The problem of possible populations: animal farming, sustainability, extinction and the repugnant conclusion

by Stijn23 min read6th Jul 20219 comments


Person-affecting viewsPopulation ethicsMoral philosophy


I propose a new population ethical theory: person-affecting excised total utilitarianism. According to this theory, we should choose the option or state of the world that maximizes the sum of expected welfares of everyone who exists or will exist in the future, except for those possible people who do not exist in all states that are available at the moment of decision, when these possible people have a low positive welfare between zero and a positive threshold value that can be democratically and autonomously chosen by those people who do exist in all available states. When a currently optimal option becomes suboptimal in the future, that currently optimal option should be excluded from the set of all available options.

This theory can avoid the counter-intuitive Very Repugnant Conclusion, Sadistic Conclusion and Extinction Conclusion that infect many other population ethical theories. As practical consequences, it says that we should not choose animal farming, not even when the farm animals have a positive welfare, and we should take some reasonable sustainability measures and climate actions to guarantee sufficiently high welfare levels of future generations.


A personal note

Population ethics is terribly difficult. My personal journey started ten years ago with a very simple theory called ‘average utilitarianism’. This theory is simple enough to express in one line: choose the option that maximizes the average welfare of the whole population. But it has many flaws, so I switched to ‘rank-discounted utilitarianism’ around 2012 (more complicated, but explained here). After coming up with some objections (some explained here), I changed my mind again and defended an even more complex ‘positive number-dampened power mean prioritarianism with negative total utilitarianism’ in my PhD-dissertation of 2014. Two years later, I realized that this theory was a big mistake, as it also involved very counter-intuitive implications (such as a reverse sadistic repugnant conclusion). So I moved on and made a stopover at the ‘minimum complaint theory’, which another two years later, in 2018, grew into a more simple but ambiguous ‘variable critical level utilitarianism’. Added to this, a few months ago (in 2021), I started to appreciate another, more complex ‘minimax net-complaint theory’. It seems that I change my mind about population ethical theories every two years. I would have given it up a long time ago, if population ethics were not so important.

Alas, when one wants to do the most good in the world and tackle the most relevant and important real-life decisions involving animals and future generations, population ethics is crucial. The moral value of possible populations, of individuals whose existence depends on our choices, is a tough nut to crack. And considering the huge numbers of farmed and wild animals and the huge number of people who could exist in the long-term future, the stakes are extremely high. Every month, humans are breeding (and killing) as many farm animals as the current worldwide human population. Humans are also causing species extinction, which means innumerable future wild animals are no longer born. And humans create new dangers that could with a non-zero probability cause human extinction. Quadrillions of future people’s lives are at stake. How bad is it when we make a decision that causes the non-existence of a number of possible people larger than any number you can realistically think of? Sure, those 100000000000000000000000000000000 non-born people cannot complain against our decision, because they will not exist as a result of our decision, but does that make our decision morally ok? What if they all could have had extremely satisfying lives? 

Considering this importance, past June I participated at a University of Oxford Global Priorities Institute conference programme. Thinking, discussing and reading about population ethical theories for another full month. At the final days of that conference, I realized that perhaps a rather simple population ethical theory (which for the moment I give a more difficult name ‘person-affecting excised total utilitarianism’) could be good enough to deal with the crucial big decisions involving possible populations.


The new theory: person-affecting excised total utilitarianism

We can start with possibly the simplest population ethical theory: total utilitarianism. Choose the option that maximizes the total welfare of the whole population. This total welfare is the sum of the welfares of every individual who exists or will exist in the future. (How to measure welfare or utility is another difficult question. Here we can assume that welfare is a properly weighted aggregate of satisfaction minus dissatisfaction of preferences and positive minus negative experiences over a lifespan. These qualitative aspects are properly aggregated such that the welfare of a person in a situation measures how strongly that person prefers that situation and everything about that situation relative to a hypothetical reference situation where everything is the same, except that person has no subjective experiences and behaves the same way but completely unconsciously.)

There is a serious issue with total utilitarianism: it often comes down to… total sacrifice. An example is the Very Repugnant Conclusion. In an initial state, a group of people is extremely happy, having an extremely high welfare. These people can totally sacrifice themselves, becoming extremely miserable (with very negative welfare), by giving birth to or breeding huge numbers of humans and animals. All these extra people only exist in this second state, and they all have lives barely worth living, with positive but very low welfare. As long as the number of extra people is large enough, the total welfare in the second state is higher than in the initial state. For example, the initial state has a billion people at a very high welfare 10, whereas the second state has those billion people at welfare minus 10, plus a trillion people at low but positive welfare 1. The total welfare in the second state is 99 times higher than in the initial state. Therefore, a ‘totalist’ would say that the second state is better, although in that state there are extremely miserable people and no-one is very happy.

The basic reason for my drifting from theory to theory, is trying to avoid this Very Repugnant Conclusion of total utilitarianism. That is not easy, because alternative theories, from average utilitarianism to minimax net-complaint theory, also have counter-intuitive implications.

A simple way to avoid the Very Repugnant Conclusion, is first making a distinction between necessary and possible people. If we make decisions, we can only choose between a limited number of options. Each option results in a different state of the world, i.e. a different future involving different people having different welfares. Necessary people are those people who exist in all available states of the world, whereas possible people do not exist in at least some available states. Second, we can apply total utilitarianism, but simply exclude all possible people from the sum of welfares. This easily avoids the Very Repugnant conclusion: the extra population of lives barely worth living are excluded from the calculation, such that the second state gets a reduced total welfare of minus 10 billion.

This reduced total utilitarianism is a ‘person-affecting theory’, which says that we should make (necessary) people happy, instead of making (possible) happy people. Person-affecting means that if a state is better (or worse) than another, than there should be at least one existing person for whom that state is better (or worse). If in the Very Repugnant Conclusion scenario a totalist says that the initial state is worse than the second state, the person-affecting theorist replies that one should then indicate at least one existing person in the initial state for whom that state is worse. As there does not exist such a person, the person-affecting theorists concludes that the initial state is better (because it is better for the necessary people).

Simply excluding all possible people is not a good idea. It makes the theory too dictatorial: only the necessary people count and can determine what is good. And the theory generates a Sadistic Conclusion, in which it is good to create people with extremely miserable lives. Consider factory farming of chickens. There is a rather strong consensus amongst farm animal welfare experts, animal ethics experts and lay people that many chickens in intensive agriculture have a negative welfare. At this moment, we can decide to breed extra chickens in the future. These future chickens are possible individuals: they do not exist if we would choose for example a vegan world. Not taking into account their negative welfare, merely because they are at this moment possible people, is extremely sadistic: meat-eaters who like chicken meat can become a little bit happier, at the cost of a huge amount of extra suffering experienced by the billions of chickens. 

To avoid this Sadistic Conclusion, let us only exclude the welfare of possible people who have a positive welfare. But now we face an anti-natalist Extinction Conclusion: global extinction of all sentience becomes the best option. A fraction of future humans can have lives not worth living, with more negative than positive experiences. And the lives of wild animals are likely more miserable, meaning a higher probability of negative welfare. If only possible people with a negative welfare count, the future can only be bad (or at best neutral, when no future person has a negative welfare). Of course, the necessary people, who cannot procreate and have to destroy the world, would suffer, but the total suffering of the huge numbers of possible future lives with negative welfare can easily be higher than the total suffering of the necessary people. This means that total extinction would be the best decision.

We can remain more faithful towards total utilitarianism and still avoid the Very Repugnant Conclusion, as well as the Sadistic Conclusion and the Extinction Conclusion, by only excluding the possible people who have welfare levels between zero and some positive threshold level. Possible people with negative welfare or sufficiently high welfare are still included in the totalist sum of welfares. After all, the Very Repugnant Conclusion clearly referred to possible people having lives barely worth living, i.e. with low positive welfare levels. So it is sufficient to only exclude those lives. Person-affecting excised total utilitarianism, that excludes (cuts out or excises) the welfare of possible people with lives barely worth living (having a low positive welfare), is the theory closest to total utilitarianism that uses a person-affecting adaptation to avoid the Very Repugnant Conclusion.

Person-affecting excised total utilitarianism introduces one arbitrariness: how high should the threshold be? To avoid unwanted arbitrariness (where the arbitrariness cannot consistently be  wanted by at least one person), I suggest that the necessary people (at least those who are cognitively capable) can individually and autonomously choose the threshold value (just as in my older theory of variable critical level utilitarianism people can autonomously choose critical levels). To make it democratic, we can simply take the average value of everyone’s chosen threshold. (One could also take a weighted average, where the chosen thresholds are weighted by how certain people feel about their chosen thresholds. Someone who feels uncertain about the best threshold or feels indifferent between different threshold values, can chose a satisfying threshold that gets a low weight in the weighted average.) 

This threshold value can also depend on the set of available states at the moment of decision. If for example the available states result in a Very Repugnant Conclusion when a low threshold is chosen, the necessary people can temporarily choose a higher threshold. In other contexts, a high threshold can result in an unwanted Extinction Conclusion, which means a lower threshold can be chosen. If the decision involves a choice between available states that generate both the Very Repugnant Conclusion and the Extinction Conclusion, the necessary people have to choose which is the least bad outcome and choose their preferred threshold values appropriately. Every necessary person can choose a preferred threshold value at every decision point in time. This chosen threshold value represents or captures population ethical preferences of that person, such as the preference to avoid the Very Repugnant Conclusion or the Extinction Conclusion. 

As a possible person does not always exist, and a non-existing person does not have any preferences (let alone population ethical preferences), it is permissible to exclude possible people from the democratic choice of threshold value. The possible people cannot complain against this exclusion, because either the possible person will not come into existence, or when the choice of the threshold value is made such that it causes the existence of that possible person, the welfare of that person was only excluded when it is positive. A positive welfare means that the person has no reason to complain against the choice of threshold value: another chosen threshold could result in the non-existence of that person.


Some technical issues (optional reading)

Before we move to concrete applications, some technical issues of this person-affecting excised total utilitarianism need to be addressed. The reader who is not interested in these technicalities, can skip this section and jump to the applications below.

First, we have to avoid dynamic inconsistency. Assume that the threshold level is 2 units of welfare: if a possible person has welfare between 0 and 2, that welfare doesn’t count. Suppose in an initial, single-person state, there is one person with high welfare 10. This person can make two consecutive decisions. The first decision is to bring into existence a second person. If the second person is brought into existence, a second decision determines how well-off both people become. Choosing the low-welfare two-person state, the first person gets a low welfare 4,5 and the newborn, second person receives welfare 6, which is higher than the threshold value of 2. In the alternative high-welfare two-person state, the first person gets welfare 9,9 and the newborn person gets a low welfare 1. Now look at the sum of welfares in each state, excluding the possible people with welfare below 2. The initial, single-person state has total welfare 10, the low-welfare two-person state has total welfare 4,5+6=10,5, and the high-welfare two-person state has total welfare 9,9 (because the second person is below the threshold and hence excluded). The low-welfare two-person state has the highest total welfare and hence should be selected. 

But, once the second person comes into existence, the initial single-person state is no longer an available option. Only the two two-person states remain. That means the second person also becomes a necessary person, because that person now exists in all available states. Hence, the second decision should be made by considering the second person as a necessary person. Now the high-welfare two-person state receives a total welfare of 9,9+1=10,9. So now this latter state should be chosen. There is an inconsistency, because this state had the lowest total welfare when facing the first decision (i.e. the choice between the three available states). We can play this game again, by introducing the option to cause the existence of a third person. In the low-welfare, three-person state, the first person gets welfare 4, the second remains at welfare 1, and the third person gets 6. The total welfare is 11. The high-welfare, three-person state gives welfare levels 9,8 to the first person and 1 to the other two people. Before the existence of the third person, that person is a possible person with welfare below the threshold, and hence excluded from the sum of welfares. That sum is 10,8, which is lower than the low-welfare, three-person state. However, once the third person exists and becomes a necessary person, the sum of welfares in the high-welfare, three-person state is 11,8. This state should now be selected. We see the welfare of the first person decrease, from 10 to 9,9 to 9,8, by adding people with lives barely worth living (welfare 1). We can play this game 200 times, until the first person gets a welfare minus 10 and 200 people are added with lives barely worth living, and behold: we end up with the Very Repugnant Conclusion.

This shows that it is very difficult to escape the Very Repugnant Conclusion (see this article “Why the repugnant conclusion is inescapable”). But we do not surrender yet. To avoid this Very Repugnant Conclusion, as well as the dynamic inconsistency, we can introduce a simple constraint. If an available state, which initially seems to be the best, is later (when possible people become necessary people) dominated by another alternative state which initially seems worse, the initial better-seeming state should be excluded from the available options of the initial decision. If you know in advance that if you choose the best state, that best state will later no longer be the best state in the future, then you should not choose that best state. In the above example, the low-welfare two-person state should be excluded from the first decision, and hence the one-person state is the remaining best state. We no longer slide towards the Very Repugnant Conclusion.

A second technicality involves situations of uncertainty. When making a decision, we are not sure who will exist in the future and how high the welfare of those people is. It is possible that we have information about probabilities. For example, choosing not to act against climate change means that future generations have some probability to become victim of climate change that results in a lower welfare. It could also mean a (possibly very small) probability of human extinction, which means humans in the far future will not exist. This issue of uncertainty can be dealt with in utilitarian ethics, by taking the expected value of welfare. If you have a 10% probability of getting welfare 6 and a 90% probability of getting welfare 0, your expected welfare equals 0,1 times 6 plus 0,9 times 0, which equals 0,6. A small difficulty arises with possible people, where a decision could mean a 10% probability of existence with welfare 6 and a 90% probability of non-existence. There is no welfare at non-existence, so how to calculate the expected value? In this case, we can simply assume that non-existence means a welfare of 0, such that the expected welfare equals 0,6. If this expected welfare is lower than the chosen threshold value, the welfare is excluded from the sum of welfares. Also note that a necessary person always has a 100% probability of existence. 

Third, we can introduce some nuanced extensions of the theory. We can make the theory ‘lexical’ when there is a tie. Suppose there are multiple optimal states, having equal sum of welfares (excluding the low positive welfares). In that case, we can break the tie by choosing the state that has the highest sum of welfares of the excludes people. Also, we can include the low positive welfares to make up for the small negative welfares, when present. (The low positive welfares cannot make up for the very bad, large negative welfares, because that would generate a weaker version of the Very Repugnant Conclusion.) Once the small negative welfares are made up for, adding more lives barely worth living does not add any value. This is a ‘soft asymmetry’ that can slightly help to avoid the Extinction Conclusion.


Applications: animal farming and climate change

Time for some real-life applications. One of my major problem areas, is animal farming. We can represent this in a simple way as a choice between three states. In the No-breeding state, humans exist (and are the necessary people) and do not cause the existence of farm animals. These humans are moderately happy. The second state Farming involves the breeding of farm animals. Humans can consume animal products such as meat, and hence are a little happier than in No-breeding. In factory farming, the animals likely have a negative welfare, but assume animal farming is possible where the animals have overall a positive welfare. They are (painlessly) prematurely killed to be eaten, so they have a low positive welfare. In the third state Sanctuary, humans cause the existence of animals, but these animals are then raised on an animal sanctuary where they receive a lot of benefits (protection, food, care, a safe and wide area,…) such that the animals are very happy. The humans, on the other hand, no longer receive benefits from consuming animal products, and have to bear the costs of taking care for the animals. So the humans now have a low welfare.

A challenging argument in favor of meat consumption, is the ‘logic of the larder’. It is perhaps the best argument against veganism. If farm animals have a positive welfare, and humans are happier because they can enjoy the taste of meat, we seem to have a win-win where everyone is benefitted, compared to the vegan situation. The happy farm animals are not worse-off than in the vegan world, because without animal farming, these animals would not exist. Hence, if one could only choose between two states, No-breeding and Farming, then Farming would be the best option: it has a higher total welfare, and no-one can complain against choosing that option (except perhaps if the animals are average utilitarians and Farming has a lower average welfare than No-breeding, which means that the animals still prefer non-existence above having a happy life at a farm). 

However, once the third option Sanctuary is available, Farming is unlikely to be the best option. For example, the total welfare in Sanctuary could be higher. A total utilitarian would then say that we have a duty to breed animals and take care of them at animal sanctuaries. This could even result in a Very Repugnant Conclusion: we have to breed a huge number of animals, sacrifice ourselves completely for the sake of those animals, who all may end up with a low welfare.

How does person-affecting excised total utilitarianism deal with this ‘logic of the larder’ animal farming problem? Can or should we eat happy meat of happy farm animals? Let us first consider happy farm animals as happy slaves. The logic of the larder, applied to human slavery, results in the counterintuitive Happy Slavery Conclusion, which says that it is permissible to breed and use human slaves who have a positive welfare. We generally agree that it is not permissible to breed and use happy human slaves. Perhaps if the total welfare with happy slavery is higher than the total welfare when slaves were set free and those slaves are so extremely happy that they themselves want the continued existence of such happy slavery and even prefer their children to be born as slaves, happy slavery could become permissible. But even if human slaves would be overall happy, it is unlikely that they would be so extremely happy in the system of slavery. The same goes for farm animals. We should avoid the Happy Animal Farming Conclusion.

Person-affecting excised total utilitarianism can avoid this Happy Animal Farming Conclusion simply by setting the threshold value slightly higher than the welfare of the farm animals. As the animals in Sanctuary have a higher welfare than the animals in Farming, it is possible that those sanctuary animals have a welfare above the threshold. That means those sanctuary animals are included in the sum of welfares. If the sum of the welfare of humans and animals in Sanctuary is higher than the sum of the welfare of humans in No-breeding, then Sanctuary should be chosen. If total utilitarianism would select Sanctuary, then so would person-affecting excised total utilitarianism. However, it could be that the welfare of animals in Sanctuary is still considered too low. Total utilitarianism would pick Sanctuary as the best option, even if it involves the Very Repugnant Conclusion, as mentioned above. The humans have a lower welfare in Sanctuary than in No-breeding, so they rather prefer No-breeding. These humans are the necessary people and hence are allowed to choose the threshold value. They can choose a threshold value higher than the welfare of the animals in Sanctuary, such that also the sanctuary animals are excluded from the sum of welfares. If the animals are excluded, Animal Farming becomes the optimal state as it generates the highest welfare for the humans. But that state involves the introduction of animals. Now we face a problem: once those animals exist, they become necessary individuals whose welfare counts. That means Sanctuary becomes better than Animal Farming once the decision is made to breed animals. As Animal Farming becomes dominated by another state, we should eliminate the option Animal Farming from the initial decision options, i.e. options available before the decision to breed animals. That means the there are only two permissible options in the initial decision: No-Breeding and Sanctuary. As in the initial decision the animals are excluded, No-breeding is better than Sanctuary. Hence, we end up with the option No-breeding. 

A second, slightly similar problem to the animal farming problem, is climate change. We have three options. First, we can decide not to procreate, such that there will be no future human generations. This is the state No-procreation, but this state is less relevant, because our welfare will be very low when we could no longer procreate. It is more likely that we choose to procreate. Then we can choose between two other states. In No-action, we do not take action to stop climate change. We can be very happy, living luxurious lives, not being bothered by climate change, and using all cheap fossil fuels as we please. In that case, future generations are born, but they will have a low welfare because they face dire climate events. In the third state Climate-action, we take effective action to avoid climate change, which means future generations are very happy, but we face the costs of action, which means we get a lower welfare compared to No-action. 

It is often told that, if we do not take climate action, our grandchildren in the future can reprimand us. They can complain against us, saying that we should have taken climate action in order to avoid those climate catastrophes that they have to face. But this is not correct. When we take climate action, we change our behaviors in such a way that different future generations will be born: none of the people who exist in the future of Climate-action also exist in the future of No-action. This is different from the animal farming problem, in which the animal in Farming is the same animal as the animal in Sanctuary. The same individual is treated differently. In contrast, in the climate change problem, the future person in Climate-action is not the same person as the future person in No-action.

The latter means there is a non-identity problem, which we also encounter in the ethics of reproduction (where one can choose for example between giving birth to a happy child with a disability or another, happier child without that disability). As long as the future people in No-action have a positive welfare, they cannot complain against us not taking action, in the sense that they cannot say that they are harmed by our inaction. If we took climate action, those extreme climate events would not occur, but also those future people would not exist (other people would exist). Only if the future people in No-action had a negative welfare, they could complain against having been personally harmed, because they prefer the state of non-existence, in which they do not suffer.

Even if the low-positive-welfare future people are not personally harmed by our inaction, they still might have a population ethical preference held by many currently living people: that the state Climate-action, in which future generations have a much higher welfare, should be chosen, especially if taking climate action is not too costly for the present generation. This reflects a preference for sustainability: that we should not lower the welfare of future generations too much compared to our own welfare (formulated by the World Commission of Environment and Development in 1987 as “meeting the needs of the present without compromising the ability of future generations to meet their own needs.”).

The person-affecting excised total utilitarianism can easily select the Climate-action state, by setting a threshold value between the average welfare of the future generations in No-action and the average welfare of the future generations in Climate-action. This threshold value of welfare is the sustainability level of welfare. 



9 comments, sorted by Highlighting new comments since Today at 2:46 PM
New Comment

Trying to summarise for my own understanding.

Is the below a reasonable tl;dr?

Total utilitarianism, except you ignore people who satisfy all of:

  • won't definitely  exist
  • Have welfare between 0 and T

Where T is a threshold chosen democratically by them, and lives with positive utility are taken to be "worth living".

If so, does this reduce to total utilitarianism in the case that people would choose not to be ignored if their lives were worth living?

That's a good summary, except that the threshold is chosen democratically by those who definitely exist. If these people choose not to ignore those people who don't definitely exist and have welfare between 0 and T, then it reduces to total utilitarianism

How do you approach identity? If ~no future people are "necessary", does this just reduce to critical-level utilitarianism (but still counting people with negative welfare, can't remember if critical level does that)? Are you ok with that?

My theory would be like critical level utilitarianism, where necessary people, possible people with negative welfare and possible people with high positive welfare have zero critical levels, and possible people with low positive welfare have a critical level equal to their own welfare. So people can have different critical levels, and the critical level might depend on the welfare of the person. 

The problem of identity could become difficult, when we consider identity as something fluid or vague. If for example copying a person (a kind of teleportation but without destroying the source person) would be possible: which of the two copies is the necessary person and which is the possible person? I guess the two copies have to fight over this for themselves. In general: once person A in state X identifies herself with a unique person B in state Y, and B identifies herself with A, only then are persons A and B considered identical. A necessary person is a person who is able to identify himself with a unique person in each other available state. 

Thank you for sharing your thoughts. What do you think of the following scenario?

In world A the risk for an existential catastrophe is fairly low and most currently existing people are happy.

In world B the existential risk is slightly lower. In expectation there will live 100 billion additional people (compared to A) in the far future whose lifes are better than those of the people today. However, this reduction of risk is so costly that most of the currently existing people have miserable lifes. 

 Your theory probably favours option B. Is this intended ?

Yes, my theory favours B, assuming that those 100 billion additional people have on expectation a welfare higher than the threshold, that the higher X-risk in world A does not on expectation decrease the welfare of existing people, and that  the negative welfare in absolute terms of having a miserable life is less than ten times higher than the positive welfare of currently existing people in world A. In that case, the added welfare of those additional people is higher than  the loss of welfare of the current people. In other words: if there are so many extra future people who are so happy, we really should sacrifice a lot in order to generate that outcome. 

However, the question is whether we would set the threshold lower than the welfare of those future people. It is possible that most current people are die-hard person-affecting utilitarians who care only about making people happy instead of making happy people. In that case, when facing a choice between worlds A and B, people may democratically decide to set a very high threshold, which means they prefer world A

Hi Stijn! Thanks for writing this---I completely agree that getting population ethics on surer footing is an important issue for the EA community. And I agree with your diagnosis that it's super difficult.

I'm wondering if there is a similar dominance argument you could apply to John Broome's argument against the "Intuition of Neutrality". Basically, imagine we're in some World A, where Worlds B & C are available and they only differ in that 1 extra person exists with utility within this range of indifference ("neutrality" in Broome's words) in both worlds, however in World C her utility is higher than in B. Standing at the vantage point of A, we're indifferent between B and C (since her utility doesn't count in either); but C is dominated by B in that this new person has a better life and no one is affected. Very curious if layering the 'range of indifference' with a dominance criteria can be shown to always escape conclusions like the one above where C>B when compared to one another, but C~B when compared from A. 

Apologies if that wasn't clear! And perhaps your dynamic consistency problem is analogous... I just didn't see it as immediately obvious and haven't spent enough time thinking through the details. Thanks again for writing such a detailed post on this!

Hi Kevin,

thanks for the comment.  My theory mostly violates that neutrality principle: all else equal, adding a person to the world who has a  negative welfare is bad, adding a person who has a welfare higher than treshold T is good, and in its lexical extension, adding a person with welfare between 0 and threshold T, is good (the lexical extension says that if two states are equally good when it comes to the total welfare excluding the welfare of possible people between 0 and T, then the state that has the highest total welfare, including that of all possible people, is the best).

There is indeed an apparent intransitivity in my theory, which is not a real or serious intransitivity, as it is avoided in the same way as that dynamic inconsistency is avoided, namely by considering the choise sets. So, worlds A, B and C are equally good when you consider the full choice set {A,B,C}, but once that extra person is added, the choice set reduces to  {B,C}, and then C is better than B (the extra person becomes a necessary person in choice set {B,C}). The crucial thing is that the 'better than' relationship depends on the choice set, the set of all available states. This excludes the serious 'money pump' intransitivities. In the full choice set {A,B,C}, I am indifferent between A and B, so I'm willing to switch from A to B. Now I prefer C over B (because that extra person has a higher welfare in C), and hence I'm willing to pay to switch from B to C. But as the choice set is now reduced to {B,C},  after choosing C, I can no longer switch back to A, even if I was initially indifferent between C and A. In the lexical extension of my theory, I would end up with world C.

Thanks! Makes sense.