How does the value placed on creating new lives versus improving existing ones affect how an effective altruist should act?

1 point

expand_less expand_more

Nov 8, 2013

I'm not familiar with the terminology used by people who talk about effective giving, so I'll be using a lot of scare quotes and this post will probably be very imprecise.

Say that you are a total utilitarian (meaning that you think doubling the number of lives is as good as doubling the utility of every life). Even if you believe that increasing the number of lives is good, it's not clear what the marginal cost of additional lives is (for you, given the options available to you). If it is high, then whether you are a total or average utilitarian doesn't matter much, except that a total utilitarian wouldn't want to fund family planning, education, etc. So the effectiveness of donations meant to increase "time lived" strongly affects how differently total-utilitarian and "typical" EAs should behave.

For concreteness, suppose someone are trying to maximize the total number of QALYs (quality-adjusted life years) lived in the next 100 years by all people who are alive at any point in this time interval - this is your "social welfare function" as an EA. A big limitation of this is that QALYs don't take non-health-related quality of life factors into account; I chose them because I wanted the social welfare function to be relatively easy to calculate.

People who donate to charities generally don't have population growth as an explicit goal, so at first it seems like a total utilitarian EA should act very differently. However, a lot of the most effective charities, as judged by GiveWell, are public health initiatives which greatly increase average QALYs. They probably increase total QALYs as well, although the effect of health improvements on fertility needs to be considered. Another issue is that due to the lack of donors, there aren't many (or any) charities that increase total QALYs in a cost-effective way. Public health initiatives that increase total QALYs probably don't do it in an optimal way, since the increase is just a side effect of their main goal of improved health.

How should the donations of a total-QALYs-maximizing EA differ from:

Your donations if you were an EA acting according to your own values,
How the typical person you know would donate if they were an EA,
How an average-utilitarian EA would donate, and
What GiveWell would recommend?

Also, I've read up about "earning to give," and it seems like a reasonable strategy for EAs who have values shared by many charities. Is this also reasonable for a total-QALYs-maximizing EA? JonahSinick thinks that EAs with "unusual values" might benefit more from earning to give, but this seems strange to me, since there's unlikely to be an effective charity working toward goals that few people share.

1 point

expand_less expand_more

Comments

sorted by

Highlighting new comments since Today at 7:17 AM

All comments loaded

3 points

expand_less expand_more

People who donate to charities generally don't have population growth as an explicit goal, so at first it seems like a total utilitarian EA should act very differently. However, a lot of the most effective charities, as judged by GiveWell, are public health initiatives which greatly increase average QALYs

Public health initiatives lower population growth.

If children are less likely to die, women get fewer of them.

3 points

expand_less expand_more

3 points

expand_less expand_more

Public health initiatives lower population growth.

Not quite. They may lower the birth rate, but, taken by themselves, I don't see why they would lower the population growth.

The reduced fertility is usually attributed to greater wealth and education. I think it would be hard to disentangle public health advances from the general correlated mix of things that happen in a society as it becomes richer.

3 points

expand_less expand_more

-2 points

expand_less expand_more

Not quite. They may lower the birth rate, but, taken by themselves, I don't see why they would lower the population growth.

It nothing that's trivaly true. Bill Gates spend a lot of money via his foundation in that area and holds the belief that public health initiatives are a good way to reduce population growth.

First, we've got population. The world today has 6.8 billion people. That's headed up to about nine billion. Now, if we do a really great job on new vaccines, health care, reproductive health services, we could lower that by, perhaps, 10 or 15 percent, but there we see an increase of about 1.3.

I don't think it's impossible to separate the effects. There are a lot lot of philantrophic projects that are focused on a specific area and you can compare that area with others.

-2 points

expand_less expand_more

Pablo_Stafforini

7 points

expand_less expand_more

Bill Gates spend a lot of money via his foundation in that area and holds the belief that public health initiatives are a good way to reduce population growth.

A GiveWell intern--now a full-time employee--spent quite some time researching the effects of global poverty interventions on population size and reached no firm conclusions. I don't think Gates' beliefs, in the absence of details about his reasoning process, provide much evidential support in favor of your original claim.

7 points

expand_less expand_more

2 points

expand_less expand_more

trivaly true

You said:

If children are less likely to die, women get fewer of them.

On the first glance the net effect looks to be zero. Remember that population growth is driven by how many children survive to reproduce.

I don't think it's impossible to separate the effecs.

Do you have any links to attempts to do so?

2 points

expand_less expand_more

0 points

expand_less expand_more

Remember that population growth is driven by how many children survive to reproduce.

If you have one child per couple, your population won't grow even if all the children survive to adulthood; conversely, if you have five children per couple, your population will grow even if half the children die before puberty.

0 points

expand_less expand_more

3 points

expand_less expand_more

"How many" is not the same thing as "which percentage of".

3 points

expand_less expand_more

1 point

expand_less expand_more

Isn't it exactly what army says? Putting all other things aside: "how many" is not the same thing as 'which percentage of" precisely because even if 90% of 100 children per couple dies we still have 10 children per couple and growth of population.

If public health initiatives lowers "percentage of" surviving children and at the same time lowers "birth rate" we still can have ether option - decrease or increase - based on "percentage * rate >< 2"

What did I missed here?

1 point

expand_less expand_more

0 points

expand_less expand_more

And “if children are less likely to die, women get fewer of them” isn't the same thing as “the number of children women get is inversely proportional to the probability that they survive”. My point still stands.

0 points

expand_less expand_more

1 point

expand_less expand_more

A total utilitarian would tend to deprioritize educational initiatives; there's an inverse correlation between education and number of offspring.

1 point

expand_less expand_more

1 point

expand_less expand_more

JonahSinick thinks that EAs with "unusual values" might benefit more from earning to give, but this seems strange to me, since there's unlikely to be an effective charity working toward goals that few people share.

My explanation for that would be "if few people share your values, it will be harder for you to convince other people to give you money to pursue that value," which suggests that by earning and then funding work towards that value, you can do more. I do suspect that most unusual causes are funding limited.

1 point

expand_less expand_more

0 points

expand_less expand_more

Your donations if you were an EA acting according to your own values,

I don't think this question is well formed. Unless you were being altruistic, you're not an EA. If your own values are altruistic, then "your own values" doesn't add anything to "EA"

2 How the typical person you know would donate if they were an EA,

The typical person isn't an EA. Their becoming an EA involves such a massive change it isn't clear that their pre-EA values would have much effect. Should by being a Sunni or a Shi'ite change what the most effective way of helping people is?

3 How an average-utilitarian EA would donate

Average-utilitarian EAs will want to do crazy things like continually murder unhappy people until only one, superbly happy (and probably sadistic, to be happy in such a scenario) person was left. Total utilitarianism is the default brand of utilitarianism.

0 points

expand_less expand_more

1 point

expand_less expand_more

I doubt sadistic... a sadist presumably wants other people around to hurt.
But certainly atypical, if their utility is consistent with everyone else in the world being dead.

1 point

expand_less expand_more

0 points

expand_less expand_more

I'm not entirely sure whether you

propose QALYs as a means to optimize a total utilitarians goals or
discuss what you as a total utilitarian could do best to optimize this goal or
something else related to both

I assume the former because you wrote

I chose them because I wanted the social welfare function to be relatively easy to calculate.

In general if you choose any oversimplified scheme to optimize for you will not earn what you want. What gets measured gets optimized.

The following quotes are from Wikipedia: http://en.wikipedia.org/wiki/Quality-adjusted_life_year

QALYs are empirically known to be oversimplified and more a theoretical economists tool to derive general optimization potential that a precise tool.

The four theoretical assumptions underlying QALYs are invalid (quality of life should be measured in consistent intervals; life years and QOL should be independent; people should be neutral about risk; and willingness to sacrifice life years should be constant over time).

They are neither recommended for individual health care decisions where they

place[s] disproportionate importance on physical pain or disability over mental health. The effects of a patient's health on the quality of life of others [..] do not figure into these calculations.

nor on the population as an aggregate where

the weight assigned to a particular condition can vary greatly, depending on the population being surveyed.

Also if you want to use it as a tool to personally rate some means you should consider that

those who do not suffer from the affliction in question will, on average, overestimate the detrimental effect on quality of life, compared to those who are afflicted.

So I propose that you choose a more elaborate tool set if you want to optimize a complex goal. Otherwise you fall into the same trap as you wat to avoid from UFAI: Overoptimizing oversimple goals.

0 points

expand_less expand_more

1 point

expand_less expand_more

I wanted a concrete discussion about how a total utilitarian (TU) should act, not one about what exactly their utility function should be. I think total QALYs are at least a better approxmiation of a TU social welfare function than other simple social welfare functions (life expectancy, GDP per capita, education, reported happiness, etc.), since they are all average measures. For all of these except happiness, you can construct a "total" version:

Life expectancy becomes total years lived,
GDP per capita becomes total GDP, and
Average education level becomes total years of education.

If you don't like how ambiguous QALYs are, you can use total years lived (QALYs without the quality adjustment) or total GDP as social welfare functions (although total GDP seems suspect because a TU might prefer two people living on, say, $10000 a year to one person living on $50000). The total number of adult years lived would also be a reasonable metric.

Basically, since the implied social welfare functions of most donors and charities seem very far from any reasonable TU social welfare function, even fairly oversimplified metrics can be much better than the status quo from a TU's perspective. In general, an effective altruist with unusual values has to worry less about oversimplifying, since even a crude social welfare function can be (from their perspective) much better than what people currently do.

1 point

expand_less expand_more

1 point

expand_less expand_more

I wanted a concrete discussion about how a total utilitarian (TU) should act,

Maybe you should have more clearly said so. And mentioned the less satisfactory measures too.

How do you handle the 'mere addition paradox' (Parfit)? Do you

assert that higher utility living is on a completely different scale from, and thus incomparable to, the bottom levels of utility, or
deny that there is anything wrong with the 'repugnant conclusion'?

(quoting Wikipedia)

It seems to be the latter.

I personally again think that it is not simple addition. Adding a life directly affects all the other lifes around it. This will never be a simple addition (except if you make it so like in Brave New World where new lifes arise in 'bottles'). Same if you 'subtract' below average lifes. Also watch out for the szenarios which from a total utilitarian perspective look good too: wire heading and other psycho engineering which just change your valuation of the state ('happy slaves').

1 point

expand_less expand_more