Which effective altruism projects look disingenuous?

post by Mati_Roy · 2021-01-03T07:28:39.662Z · EA · GW · 5 comments

This is a question post.

Contents

  Answers
    Mati_Roy
None
5 comments

Answers

answer by Mati_Roy · 2021-01-03T08:11:20.136Z · EA(p) · GW(p)

Importance: not really important to read this comment

Update: I updated; see my reply

GivingMultiplier's description according to the EA newsletter^1:

Let's assume Effective_Charity and Local_Charity. 

If you were going to give 100 USD to Local_Charity, but instead donate 10 USD to Effective_Charity and 90 USD to Local_Charity, GivingMultiplier will give 9 USD to Local_Charity and 1 USD to Effective_Charity, so there's now 99 USD going to the Local_Charity and 11 USD going the Effective_Charity. GivingMultiplier would give the money to Effective_Charity anyway. So for the donor, this is indistinguishable from donating 99 USD to Local_Charity and 1 USD to Effective_Charity, but it's done in a more obscure way.^2

Also, sure they are rather transparent about their process – at least in the newsletter; it wasn't obvious from the main page of the website –, but still, their scheme mostly works only insofar as people don't understand what's going on.

Potential motives

A bunch of people don't know Why you shouldn’t let “donation matching” affect your giving, and so they will be misguided by donation matches. If EA charities don't use them, then they might be at a disadvantage. So their reasoning might be that the game theory favors also using this technique under a consequentialist moral framework – sort of like a tit-for-tat with other charities, with deceiving donors as an externality.

One could argue that they should link to the piece against donation matching on their website, but maybe both memes are fit to different environments – maybe it would mostly reduce how much people use that specific service to fill their donation matching need, or something like that. I don't know, I'm trying to steelman it.

They might also want to know where people donate money, so they allow people to choose where some money goes among those 9 charities in exchange for knowing where they donate the rest of the money. And at the same time, they signal support for those 9 charities.

Consequences on the donors

If donation matches don't change how much donors give, but just where they give (which seems plausible to me), biasing them equally against all charities might actually help them make decisions that are more aligned with their worldview than if they were less biased with only a subset of them.

Footnotes

1) There website is actually giving different numbers, but the idea is the same.

2) Sure, there's the real choice of choosing which of the 9 Effective Charities receive the money, but:

a) The part about local charities is a red herring

b) Those charities probably sort-of have reached market efficiency (in the sense that large donors can rebalance their donations according to  how much total funding they want each of them to have)

(a) is my main objection.

comment by Lucius_Caviola · 2021-01-04T18:31:52.426Z · EA(p) · GW(p)

You are right that if someone only cares about their favorite charity, then donating through GM doesn't give them any value. After all, GM never helps you to get more value for your favorite charity than you could get by donating directly to your favorite charity. But we also don't claim that we do that. On our website, we say: "Give to both your favorite charity and a super-effective charity recommended by experts. We'll add to your donations." (The EA newsletter text frames things slightly differently and perhaps that's indeed not the optimal way of promoting GM.)

But if someone cares about both their favorite charity and about giving effectively, donating through GM can get them more value. Keep in mind that our target audience are non-EA donors, many of whom haven't heard of EA or about our highly effective charities before.

In our studies we find that many non-EA people (ca. half of our MechanicalTurk participants!) are willing to split their donation 50/50 between their favorite and a highly effective charity when they are offered such a splitting option even if no matching is offered. This shows that surprisingly many people do have a preference to give to effective charities. They just don't know about effective charities yet and don't consider the option to split their donation. The point of GM is to inform non-EA donors about effective charities and offer them this splitting option.

Suppose you have a donor who cares about their favorite charity and a very effective charity. They want to give 90% to their favorite and 10% to the effective charity. They could either donate directly to these two charities or they could donate through GM. If they donate through GM, the system adds on top of their donations. 

The part that is added on top of their favorite charity is clearly counterfactual because the matching funder wouldn't have given to that charity. The part that is added on top of the effective charity is less counterfactual because the matching funder would have given anyway to effective charities. But in expectation it is partly counterfactual because the donor can influence which specific effective charity this part of the funding should go to (and many donors may care much more about some effective charities than others). (Your efficient market hypothesis is interesting and I haven’t considered it. But I doubt that the market for effective charities is completely efficient.)

As Aaron pointed out, all of this is transparently explained on our FAQ page.

Do my matched donations have an impact?

Yes. The donors who provided the matching funding would likely not have donated to the specific charities that you have chosen. Therefore, by making a donation through Giving Multiplier, you don't just decide to which charities your own money goes to but you also decide to which specific charities the added (i.e. matched) amounts—that were provided by the matching funders—go to. Note that most matching funders likely would have donated their amounts to a highly effective charity per default. But they would not have donated to your favorite charity and it's unlikely that they would have donated to exactly the effective charity that you have chosen.

Our website is new and if there are ways to improve, we'd consider these. But to be clear: there is absolutely no intent of deceiving donors. 

comment by Mati_Roy · 2021-01-05T03:37:52.448Z · EA(p) · GW(p)

hummm, I guess it's fine after all. I change my mind. People can just give whatever fraction they were going to give to local charities, and then be matched. And the extra matching to effective charities is a signal from the matcher about their model of the world. I don't think someone that was going to give 100% to another charity than those 9 should use GivingMultiplier though (unless they changed their mind about effective charities). But my guess is that this project has good consequences.

comment by Aaron Gertler (aarongertler) · 2021-01-04T07:52:01.496Z · EA(p) · GW(p)

(Epistemic status: Writing this in a hurry on my way to something else, trying my best to be accurate but I'd love to know if I miss any details. Also, not writing this as a mod or a CEA employee-- just doing my best to understand this organization.)

It looks like GivingMultiplier has changed their match structure a bit since I sent out the EA Newsletter, so that's not the best reference to use.

In any case,  your argument has some questionable bits, whether you use the Newsletter's out-of-date numbers or the GivingMultiplier numbers.

First, the higher a % of the donation is given to an effective charity, the greater the match. If I give 90/10 to non-EA/EA, I get $3 in matching funds; if I give 50/50, I get $15 in matching funds. This doesn't change the "indistinguishable from if I gave X" property, but it is a thing that would have been easy to check before posting.

Second, point (b) matters. It seems like a bold assumption to assume that EA charities have reached "market efficiency" -- very few donors, even in our community, are paying such close attention to all charities' total funding that they would rebalance this way on the fly. I work with a donor who gives nearly $1 million to EA orgs each year, sometimes has to make allocation decisions within a few weeks, and has nothing like the time he would need to examine all the EA orgs he cares about and figure out how to contribute to the optimal balance of funding between them. You could argue that some giant, thoughtful funder like Open Philanthropy might balance out the rest of the "market", but even Open Phil has a limited scope and makes relatively infrequent grants (giving the market plenty of time to become "imbalanced" again).

Thus, if you actually think one of the "EA" choices at GivingMultiplier is more valuable than the rest, it seems very likely that you contribute more to their work by choosing them to be matched. The below is only true if you assume that every charity in the category "Effective_Charity" is equally cost-effective.

GivingMultiplier would give the money to Effective_Charity anyway.

Also, note that GivingMultiplier generally doesn't decide where unused funding goes -- that's at the discretion of the original donor:

Any matching funds that are unused after 6 months will automatically be donated to the effective charity that you chose.

Finally, GivingMultiplier explains their philosophy pretty clearly on their website. Did you see anything on the site that actually seemed false to you? I read their program as quite simple:

  1. Give people an incentive to think about splitting their donation between "heart" and "head", by...
  2. Setting up a match such that people who give more, and give more to the effective charity, really are redirecting matching funds to the effective charity of their choice (rather than the others on the list) and the local charity of their choice (rather than whatever local charity future match-ees would have chosen, assuming the matching funds run out at some point).
    1. To quote the site: "Note that most matching funders likely would have donated their amounts to a highly effective charity per default. But they would not have donated to your favorite charity and it's unlikely that they would have donated to exactly the effective charity that you have chosen."

I think GiveWell is correct that matching campaigns are often disingenuous. But this one is, by contrast, relatively open and clear about how it works.

If you think they could have been even more clear, or think that most donors will believe something different despite the FAQ, you could say so. But to say that people who use the match "don't understand what's going on" is both uncharitable and, as best I can tell, false.

My take on GivingMultiplier: They offer a small but objective benefit to donors who don't  value all charities equally, and also offer the much larger benefit of being a well-designed entry point into effective giving for donors who haven't thought much about it before -- but who, through the website, may discover charities that let them do much more good than they were doing before. It's not the Platonic ideal of a match that GiveWell is trying to achieve through their podcast advertising, but it's honest and real.

Replies from: Mati_Roy
comment by Mati_Roy · 2021-01-04T08:09:45.816Z · EA(p) · GW(p)

This doesn't change the "indistinguishable from if I gave X" property, but it is a thing that would have been easy to check before posting.

I did check. As you said, it doesn't change the conclusion (it actually makes it worse).

Second, point (b) matters. It seems like a bold assumption to assume that EA charities have reached "market efficiency"

I'm >50% sure that it doesn't fare better, but maybe. In any case, I specified in my OP that my main objection was (a). 

Thus, if you actually think one of the "EA" choices at GivingMultiplier is more valuable than the rest, it seems very likely that you contribute more to their work by choosing them to be matched. 

Yep, I did  mentioned that in my OP.

Did you see anything on the site that actually seemed false to you?

No,  I also mentioned this in OP.

  1. Give people an incentive to think about splitting their donation between "heart" and "head", by...

There's not really a real incentive though. I feel like there's a motte-and-bailey. The motte is that you get to choose one of the 9 charities, the bailey is that the matching to the local charity is actually meaningful.

and the local charity of their choice

That's meaningless as I showed in OP.

If you think they could have been even more clear, or think that most donors will believe something different despite the FAQ, you could say so. But to say that people who use the match "don't understand what's going on" is both uncharitable and, as best I can tell, false.

[...]

I disagree. shrug

Replies from: aarongertler
comment by Aaron Gertler (aarongertler) · 2021-01-04T08:51:28.117Z · EA(p) · GW(p)

That's meaningless as I showed in OP.

I don't understand what you mean.

Let's say that GM has $100 in matching funds to distribute. I like Doctors Without Borders and AMF. You prefer March of Dimes and Clean Air Task Force.

I give a $333/$333 split to my charities. That's a 50/50 split, which gets a 15% match from GM, which equates to $100. 

If I get there before you, Doctors Without Borders and AMF both get an extra $50. 

If you get there before me and do the same split, March of Dimes and CATF get $50.

Those are different states of the world, determined by which of us gets the match. 

If neither of us had used the match, GM would have given $100 to the charity chosen by whichever donor was matching us. That's a third possible state of the world.

If we assume that GM has limited funding, every person who gets a match is theoretically taking funds for their charities, at the expense of someone who would have used those funds for other charities. If this person likes their charities more than most other charities, they are benefiting in some way.

In theory, you could argue that the original matching donor is hurt, because they lose money that would have gone to a charity of their choice -- but they chose to fund a match, likely because they wanted to encourage people to think more carefully about funding effective charities and were willing to "pay them" to do so.

What do you think is wrong about this model?

Replies from: Mati_Roy
comment by Mati_Roy · 2021-01-07T06:26:14.845Z · EA(p) · GW(p)

I guess I was working on the assumption that it was rare that people would want to split their donation between local and effective a priori, and my point was that GM wasn't useful to people that didn't already want to split their donations in that way before GM's existence -- but maybe this assumption is wrong actually

comment by Mati_Roy · 2021-01-04T12:09:13.636Z · EA(p) · GW(p)

I'm henceforth offering a MetaGivingMultiplier. It's the same structure than GivingMultiplier, but replace "local charities" with "GivingMultiplier" and "super-effective charities" with "a cryonics organization" (I recommend https://www.alcor.org/rapid/ or https://www.brainpreservation.org/). Anyone wants to take advantage of my donation match?

h/t: came up with this with Haydn Thomas-Rose

5 comments

Comments sorted by top scores.

comment by Kit · 2021-01-03T11:37:59.552Z · EA(p) · GW(p)

I'm totally not a mod, but I thought I'd highlight the "Is it true? Is it necessary? Is it kind?" test. I think it's right in general, but especially important here. The Forum team seems to have listed basically this [? · GW] too: "Writing that is accurate, kind, and relevant to the discussion at hand."

I'm also excited to highlight another piece of their guidance "When you disagree with someone, approach it with curiosity: try to work out why they think what they think, and what you can learn from each other." On this:

  • Figuring out what someone thinks usually involves talking to them. If posting here is the first someone has heard of your concern, that might not be a very good way of resolving the disagreement.
  • Most people running a project in the community are basically trying to do good. It sounds obvious, but having a pretty strong prior on disagreements being in good faith seems wise here.
Replies from: aarongertler
comment by Aaron Gertler (aarongertler) · 2021-01-04T07:58:47.802Z · EA(p) · GW(p)

I agree with everything that Kit has said here. This post might have been in sufficient violation of the Forum's rules to remove (being slightly inaccurate and slightly unkind), but I'm leaving it up (without asking the author to consider changes, as I typically would -- see following comment) because I think Kit's comment suitably addresses my concerns.

EA orgs aren't run by angels. Any community where money changes hands will attract people who want to deceive others, with or without good intentions. But it's really good to reach out to people before accusing them of deception; they could be making an honest error, you could be making an honest error, or the issue could simply be a difference of opinion within a moral gray area. We're working in a field with many complex questions (moral and logistical), and the best first reaction to confusion is communication.

Replies from: Mati_Roy
comment by Mati_Roy · 2021-01-04T08:17:55.956Z · EA(p) · GW(p)

I agree with what Kit said as well.

But that the only reason you're not removing it is because of Kit's comment makes me pretty concerned about the forum.

I also disagree that private communication is better than public communication in cases like this.

Replies from: aarongertler
comment by Aaron Gertler (aarongertler) · 2021-01-04T08:41:33.473Z · EA(p) · GW(p)

I should have been more clear on that point -- thanks for the comment. I've changed my reply to add the phrase "without asking the author to consider changes, as I typically would". I can see how the original reply could have been concerning.

On handling posts that may violate Forum rules:

My first act for any post that seems to violate rules is to contact the author and express my concerns; I've probably done this ~10 times in the last two years. (The exception to this is for a post that is in stark violation of rules -- e.g. an insult with no further content, or obvious spam.) 

If  the ensuing discussion doesn't lead me to change my view on whether the post violated a rule, and the author declines to make changes to the content in accordance with the Forum's rules, the post might (again, might) be moved back to "draft" status (we don't delete non-spam content -- we want the author to be able to share things elsewhere even if the Forum doesn't permit them).

Of the aforementioned ~10 instances, I removed content one time when the author never replied (this was a comment that shared provably false and inaccurate information about a named person in a way that was hard to correct with a reply). On one or two other occasions, authors chose to remove their work. 

In every other case, I was convinced by the author, the author made light edits (generally of the "softening tone without changing substance" variety), or a discussion developed that seemed valuable enough for leaving the post up to be a net positive.

On private vs. public communication:  

The big difference is that, in most cases, you can move from a private to a public discussion more smoothly than vice-versa. Once a public accusation has been made, confusion and concern tends to linger, whatever the substance of the accusation. You can see this in action when a false Tweet gets 50 times as many retweets as the correction.

Public accusations also tend to lead to bitter fights that could have been avoided with a private conversation: Forum User A leaps to defend the accused, Forum User B fires back, and meanwhile the person/org in question would have been happy to clarify their point/edit their website/etc. if only someone had told them.

(On that note, I've sent this post along to Lucius of the GivingMultiplier team.)

Replies from: Mati_Roy
comment by Mati_Roy · 2021-01-04T11:46:17.443Z · EA(p) · GW(p)

On handling posts that may violate Forum rules:

Thanks for the clarifications.

On private vs. public communication:

I don't want to argue for what to do in general, but here in particular my "accusation" consists of doing the math. If I got it wrong, am sure other got it wrong too and it would be useful to clarify publicly.

On that note, I've sent this post along to Lucius of the GivingMultiplier team.

Thank you.