualism, see “Population ethics and the veil of ignorance.”
Aggregating welfare by taking the sum of each individual’s welfare, adding a rights based restriction of deontological ethics in order to avoid demandingness, and adding a person-affecting restriction of population ethics in order to avoid repugnant conclusions, gives us the moral theory of everything. In its full formulation, it states:
Choose the option that maximizes the expected total welfare (the sum of each individual’s strength of preference for that option, summed over all individuals who ever exist in that option and weighted by their probability of existence), whereby
- those individuals who could increase the total welfare at most a bit further by decreasing the total welfare of the group of individuals who do exist, do not exist, and
- the group of individuals who do or could increase the total welfare at most a bit further by decreasing the total welfare of their own group, does not have to exist.
Notice a weak symmetry between the first, actualist population ethical restriction and the second, libertarian deontological restriction.
This theory has two free parameters: “a bit” in the first restriction and “a bit” in the second restriction. To avoid unwanted arbitrariness, moral agents have to decide democratically how much “a bit” is in specific choices that they face.
For more on this formulation of the theory of everything, see “The standard model and formalism of normative ethics.”
The problem with all wefare estimates is their inherent imprecision, as well as the oversight that in any altruistic project—such as that of the EA community—the key factor is the motivation of the altruistic agents. Calculating any objectively altruistic action means ignoring the subjective element of the altruistic act.
The only effective calculation is one that prioritizes the motivation of the altruistic agents. Appealing to mere altruistic principles based on a cognitive understanding is insufficient. Hence the twenty-year failure of the EA community in its unsuccessful attempt to rationalize altruism. Altruism cannot be rationalized outside of a cultural conception of the altruistic agent's motivation.
One contribution in this regard is that altruism can be cultivated as a necessary element within the development of a culture of benevolence, which might be sufficiently motivating to generate altruistic acts consistent with conventional welfare estimates.
But this is just one contribution among many others that could be made.