Monday, July 16, 2012

Rawls' original position, potential people, and Pascal's Mugging

tl;dr: If we take possible people into account, even endorsing the Repugnant Conclusion would only provide a negligible chance of getting to exist. So in the Rawlsian original position, they would be concerned with other features of society than population.

Recall political philosopher John Rawls' device of the original position. Disembodied souls must agree to set policies defining the basic structure of a society that they will be incarnated in, without any knowledge pinpointing which of the people in the society they will be. In Rawls' original formulation, the souls all know that they will be present in the society. I would say that given that knowledge, souls voting based on their self-interest would reject Derek Parfit's Repugnant Conclusion and prefer to create a society with few citizens at a high standard of living, rather than many more with a lower standard of living. This would improve the chances of a randomly selected "slot" in society being desirable to a soul.

One can adjust the original position scenario to make it more pronatalist by considering the interests of possible people. On this account the number of souls in the original position is large, and if they create a low-population society, then many souls will not have an opportunity to incarnate at all. For example, say there are 1,000 souls, and they consider whether to create a society with 10 souls each of which has 10 units of welfare, or 1000 souls each of which has 1 unit of welfare. In the first society, the souls would have only a 1% chance of being incarnated, so the expected welfare of a soul would be 0.1 units. In the second society each soul would be guaranteed incarnation and 1 unit of welfare, a better expected outcome. This example makes it seem as though this parliament of possible people would endorse the Repugnant Conclusion.

What would happen if the souls were far more numerous than the potential number of slots in even a populous society? After all, there are more possible patterns of human DNA, let alone histories of brain states, than there are atoms in the observable universe. And if we further differentiate people by their spatiotemporal locations, there is just no way for more than a ludicrously small proportion of the possible people to exist. With choices like this, it no longer seems obvious to me that the parties in the original position would much care about population size.

Even if population size is increased a trillion trillion trillion times, the chance of being incarnated would still resemble Pascal's Mugging, small enough to be negligible for a human with a realistic psychology (better approximated with a bounded rather than unbounded utility function). So long as the souls care at all about any features of the society that are likely to be actually present, those concerns would dominate the ludicrously improbable chance of actually being incarnated there. A desire for the society to include great art, or to have an aesthetically pleasant high standard of living, or to be egalitarian would easily dominate the miniscule expected utility of actually getting to live in it. Souls might value an egalitarian society at 1 unit of utility, and getting to exist in it at 1,000,000,000 units, but the former would dominate.

Basically, the idea here is that if we are taking the concerns of possible people into account, then we can satisfy preferences that do not involve existing themselves for some meaningful portion, but this is not possible for the creation of chances to exist, so we would maximize preference-satisfaction of possible people (given a choice of some definite algorithm for weighting) by focusing on the former at the expense of the latter.

This reasoning would not apply if the way we cared about possible people was that we thought bringing people into being and achieving their goals was good, but not achieving the goals of those who do not exist.

Still, the original position scenario has a lot of intuitive appeal for many people, and this extension seems interesting to me.


Robin Hanson said...

This is a good and interesting point.

Paul Christiano said...

Interesting point. I think I basically buy it as a consequence of applying the original position to possible people. I am skeptical about trying to take some moral intuitions and saying they should come from considering the original position, but letting others stand as "real" values (which then get amped up by considering the original position), so this seems to mostly serve as a reductio against trying to justify aggregative moral intuitions via the original position (this being said from the perspective of a metaethics where we honor intuitions about what we should strive for, and where explicit metaethical reasoning impinges on those intuitions only indirectly).

In particular, it seems incongruous to say that my desire for aggregate happiness comes from considering the original position while my desire for average happiness is some more direct kind of altruism (particularly if we then go on to say that the original position doesn't actually justify aggregative intuitions). Before bringing up average happiness the argument felt more plausible applied to aesthetic/scientific progress etc., but now it seems objectionable everywhere.

Carl said...

Paul, yes there are extra degrees of freedom there to mess around with conclusions. Regarding the mention of average welfare, I was thinking of aesthetics: "this society is more beautiful than Muzak-and-potatoes, despite its smaller size." I do think there is a distinction between "you are obligated to participate in this institution which was best for you ex ante" and "you like the aesthetics of a society that looks like so-and-so."

Peli Grietzer said...

I think part of this argument can be generalized: every population-maximizing principle that appeals to the preference of possible persons to exist faces the problem of explaining why it's not equally valuable to satisfy other preferences of possible persons. What I think is really interesting is that replying that the point is not to satisfy the preference of a possible person but to bring a possible satisfied preference into existence can't work: it's obviously bad to bring into existence a suicidal person whose preference for never existing is exceeded only by her preference that the moon orbit the earth (or some other fact that obtains whether she exists or not), even though bringing her into existence is a net increase in the amount of satisfied preferences in existence.

(I hope this didn't stray too far off-topic by attempting to generalize.)

Carl said...

Peli, your comment is welcome.

Thrasymachus said...

Hello Carl. I've also been interested in extending original position ideas into different-number cases, through the lens of a paper on life-extension I'm writing. Apologies if the below is long, unclear, and meritless:

What would motivate us to use 'all possible people' for our parliament, instead of convening a parliament on a choice-by-choice basis and restricting entry of souls to 'all possible people who could exist in at least one possible world we can choose?

The parliaments will be much smaller if we use the latter method, as the souls will not exceed the slots, and so the pro-natalist character remains. I have some intuitions that prefer the latter method: a parliament of souls blinded to the counter-factuals of their own existence but not to their brain history or DNA or whatever their personal identity resides in could find out whether they have any chance of existence once informed of the laws of nature, in what circumstances people could exist and choices available, etc. Those who have no chance have no stake (and so indifference), and the 'remaining players' may have enough of a stake that even bounding their utility makes their desire to increase their odds of incarnation important.

Do you see anything that motivates using your method of convening parliaments over this alternative, or costs of this 'other method' that I've missed?

In any case, if we get varying results on 'diluting the stakes' for a given soul is intriguing. Some of the perplexity may boil down to old problems re. (un)bounded utilities, but further exploration might give reasons to reject original position-like reasoning if (like person-affecting views) it proves very problematic to deploy it in different number cases.

Carl said...

"The parliaments will be much smaller if we use the latter method, as the souls will not exceed the slots, and so the pro-natalist character remains."

I don't think so. Remember the non-identity problem: almost every minor action we take scrambles which sperm and eggs unite, the times of conception and birth, etc. There are more possible humans who could be born next year depending on our actions today than there are atoms in the universe. As we consider more ways in which people can vary (developmental noise in different organ systems, different alleles for genetic difference, etc) the number of possible people goes up exponentially.

Thrasymachus said...

Hello Carl,

"I don't think so. Remember the non-identity problem: almost every minor action we take scrambles which sperm and eggs unite, the times of conception and birth, etc. There are more possible humans who could be born next year depending on our actions today than there are atoms in the universe. As we consider more ways in which people can vary (developmental noise in different organ systems, different alleles for genetic difference, etc) the number of possible people goes up exponentially."

My suggestion was poorly stated. Let me try a renovation:

Consider having to make a choice, A or B, at a given time t. (For concreteness, suppose A = banning life extension, B = not banning life extension). It seems to me we can can consider the future of A-world or B-world forward from t and see which (and how many) people exist. Then we convene a parliament with a soul for every possible person who will exist in one or both of these worlds.

Although due to the 'scrambling effects' you mention A-world and B-world will have fewer shared persons than we might at first suspect, doing things this way seems to 'bracket out' all the future persons contingent on time of conception, other choices, etc. So even on bounded utility the souls in this parliament look to have a strong interest in picking whichever of A or B has more opportunities for incarnation.

No obvious reason arises to me why deploying 'sub parliaments' for each choice available is worse than having a 'grand parliament' try and decide every different number question together, leading to the swamping concerns you outline.

Sorry if I'm still just not 'getting it'. Any further forbearance gratefully appreciated.

Carl said...

At any given time you never have just two choices, you have billions. Simply announcing your life extension proposal in different words or at different times scrambles people.

Brian Tomasik said...

I agree with others that this is an intriguing idea.

If we naively weighted just by a strict count of "number of possible people in person-space with a given preference," we'd run into troubles. There might be just as many preferences for great art not to exist as for it to exist. It's easy enough to imagine "art hating" minds, and they might be about as common in mind-space as "art loving" minds -- just hook up the wiring to disgust centers instead of beauty centers of the brain. For that matter, there may be as many sadists as altruists in possible-mind-space.

A more intuitive weighting would be the actual prevalence of evolved or built minds in the multiverse. Then, for instance, suffering reducers would be much more common than suffering increasers because organisms evolve to dislike suffering by themselves, their kin, and their reciprocal trading partners. Societies should often develop norms against cruelty for collective benefit.

Actually, weighting by prevalence in the multiverse has an obvious justification: Those people don't just potentially exist but actually exist! So it's more clear that their preferences should be satisfied, even to a run-of-the-mill preference utilitarian as long as she cares about preferences that are fulfilled outside the organism's own light cone or universe. The analogue of Carl's point is that if gazillions of aliens care even a tiny sliver about how human society is constructed, those preferences may dominate humans' own views on how their society is constructed. Of course, one might imagine commonplace views that place high value on the autonomy of societies to choose their own preferred destinies.

If only by quantum randomness, all possible people exist with some measure, but the weighting matters, and the natural weighting is their actual frequency of existence. Talk about possible-but-not-actual souls is no longer needed.

esrogs said...

Does this argument still work in an infinite universe? To make it go through, would you just argue that most minds / brain histories have low measure?