by Katja Grace (source)
[Epistemic status: sloppy thoughts not informed by the literature. Hoping actual population ethicists might show up and correct me or point me to whoever has already thought about something like this better.]
Person-affecting views say that when you are summing up the value in different possible worlds, you should ignore people who only exist in one of those worlds. This is based on something like the following intuitions:
- World A can only be better than world B insofar as it is better for someone.
- World A can’t be better than world B for Alice, if Alice exists in world A but not world B.
The further-fact view says that after learning all physical facts about Alice and Alice’—such as whether Alice’ was the physical result of Alice waiting for five seconds, or is a brain upload of Alice, or is what came out of a replicating machine on Mars after Alice walked in on Earth, or remembers being Alice—there is still a further meaningful question of whether Alice and Alice’ are the same person.
I take the further-fact view to be wrong (or at least Derek Parfit does, and I think we agree the differences between Derek Parfit and I have been overstated). Thinking that the further-fact view is wrong seems to be a common position among intellectuals (e.g. 87% among philosophers).
If the further-fact view is wrong, then the what we have is a whole lot of different person-moments, with various relationships to one another, which for pragmatic reasons we like to group into clusters called ‘people’. There are different ways we could define the people, and no real answer to which definition is right. This works out pretty well in our world, but you can imagine other worlds (or futures of our world) where the clusters are much more ambiguous, and different definitions of ‘person’ make a big difference, or where the concept is not actually useful.
Person-affecting views seem to make pretty central use of the concept ‘person’. If we don’t accept the further-fact view, and do want to accept a person-affecting view, what would that mean? I can think of several options:
- How good different worlds are depends strongly on which definition of ‘person’ you choose (which person moments you choose to cluster together), but this is a somewhat arbitrary pragmatic choice
- There is some correct definition of ‘person’ for the purpose of ethics (i.e. there is some relation between person moments that makes different person moments in the future ethically relevant by virtue of having that connection to a present person moment)
- Different person-moments are more or less closely connected in ways, and a person-affecting view should actually have a sliding scale of importance for different person-moments
Before considering these options, I want to revisit the second reason for adopting a person-affecting view: If Alice exists in world A and not in world B, then Alice can’t be made better off by world A existing rather than world B. Whether this premise is true seems to depend on how ‘a world being better for Alice’ works. Some things we might measure would go one way, and some would go the other. For instance, we could imagine it being analogous to:
- Alice painting more paintings. If Alice painted three paintings in world A, and doesn’t exist in world B, I think most people would say that Alice painted more paintings in world A than in world B. And more clearly, that world A has more paintings than world B, even if we insist that a world can’t have more paintings without somebody in particular having painted more paintings. Relatedly, there are many things people do where the sentence ‘If Alice didn’t exist, she wouldn’t have X’.
- Alice having painted more paintings per year. If Alice painted one painting every thirty years in world A, and didn’t exist in world B, in world B the number of paintings per year is undefined, and so incomparable to ‘one per thirty years’.
Suppose that person-affecting view advocates are right, and the worth of one’s life is more like 2). You just can’t compare the worth of Alice’s life in two worlds where she only exists in one of them. Then can you compare person-moments? What if the same ‘person’ exists in two possible worlds, but consists of different person-moments?
Compare world A and world C, which both contain Alice, but in world C Alice makes different choices as a teenager, and becomes a fighter pilot instead of a computer scientist. It turns out that she is not well suited to it, and finds piloting pretty unsatisfying. If Alice_t1A is different from Alice_t1C, can we say that world A is better than world C, in virtue of Alice’s experiences? Each relevant person-moment only exists in one of the worlds, so how can they benefit?
I see several possible responses:
- No we can’t. We should have person-moment affecting views.
- Things can’t be better or worse for person-moments, only for entire people, holistically across their lives, so the question is meaningless. (Or relatedly, how good a thing is for a person is not a function of how good it is for their person-moments, and it is how good it is for the person that matters).
- Yes, there is some difference between people and person moments, which means that person-moments can benefit without existing in worlds that they are benefitting relative to, but people cannot.
The second possibility seems to involve accepting the second view above: that there is some correct definition of ‘person’ that is larger than a person moment, and fundamental to ethics – something like the further-fact view. This sounds kind of bad to me. And the third view doesn’t seem very tempting without some idea of an actual difference between persons and person-moments.
So maybe the person-moment affecting view looks most promising. Let us review what it would have to look like. For one thing, the only comparable person moments are the ones that are the same. And since they are the same, there is no point bringing about one instead of the other. So there is never reason to bring about a person-moment for its own benefit. Which sounds like it might really limit the things that are worth intentionally doing. Isn’t making myself happy in three seconds just bringing about a happy person moment rather than a different sad person moment?
Is everything just equally good on this view? I don’t think so, as long as you are something like a preference utilitarian: person-moments can have preferences over other person-moments. Suppose that Alice_t0A and Alice_t0C are the same, and Alice_t1A and Alice_t1C are different. And suppose that Alice_t0 wants Alice_t1 to be a computer scientist. Then world A is better than world C for Alice_t0, and so better overall. That is, person-moments can benefit from things, as long as they don’t know at the time that they have benefited.
I think an interesting feature of this view is that all value seems to come from meddling preferences. It is never directly good that there is joy in the world for instance, it is just good because somebody wants somebody else to experience joy, and that desire was satisfied. If they had instead wished for a future person-moment to be tortured, and this was granted, then this world would apparently be just as good.
So, things that are never directly valuable in this world:
- Someone getting what they want and also knowing about it
- Anything that isn’t a meddling preference
On the upside, since person-moments often care about future person-moments within the same person, we do perhaps get back to something closer to the original person-affecting view. There is often reason to bring about or benefit a person moment for the benefit of previous person moments in the history of the same person, who for instance wants to ‘live a long and happy life’. My guess after thinking about this very briefly is that in practice it would end up looking like the ‘moderate’ person-affecting views, in which people who currently exist get more weight than people who will be brought into existence, but not infinitely more weight. People who exist now mostly want to continue existing, and to have good lives in the future, and they care less, but some, about different people in the future.
So, if you want to accept a person-affecting view and not a further-fact view, the options seem to me to be something like these:
- Person-moments can benefit without having an otherworldly counterpart, even though people cannot. Which is to say, only person-moments that are part of the same ‘person’ in different worlds can benefit from their existence. ‘Person’ here is either an arbitrary pragmatic definition choice, or some more fundamental ethically relevant version of the concept that we could perhaps discover.
- Benefits accrue to persons, not person-moments. In particular, benefits to persons are not a function of the benefits to their constituent person-moments. Where ‘person’ is again either a somewhat arbitrary choice of definition, or a more fundamental concept.
- A sliding scale of ethical relevance of different person-moments, based on how narrow a definition of ‘person’ unites them with any currently existing person-moments. Along with some story about why, given that you can apparently compare all of them, you are still weighting some less, on grounds that they are incomparable.
- Person-moment affecting views
None of these sound very good to me, but nor do person-affecting views in general, so maybe I’m the wrong audience. I had thought person-moment affecting views were almost a reductio, but a close friend says he thought they were the obvious reasonable view, so I am curious to hear others’ takes.
An interesting thing to point out here is that what Katja describes as the further-fact view is terminologically equivalent to what we here call Closed Individualism (cf. Ontological Qualia). This is the common-sense view that you start existing when you are born and stop existing when you die (which also has soul-based variants with possible pre-birth and post-death existence). This view is not very philosophically tenable because it presupposes that there is an enduring metaphysical ego distinct for every person. And yet, the vast majority of people still hold strongly to Closed Individualism. In some sense, in the article Katja tries to rescue the common-sense aspect of Closed Individualism in the context of ethics. That is, by trying to steel-man the common-sense notion that people (rather than moments of experience) are the relevant units for morality while also negating further-fact views, you provide reasons to keep using Closed Individualism as an intuition-pump in ethics (if only for pragmatic reasons). In general, I consider this kind of discussions to be a very fruitful endeavor as they approach ethics by touching upon the key parameters that matter fundamentally: identity, value, and counterfactuals.
As you may gather from pieces such as Wireheading Done Right and The Universal Plot, at Qualia Computing we tend to think the most coherent ethical system arises when we take as a premise that the relevant moral agents are “moments of experience”. Contra Person-affecting views, we don’t think it is meaningless to say that a given world is better than another one if not everyone in the first world is also in the second one. On the contrary – it really does not matter who lives in a given world. What matters is the raw subjective quality of the experiences in such worlds. If it is meaningless to ask “who is experiencing Alice’s experiences now?” once you know all the physical facts, then moral weight must be encoded in such physical facts alone. In turn, it could certainly happen then that the narrative aspect of an experience may turn out to be irrelevant for determining the intrinsic value of a given experience. People’s self-narratives may certainly have important instrumental uses, but at their core they don’t make it to the list of things that intrinsically matter (unlike, say, avoiding suffering).
A helpful philosophical move that we have found adds a lot of clarity here is to analyze the problem in terms of Open Individualism. That is, assume that we are all one consciousness and take it from there. If so, then the probability that you are a given person would be weighted by the amount of consciousness (or number of moments of experience, depending) that such person experiences throughout his or her life. You are everyone in this view, but you can only be each person one at a time from their own limited points of view. So there is a sensible way of weighting the importance of each person, and this is a function of the amount of time you spend being him or her (and normalize by the amount of consciousness that person experiences, in case that is variable across individuals).
If consciousness emerges victorious in its war against pure replicators, then it would make sense that the main theory of identity people would hold by default would be Open Individualism. After all, it is only Open Individualism that aligns individual incentives and the total wellbeing of all moments of experience throughout the universe.
That said, in principle, it could turn out that Open Individualism is not needed to maximize conscious value – that while it may be useful instrumentally to align the existing living intelligences towards a common consciousness-centric goal (e.g. eliminating suffering, building a harmonic society, etc.), in the long run we may find that ontological qualia (the aspect of our experience that we use to represent the nature of reality, including our beliefs about personal identity) has no intrinsic value. Why bother experiencing heaven in the form of a mixture of 95% bliss and 5% ‘a sense of knowing that we are all one’, if you can instead just experience 100% pure bliss?
At the ethical limit, anything that is not perfectly blissful might end up being thought of as a distraction from the cosmic telos of universal wellbeing.