I've mainly worked at the cause area level, where I think these considerations are less relevant. But I have observed that if people are so enthusiastic about an area that many will volunteer (or take lower salaries), that does make progress in an area easier.

Maybe if the offsetting is convincing people to go vegan. But I think a lot of the offsetting is corporate campaigns to get higher welfare standards. That has much higher leverage than individual actions to eat higher welfare meat. Maybe once all the corporations have reformed, individual actions could start being competitive, but we can re-evaluate then.

About that example: I think you meant to write "at charity B" instead of "at charity A"?

Yes - thank you - fixed!

* Charity C: 1 employee and 10000 volunteers, 10001 impact total. Should the donor really get 10001 impact? To me that feels really off

Very unlikely to be the case in the real world, but if the donor makes the whole chain of events happen, then I think that is the counterfactual impact.

* Charity D: The employee also becomes a volunteer but does the same job. In a way the employee is "donating" their salary and becomes the donor. Should the employee now get "2 impacts"? That again feels off? Are they somehow better than the other volunteer?

Again, if without that person, there would be no impact, I think that's the counterfactual. In general, a volunteer who could manage other volunteers I think would be more valuable.

* Charity E: 100 donors, all donating $1. It's a blind kickstarter campaign and the charity can only do the intervention if they raise 100, otherwise money back, but the donors don't know how many others already donated, so order doesn't matter. In a way each donor should get the full "counterfactual" impact because without each individual donor the fundraise would've failed. But again attributing the same full impact 100 times feels wrong.

Yeah - an even more extreme case would be an election with 100,000,001 votes versus 100,000,000 votes - would all 100,000,001 people get to claim they cast the deciding vote for the election? Perhaps someone who has thought more about this wants to weigh in? @aaronhamlin ?

>A disadvantage of starting with market rate is that the person who would've accepted >the NGO rate might on average pay themselves more because they would more >consciously have to give up money. 

This is implying we are right now tricking people into giving up more than they consciously would. I'm not sure that's what we would want as a movement.

I don't think so-there's an analogy here with opt in versus opt out. You could say that people have a certain preference to be an organ donor and that whether it is opt in or opt out, you should get the same percent of participation of being an organ donor. However, we find that in reality, whether the system is opt in or opt out has a huge impact on the actual participation rate. This is similar for retirement programs.

I think this might vary heavily depending on the focus area. In AI for sure, not sure about the others.

Claude says 98th percentile salary globally is USD48k, and it estimates the median at AI-aligned orgs to be $65k for animals, $85k for global health, meta $85k and AI safety $115k.

I'm aware of the EA unemployment but I'm not sure if we should optimize for EAs working in EA orgs. Shouldn't the best talent or fit win? Especially for hard-to-fill roles like biorisk researchers, AI alignment people, a GiveDirectly country operation manager role, etc. Of course, value alignment shoudl play a part in assessing job match, but I'm not sure willingness to take a low salary is the right selector.

I agree that we should be maximizing overall effectiveness, and that reducing EA "unemployment" should not be a primary goal. However, I do think that there are a lot of well-qualified EAs who have not been able to get an EA job (see here for links).  I think there are social pressures to have EA wages not lower than other nonprofits. As evidenced by some EAs' donating patterns, they are willing to accept lower wages than most. So I think that total effectiveness would be higher with lower wages. Of course it does not apply in all cases.

Either way, all these points are downstream of the main point I'm trying to make: We should take market rate equivalents in cost effectivness calculations so we don't overpromise impact to donors. We can do this and then still decide to only accept candidates that take a 50% pay cut.

I definitely agree with the point that we should be recognizing the contribution of people who work for less money, including volunteers. Unlike Shapely values, counterfactuals can add up to more than 100%. This might be better in the other comment thread about volunteers, but here goes anyway. Let's say we have charity A that does one unit of impact per employee, and charity B that does one unit of impact per employee but each employee also manages volunteers that produce another unit of impact. I think that the counterfactual impact of the donor by funding the salary of the employee at charity B is two units, because without the donor, you would not get any impact. Then I think the counterfactual impact of the volunteers is one unit (because it wouldn't happen without them). Let's assume the employee is paid at market rate and replaceable for this thought experiment. So then we have two units of impact but the counterfactual impact is three units. So obviously Shapely values that must add to 100% you can't give all the impact to the donor. And you can easily come up with more complicated scenarios where this will not work out as cleanly. But I do think it's possible that using volunteers (or lower salaries) can increase the counterfactual impact of the donor.

Edited to switch to charity B.

But the fish only got saved once. It would be nearly impossible to determine who should get credit for how many fish saved. So from a marketing perspective, I think the smart move is to let everyone feel like the hero. 

Though Shapley values of impact have to add up to 100%, counterfactuals can add up to >100%.

Their cost effectiveness was lower than that of their "competitor" because their competitor paid themselves a lower salary.  This is a race to the bottom that is unhealthy for the movement.

There are many EAs would would like to get a job at an EA org, and lower salaries means that more jobs can be offered. So I think EA "unemployment" is a bigger problem for the movement than the people who do have EAs jobs not making enough money.

1. It would attract talent that otherwise would've been inaccessible because that talent wouldn't have been willing to take as big of a sacrifice.

Yes, but I generally think there is talent available in the EA community.

2. It would allow less privileged groups to join the charity world that otherwise just wouldn't have the means to take paycuts like this.

Most salaries at EA orgs are in the top couple percent of the world, so I don't think this is a big issue.

3. It would give employees a tangible measure to estimate how much impact they are having with their job.

Most employees already know the market rate, so they can subtract. And I think most orgs do appreciate their employees for working for less than market rate. But I could see it helping with getting credit for GWWC.

A disadvantage of starting with market rate is that the person who would've accepted the NGO rate might on average pay themselves more because they would more consciously have to give up money. They means a combination of more donor money required and more EA "unemployment", which I think would reduce effectiveness overall.

Adding AIs as potential bad actors themselves complicates this, and should be considered a separate category that might end up dominating all three of these categories.

Yes - as I said to Chris:

"Many people think that if the AI wants to take over, it will just win. But the AI may be in a hurry because it could be deactivated in a year, so it may try to take over when its chances of success are relatively low."

I think the bigger impact of AI than agricultural productivity is increased demand for manual labor as knowledge work is automated, so low income countries should get ready to export a lot.

But Floramérica didn’t just grow flowers; it engineered a multinational cold chain from scratch. It negotiated with airlines to create cargo space, designed specialized refrigerated trucks to navigate Andean mountain roads, and mastered the stringent plant import standards of US Customs. 

 

There is general resistance to buying things made in poor countries, which EA has argued against at length. But there is even more resistance to buying things made in poor countries that have higher greenhouse gas emissions than local products, like flown cut flowers. However, since the US is willing to spend around $10 million to save a statistical life, and since much of the total cost of saving a life with an intervention to reduce radon concentrations in homes is the energy to power the ventilation fan, it is probably millions of dollars worth of energy to save a life. I've estimated that it costs much less in energy to save a life in a poor country by giving them employment even for something like flown cut flowers.

The dexterous manipulation of malleable, unpredictable fabric remains one of the genuinely unsolved challenges in robotics. Even where automation is technically feasible, the economics of deploying it against a workforce earning $65 a month remains deeply unfavorable—robots have high fixed costs, must be retooled for different product lines, and, as orders fluctuate, cannot flex up or down the way a human workforce can. Human labor in physical manufacturing may eventually be made redundant, but even apparel manufacturers in China today still rely on human tailors to drive their production.  

Yes, and I think that there will be a window where AI automates most knowledge work, but then the large economic growth that results would mean that there would be much greater demand for products that require physical labor. So I think it's important for poor countries to get ready to export a lot of these goods.

During the FTX era, it lost that for a while. Getting your project funded basically required that you had a pulse and sounded like an EA.

Do you mean just in AI safety/meta? Because the FTXFF was only funding long-term projects, so I know there were still huge funding gaps in global poverty and animal welfare. And even in bio and nuclear security, there were still very large funding gaps.

I have been offered jobs that are billed as impactful and that pay vastly more money than I need while not providing any clear benefit for the world.

That's great that you are so successful. But there are many EAs/rationalists who are well-qualified who still have not gotten an EA job (probably thousands). I think part of the disconnect between the people saying there is lots of money and the people who are saying that they can't get a job is that EAs are so well-qualified that they are used to only having to apply for a few jobs per offer. The number of applicants per job is something like 30-300, so it makes sense that the average person needs to apply to that many to get a job. But when EAs apply to 10 to 30 EA jobs and don't get one, they tend to get frustrated. Also there are just not that many EA jobs that would be relevant to an individual EA's qualifications.

Furthermore, there are enormous potential effective uses of money that don't require EA labor, such as Give Directly, stockpiling PPE, etc.

So overall, I do think we should be careful to limit grift, but there is enormous room for effective use of funding in EA, one estimate was ~$1 trillion.