The two sources of moral value
answering the age-old question of why put a value on human life
Before we get going with this post: Hello to all the new subscribers arriving from the nerd oscars! I finally added something to the about page if you’re curious about the backlog. This has all been very flattering and very surreal, and I hope at least some of you end up liking it here.
Today’s general musing piece is about as political as I get here (I generally try to avoid getting too political). I should have another spicy-but-not-especially-partisan piece coming up soon,but aside from that I can hopefully stick to stories, weird social mathematical models, and reopening my dating page.
It’s often said that people’s lives have moral value1. But if we ask why people have moral value, there’s two distinct answers2:
Consumption value: People have value because of their ability to experience things. People experiencing joy is good and people experiencing pain is bad, and causing positive emotions in someone is a terminal good.
Production value: People have value because of their ability to produce things. A factory worker has value because he can produce widgets, a mother has value because she can produce and raise children, a soldier has worth because he can defend his society, and so on. Outside of jobs, people still produce massive amounts of value in their daily lives, from buying groceries for the family to helping a friend buy furniture. Society actively needs the results of all this production to survive and function, so the people who produce them are valuable.
Note that we can mostly treat either of these as being downstream of the other: Production can be thought of as only having value because it creates consumption goods (people’s work creating good feelings in other people). Conversely, consumption values may have value because they’re signals of production values (people feel happy when they’re effective and productive and depressed when they feel useless).
These break down in different ways when we reach weird moral edge cases. If we only values production, we can imagine reaching a world with a fully automated economy with no one to enjoy it. As Bostrom put it3:
We could thus imagine, as an extreme case, a technologically highly advanced society, containing many complex structures, some of them far more intricate and intelligent than anything that exists on the planet today – a society which nevertheless lacks any type of being that is conscious or whose welfare has moral significance. In a sense, this would be an uninhabited society. It would be a society of economic miracles and technological awesomeness, with nobody there to benefit. A Disneyland with no children.
This doesn’t seem like it has much value. Even on a more low-tech level, a town of people who go to work cheerfully every day seems better than one where they do it miserably, even if the two are otherwise identical4.
The weird extreme for a fully consumption-based value system is utility monsters and misery exporters. If we follow it off a scifi cliff it tells us to tile the universe with raven brains5 on heroin. Slightly more prosaically, it gives moral weight to groups of people by their ability to experience, which breaks all sorts of locality expectations. For example, if we suddenly discover a large island population who, for some hypothetical reason, is completely incapable of ever being productive, it obligates us to spend the effort to take care of it. At the extreme, you could imagine it being a pure misery exporter - a people that produce nothing and live off the aid we feel morally obligated to give while actively trying to maximize both their number of children (which we are more sympathetic to, since we hold them morally blameless for their environment), and those children’s misery (so that we’re maximally incentivised to send them aid). Each step along the chain to enabling this society makes sense given our moral fundamentals, but the end result is something I don’t think anyone actually wants to end up with.
As this example shows, attempting to solve the value conflict by just valuing both production and consumption independently is doomed to failure, since both utility monsters and disneylands without children would still have value6.
In Planecrash, Eliezer describes how his hypothetical sane society of dath ilan would resolve the value conflict. The society has fundamental values that are deeply consumption-based (they’re horrified by the notion of anyone truly dying, ever, in any way), but their governance and society is completely production based (there’s basically no form of welfare that could ever create misaligned incentives7, which means hardly any at all). dath ilan (and, presumably, Eliezer) square away the parts they can’t solve by putting anyone they can’t support on cryonics and kicking the can down the road to when we have friendly superintelligence.
Okay but seriously, which one is better?
So while both sides have a point here, I think the production side is mostly better. It scales better at the extremes - the dangers of approximating a disneyland without children seem much less risky to me than those of approximating a utility monster. And production focused values have locality, which is a property that generally helps with scaling.
This may not be a universal thing. Asian societies may be too production focused - you can see this in Korean society (which, from everything I’ve heard, seems obsessed with achievement to the point where everyone is miserable and, increasingly, childless), or China’s export-focused economic policies, which boost chinese production over consumption8, creating a massive export surplus. It seems plausible that they should try to be a little happier there9.
Putting dollar values on our values
We might ask, what does this tell us about the dollar value of a statistical human life. Most (american) estimates have it at around 8-10 million dollars. Actually getting this number is surprisingly straightforward: Just take the total GDP per capita of the United States (about $82K), then multiply it by the average lifespan (about 77 years), to get a number of about 6.2 million10. This value is actually independent on whether you assume a consumption- or production-based valuation of human life: Any economic transaction between people involves production on one end and consumption on the other (e.g. if you make someone an ice cream sandwich, you can account for the value produced as the production you put in to create it, or the enjoyment he got from consuming it, but either way you end up adding the same number), so when you look at the average across society, you end up with the same number.
This is a neat way of getting around the problem of what we value about life by averaging it out and only looking at the value of a hypothetical statistical life. But it doesn’t tell us the value of any specific individual life.
The first obvious problem is age - we multiplied GDP by average lifespan, but a ninety year old getting hit by a truck a day before he would otherwise have had a fatal heart attack, while still tragic, seems, to be less of a loss than a twenty year old getting hit by the same truck. Should we multiply the value of life by average remaining lifespan? In practice healthcare systems often do, by using a metric of QALYs to measure how useful an intervention is11.
The other more controversial question is whether everyone’s life is worth the same. It’s probably based to assume everyone has equal consumption value12 . People almost certainly vary wildly in production value though. If you’re in a trolly that can run over a random stranger or Norman Borlaug, you obviously kill the stranger. This implies that, for consistency, we should assign some people’s lives a higher monetary value based on their productivity.
Finally, there’s the national origin issue. We got the above number by using american life expectancy and GDP, but other countries also exist13, and they generally have lower GDP per capita. Does this imply their lives are worth less?
To some degree I’m willing to bite this bullet. It’s worth noting the gap in local real value (as opposed to nominal dollar) terms isn’t nearly as big as you’d expect based purely on GDP numbers, since you have to adjust for PPP (and possibly more), and once you do that the gap between america and other first-world countries mostly vanishes.
But since we used a purely production-based metric to get this number, we’re going to see a value roughly proportional to GDP per capita, which means for third-world countries it’s going to be extremely low. If you’re a production value purist you may be fine with this, but if you’re a consumption value believer (even in part), there’s a huge amount of moral value arbitrage to be had from this. Doing the arb trade on this is what we normally refer to as Effective Altruism.
Some political implications
As much as I hate to do politics on here, there is an interesting note about left/right politics here. Broadly speaking, the left uses consumption-based values, and the right uses production-based values.
This isn’t universal and is a bad guide to practical politics (please don’t read this and go “well I value people’s internal experience, so I should always vote for my local left-wing party”, or vice versa). The real world has plenty of path-dependent idiosyncrasies and issues that aren’t all caught by this dynamic, like which side’s politicians are smarter or more honest, or issues like abortion that aren’t really described by this dynamic at all. And then there’s issues like immigration, where people make both consumption- and production-based arguments for both sides (which may explain why the politics of immigration have historically been so inconsistently partisanly aligned).
But still. If you’re ever in a position where you genuinely don’t understand how someone could reach political opinions that are so obviously wrong you can’t even imagine how they got there - consider that it may be because the two of you are valuing the two different options on this axis.
Citation needed.
Well, two distinct utilitarian answers. Non-utilitarian ethics, which are mostly designed for working within the system, can give different answers. A virtue ethicist might say your friends have value because loyalty to friends is virtuous, for example. But these are mostly differences about how to act in life rather than core values, and will be ignored here.
I’m linking Scott’s post quoting Bostrom here, since I can’t find where he got the original quote from.
Although it’s possible that since happiness happens for a reason, this is in fact impossible and two towns with different moods would also inevitably differ in some other important metric like slack or social trust, even if they had identical economies.
Which have more neurons per mass than rat brains and are thus more effective.
If you’re a mathematician, this is where you might be thinking “well what if we just took the product to be our value?”. That wouldn’t be a direct disaster, but I’m not sure what the product of production and consumption even means, as a value.
For example, if you can’t work to support yourself and don’t have kids you can go live cheaply alone in a minimalist quiet city. But if you’ve ever had kids or would want to in the future you’re not allowed that, to avoid creating an incentive for people to take advantage of society by having kids they can’t support.
I’ve heard Shanghai described as being futuristic and great in every way New York fails - clean, modern infrastructure, no social disorder and an ever expanding metro system - except that everyone in Shanghai is miserable while New Yorkers are mostly happy. I do not have the knowledge to judge this claim for myself.
If true, this has the saddest of all possible implications: That building more and better trains may not be enough to singlehandedly make everyone truly happy.
Places with a history of both western and eastern influence, like Hong Kong and Singapore, seem in something of an odd middle. Like mixed-origin hair in the stormlight archive, they have a mix of pure strands of each culture in different places instead of being a smooth midpoint of the two.
The gap between this number and the earlier value of $8-10 million is mostly explained by the expectation of GDP growth: People are expected to spend much of their lives in the future, where GDP is expected to be higher, so their average lifetime GDP per capita is adjusted up.
This still assumes years are (up to quality) equivalent, but they might not be - from a consumption-based view a nine year old’s year may have more value since he experiences things more intensely, while from the production-based view a forty-year old’s year might be worth more since he produces more with it. I can see why healthcare systems would prefer to avoid making this kind of judgement.
It could hypothetically be the case that some people experience qualia less than others, but there’s almost certainly no good way to determine this, and there’s a lot of ways that going down this line could end badly.
Citation needed.