Conditional interests, asymmetries and EA priorities

by MichaelStJules13 min read21st Oct 201923 comments


The aim of this post is to introduce the concept of conditional interests. What I focus on is the following claim, its justification and its implications, including for EA priorities:

Only Actual Interests: We accomplish no good by creating and then satisfying an interest, all else equal, because interests give us reasons for their satisfaction, not for their existence or satisfaction over their nonexistence.

This could be described as an asymmetric "interest-affecting view", and the procreation asymmetry follows, because individuals who wouldn't otherwise exist have no interests to satisfy. I think such a view accords best with our intuitions about personal tradeoffs.

It therefore (in theory) allows individuals to make personal tradeoffs between experiences of pleasure and suffering as normally understood, unlike strong negative hedonistic utilitarianism, but it also doesn't give reasons to individuals who have no interest in wireheading, Nozick's experience machine or psychoactive drugs to subject themselves to these. If you don't have an interest in (further) pleasure for its own sake at a given moment, you are not mistaken about this, despite the claims of classical utilitarians. As such, while strong negative utilitarianism may, to some, be counterintuitive if it seems to override interests (see responses to this objection by Brian Tomasik and Simon Knutsson), classical utilitarianism effectively does the same, because it can prioritize the creation and satisfaction of new interests (in the same individual or others) over the satisfaction of actual interests. So, in my view, if negative hedonistic utilitarianism (in any form) is wrong because it overrides individual interests, so is classical utilitarianism. However, I think we can do better than both strong negative hedonistic utilitarianism and classical utilitarianism, which is the point of this post.

Furthermore, by seeing value in the creation of new interest holders just to satisfy their interests and sometimes prioritizing this over interests that would exist anyway (in a narrow or wide sense), classical utilitarianism (and any other theory prescribing this) treats interest holders as mere receptacles or vessels for value in a way that Only Actual Interests prevents. There have been different statements of this objection, but I think this is the clearest one.

The claim Only Actual Interests is basically from Johann Frick’s paper and thesis which defend the procreation asymmetry. I recently wrote a post on the forum that referred to his work, but this post considers essentially his approach, its justifications and its implications. Christoph Fehige's antifrustrationism, developed earlier, is also basically the same, but concerned only with preferences, specifically.

In the first section, I give definitions and in the second, I state some claims. In the third section, I list a few basic implications. In the fourth section, I describe the relationship to Buddhist axiology or tranquilism. In the fifth section, I defend the claim, primarily through examples with our common sense understanding of interests. In the sixth section, I consider some other more abstract theoretical implications. In the seventh section, I describe implications for EA priorities, starting from 80,000 Hours' cause analyses; the main conclusion is that existential risks should receive less priority. In the last section, I include some thoughts about the possibility of prioritizing (an individual's) current interests over (their) future ones.


Outcome: The entire actual history of all that is ontological (universe, multiverse, possibly things beyond the physical), past, present and actual future.

Interest, interest holder: An interest is a value held by some holder, the interest holder, and that can be more or less satisfied (according to some total order) so that for the interest holder, it is better that it be more satisfied than less, all else equal.

Note: a value could a priori be an interest with itself or the universe as its holder. We might say the value of pleasure itself has an interest in further pleasure, or the universe has an interest in further pleasure, although I think this is wrong; see the discussion following Only Actually Conscious Interests in the Claims section.

Actual interest: An interest is actual in a given outcome if it is held in that outcome.

Conscious interest: An interest is conscious if its satisfaction or unsatisfaction can be experienced consciously by the holder in some outcome.

Actually conscious interest: An interest is actually conscious in a given outcome if its satisfaction or unsatisfaction is experienced consciously by the holder in that outcome.

Actually conscious interests are actual interests and conscious interests.

Experiential interest: An interest is experiential if its degree of satisfaction is determined solely by the conscious experiences of its holder.

Pleasure: A conscious experience is pleasurable if the experience comes with a conscious interest in itself over its absence for its own sake, and this experience is experienced by the holder of this interest. Pleasure is the conscious experience and the conscious interest of the holder.

Suffering: A conscious experience involves suffering if the experience comes with a conscious interest in its absence over the experience itself, and this experience is experienced by the holder of this interest. Suffering is the conscious experience and the conscious interest of the holder.


This is just a set of claims of interest for this post. I am not actually making all of these claims here.

Experientialism: The only interests that matter are the holders’ interests in their own conscious experiences.

Only Conscious Interests: The only interests that matter are conscious interests.

Hedonism: Experientialism and Only Conscious Interests are true, and specifically, pleasure and suffering are the only kinds of interests that matter.

Hedonism is one of the main claims of hedonistic utilitarianism, including classical utilitarianism.

Negative Hedonism: Experientialism (or Hedonism) and Only Conscious Interests are true, and specifically, suffering is the only kind of interest that matters.

Negative Hedonism is one of the main claims of (strong) negative hedonistic utilitarianism.

Now, the main claim of this post, restated:

Only Actual Interests: Interests provide reasons for their further satisfaction, but not their existence or satisfaction over their nonexistence.

In particular, an interest is neither satisfied nor unsatisfied in an outcome if it does not occur in that outcome, and this outcome is not worse than one in which the interest occurs, all else equal. (I don't say that only actual interests matter, since that's either confusing or inaccurate.)

You could call this an "interest-affecting view", and this could be interpreted in a narrow or a wide way. Under a narrow view, we wouldn't compare the degree of satisfaction of different interests in different outcomes, only the degree of satisfaction of the same interests common to different outcomes. I'm not sure if such a view can be made both transitive and independent of irrelevant alternatives, although we might also reject these requirements in the first place.

Under a wide view, we might say that it's better for interest to exist and be satisfied to degree than for interest to exist and be satisfied to degree , especially if and are interests of the same kind (e.g. both are interests in pleasure, both are interests in not suffering, both are interests in gaining knowledge), and , so that would be satisfied to a greater degree than .

See the nonidentity problem for some discussion of person-affecting views, in which the interests at stake are the wellbeing for two different people, exactly one of whom will be born. Should we prefer for a person with a better life to be born than a person with a worse life, even if they would be different people? If yes, we should reject a purely narrow person-affecting view.

We might also want to restrict further to interests that are held presently or in the actual future, excluding past interests.

Only Actually Conscious Interests: Interests provide reasons for their further satisfaction when they are consciously experienced and through their conscious experience by their holders, but they don’t provide reasons for their existence or satisfaction over their nonexistence.

In particular, an interest is neither satisfied nor unsatisfied in an outcome if the interest (or its satisfaction/unsatisfaction) is not experienced, and this outcome is not worse than one in which the interest (or its satisfaction/unsatisfaction) is experienced, all else equal. As such, the holders of the interests may as well be the conscious experiences themselves. Or, the universe as a whole may be the holder of conscious interests, but the interests are (in practice, not a priori) localized: in locations where there are no conscious experiences, there are no conscious interests to be satisfied. To illustrate, what you normally understand to be my conscious interests for my own conscious states are interests that are directed at conscious states in the parts of space that will be occupied by what can vaguely be defined as my body. If "I" didn't experience this interest, neither would the universe as a whole.

Some basic implications

In this section, I list a few simple implications of Only Actual Interests.

1. That I could induce a craving in someone and then satisfy it is not a reason for me to actually do so.

2. That someone would have their interests satisfied or be happy or have a good life is not a reason to bring them into existence, although there may be other reasons. That they will almost certainly have some unsatisfied interests is a reason to not bring them into existence, but the reasons to do so could be stronger in practice. This is the procreation asymmetry. In particular, you would never be required to sacrifice your own wellbeing (even into hedonic negative) to bring new individuals into existence, all else equal, although, again, all else is rarely equal.

3. If someone has no interest in psychoactive substances (or, specifically, the resulting experiences), that they might enjoy them is not on its own a reason to try to convince them to take them.

Buddhist axiology or tranquilism

If we combine Only Actually Conscious Interests with Experientialism or even Hedonism, we don’t actually get Negative Hedonism (one of the main claims of negative hedonistic utilitarianism). Pleasure and suffering can both matter, and I am not claiming a conscious interest in further pleasure is necessarily an instance of suffering as I’ve defined these terms, but that the absence of this conscious interest is never in itself bad, while its unsatisfaction is. So, if someone has an unsatisfied interest in further pleasure, this is worse than not having this interest at all, even if they are happy overall, and it's also worse than not increasing their pleasure to satisfy this interest.

This does lead to an asymmetry between pleasure and suffering, but not one that does not count pleasure at all:

Asymmetry between pleasure and suffering: In the absence of an interest in further pleasure, there's no reason to increase pleasure, but suffering by its very definition implies an interest in its absence, so there is a reason to prevent it.

This is effectively Buddhist axiology or tranquilism, framed slightly differently.

In particular, if you’re a utilitarian who also accepts Only Actually Conscious Interests and Experientialism (or Hedonism), you’re basically a negative preference utilitarian who cares only about the conscious satisfaction/unsatisfaction of preferences about conscious experiences. This can include the conscious preference for more pleasure.

Negative preference utilitarians see the unsatisfaction of a preference as worse than its nonexistence, and the complete satisfaction of a preference no better than its nonexistence. If preferences are interests, with Only Actual Interests, they could potentially provide reasons for their satisfaction, but not their existence.

Why Only Actual Interests?

Hedonistic consequentialists defend something like or as strong as Only Conscious Interests. If we can convince them of Only Actual Interests, then they should accept Only Actually Conscious Interests. Or, we can convince them directly of Only Actually Conscious Interests.

We might try to shift the burden of proof: If, according to Only Conscious Interests, an interest matters only if it can be experienced consciously, why should it matter (i.e. detract from an outcome) if it is not actually experienced at all?

However, rather than just shifting the burden of proof, we can defend Only Actual Interests by analogy and generally (not just to those who accept Hedonism), based on the examples Frick gives (the first two), and one more of my own:

1. That you've made a promise to someone is a reason to keep the promise, but the fact that you could keep a promise is not in itself a reason to make it in the first place. Promises provide reasons to be kept, but not reasons to be made.

2. That you have the gear and other means necessary to climb mount Everest successfully doesn't give you a reason to actually do it; you must already (or either way, expect to) have an interest in doing it.

3. That I could induce someone to want and then buy a product (e.g. through marketing) is not a reason for me to actually do so.

I find it pretty intuitive that this is how interests should work. The claims Frick defends are slightly more general than mine in this post, using "normative standard" instead of "interest" and "bearers" instead of "holders". The rejection of the Transfer Thesis for each interest F is basically equivalent to a claim similar to Only Actual Interests:

No Transfer: Interests provide reasons for their further satisfaction, but neither an interest nor its satisfaction provides reasons for the existence of that interest's holder over its nonexistence.

Transfer Thesis: If there is reason to increase the extent to which F is instantiated amongst existing potential bearers, there is also reason to increase the extent to which F is instantiated by creating new bearers of F.

People value a lot of things, but it doesn't seem like these things justify the existence of people themselves. Is a world worse for not having value X if no one is around to miss it? If not, why would adding people just to achieve X do any good? Take X to be any value from that list, "Abundance, achievement, adventure, affiliation, altruism, apatheia, art, asceticism, austerity, autarky, authority, autonomy, beauty, benevolence, bodily integrity, challenge, collective property, commemoration, communism, community, compassion, competence, competition, competitiveness, complexity, comradery, conscientiousness, consciousness, contentment, cooperation, courage ..."

Other theoretical implications

1. The point of a hedonium shockwave, if any, would be to eliminate otherwise unsatisfied interests, not to create happiness. The prevention of future interests by destruction could be a good thing, generally. However, both are wildly speculative, and there are good consequentialist and nonconsequentialist reasons to not pursue either, e.g. for consequentialist ones, moral cooperation and trade, given how much opposition there would be to both. Value may also be more complex than Hedonism allows.

2. It avoids the Repugnant Conclusion, which is a consequence of classical utilitarianism. The repugnant conclusion is basically that for the following populations in which everyone has a life worth living,

the following relations hold

and so by transitivity and the independence of irrelevant alternatives,

The first step from to (Mere Addition) supposedly follows because the extra lives are worth living so adding them can't make the situation worse; the step from to (Non-Anti-Egalitarianism) supposedly follows if we ensure the average welfare is high enough and we aren't so anti-egalitarian that we think a total loss to the best off individuals can only be made up for by a much larger total gain for the worst off individuals, and the last step actually doesn't do anything, since and are identical populations (the division is only illustrative).

Then, doing the same by adding a very large population of lives barely worth living in instead of what's shown there, and would be very flat and close to 0. So, a very large population of lives barely worth living would be better than a small population of very good lives, e.g. here:

For some defenses of the Repugnant Conclusion, see "In defense of repugnance" by Michael Huemer.

The Repugnant Conclusion is avoided in two possibly different ways:

a. It denies the premise that there could be positive existences.

b. Assuming there are positive existences (using a different measure of value than one based on conditional interest satisfaction), adding a population of lives barely worth living is in fact bad, so is false.

To further defend this last point, under a wide view, the step from to of adding a population of lives barely worth living is equivalent to also making everyone in swap places with extra individuals from of the same number. That is, starting from , we make everyone in as badly off as would be the extra individuals in , and add extra individuals with the same wellbeing as the originals in and even more with the lower level. From the point of view of the original individuals in , this could make worse, and adding the extra individuals would not compensate, because they have no interests in being brought into existence.

3. By denying the possibility of positive existences, it avoids Arrhenius's major impossibility theorems like this one (I think this is probably the strongest statement, since it only assumes ordinal, but interpersonally comparable, welfare). Alternatively, if we do interpret positive existence using in hedonistic terms or terms preferences for continued life, then it violates Non-Sadism and implies the Sadistic Conclusion: it can be worse to add a population of positive existences than one of negative existences (usually with a much larger population of positive existences). In response, the Sadistic Conclusion might not be so bad, at least compared to the Repugnant Conclusion, and even more plausibly so since we've already accepted the procreation asymmetry and rejected Mere Addition. (Aside: Arrhenius was a negative utilitarian of some kind in the past; I don't know if he still is.)

4. Whether death is good or bad in itself (ignoring effects on others, which we should not ignore) depends on the nature of interests we count and how we count them. If we accept Only Actually Conscious Interests, then death would be good in itself (again, ignoring effects on others). If we prioritize an individual's current interests over their future ones, then their interests in continuing to live would be given greater weight, see the last section for some thoughts on this.

Implications for EA priorities

In this section, I describe how we might rerank the different cause areas in 80,000 Hours' list here.

1. We should reject Bostrom's astronomical waste argument and give less priority to preventing extinction. That does not mean we have no reasons to care about the (far) future or prevent extinction, but the fact that future humans who would not otherwise exist would be happy (rather than not exist, or fewer of them exist) is not a reason for intervention. This significantly reduces the value of working to prevent existential risks, although they may still be very important, if we think our continued existence would be sufficiently helpful in expectation to, say, wild animals (if they would also continue to exist after our extinction), or aliens. If you don't think extinction is much worse than almost everyone dying, you can see how 80,000 Hours' tool reranks the cause areas. Assuming you answer the previous questions in a way to not cause reranking (although you may very well disagree with the underlying assumptions), answering "(C) Not more than twice as bad" to question 4 reranks the list as follows:

Re-ranked list:
1. Global priorities research - 26
2. Promoting effective altruism - 25 ⇩ (-1 point)
3. Risks posed by artificial intelligence - 23.5 ⇩ (-3.5 points)
4. Factory farming - 23
5. Health in poor countries - 21
6. Reducing tobacco use in the developing world - 20
7. Nuclear security - 20 ⇩ (-3 points)
8. Land use reform - 20
9. Biosecurity - 20 ⇩ (-3 points)
10. Climate change (extreme risks) - 18 ⇩ (-2 points)

Question 4 is

Question 4: Here’s two scenarios:
A nuclear war kills 90% of the human population, but we rebuild and civilization eventually recovers.
A nuclear war kills 100% of the human population and no people live in the future.
How much worse is the second scenario?

If you want to avoid reranking before question 4, you should answer 1. (A), 2. (A) and 3. (B).

Note that AI risk remains above both Factory farming and Health in poor countries. The far future can still indeed be overwhelmingly important, and we may expect AI to shape it even if we don't go extinct. Furthermore, if we do go extinct, that is a lot of early deaths. However, they didn't provide options which go further than "(C) Not more than twice as bad", and your other answers to other questions can influence the rankings. Even if we ignore future generations, it might be better for everyone to go extinct than for 90% of the population to die out, because the surviving 10% may have very bad lives in such an outcome.

Furthermore, if the probability of extinction is around 1% or less (80,000 Hours' best guess seems to be 1-15% in the next 50 years, according to question 3., with answer (B)), then the non-existential risk causes should go up in priority, since there's a greater chance that the work we do for those causes isn't wasted. E.g. ending factory farming and then us going extinct immediately after isn't much better for factory farmed animals than us just going extinct, because factory farming will end anyway if we do go extinct (although we're likely to achieve considerable progress and prevent a lot of suffering up until extinction if we do work on factory farming).

2. There's also a question of the degree to which death is bad. If death seems less bad, then this could further reduce the priority to existential risks. This might also have an effect on the value of some, but not all global health and poverty interventions. If death is bad, I think its badness is unlikely to be roughly proportional to the number of years of life lost, since existing interests are likely to change for many people as they age, but GiveWell doesn't explicitly use such a measure anymore, anyway (see here and here), and I don't know to what degree analysts rely on such an intuition. With Experientialism or Hedonism, death in itself is not bad, but the process of dying and the impacts on loved ones are of course often very bad, perhaps especially an unexpected early death (but if early and later deaths are equally bad, then postponing death doesn't look very good in hedonistic terms). Overall, I don't think global health and poverty as a cause area would necessarily look worse since many of the best interventions do not derive most of their value from life extension. Existential risk cause areas would probably look worse if we thought before that the badness of extinction came primarily from early deaths (and astronomical waste).

3. Global health and poverty interventions might decrease the rate of population growth, and this might be in itself good. Family planning interventions and education in developing countries might look better than otherwise, specifically, for this reason.

4. We should reject the logic of the larder. That is, if animals bred and used for human purposes would have good lives, this is not a reason to breed and use them in the first place, and the fact that they will almost certainly have unsatisfied interests is a reason to not do so. There could be other reasons for their breeding and use, but they need to be even stronger.

Prioritizing current interests over future ones?

I've been wondering lately if there's a plausible consequentialist theory (or more generally, theory of value) which assigns more value to the immediate satisfaction of an individual's current interests over the satisfaction of their future ones in such a way as to be compatible with common sense notions of non-paternalism and consent. In this way, could violating an individual's interests now to better satisfy their own future interests be usually bad in itself? I think this would still be compatible with our understanding of impartiality. We could just use some kind of discounting of interests within individuals, but I'm not sure if this quite does it.

However, if we don't think people's personal identities persist over time (which seems likely to me), this wouldn't mean much. If we don't think they persist, nothing can be paternalistic, since there would be no personal tradeoffs, only interpersonal tradeoffs.

If we also give greater weight to current interests than to future interests generally, not just within individuals, our theory could also look much more deontological in practice, but it would be harder to call this impartial, since it gives less weight to the interests of future people. It might also be difficult to ground from an impartial perspective, because of the relativity of simultaneity. There's no such physical obstacle for personal tradeoffs because we can use an individual's own frame of reference.

If we're not careful, there might be issues with dynamic consistency: the decisions that look best now could systematically look worse in the future, and you would regret them, even with perfect certainty ahead of time.