Consider two scenarios:

  1. I flip a switch, saving five lives, but leading one person to be run over by the stray train trolley. I create an "impact certificate", which sells for $M.
  2. I push a fat man off a bridge, ending his life, but saving five lives.  I create and sell an "impact certificate" for $N.

Will N and M be the same? It depends who's buying. If the buyers are ideal utilitarians (and careful assumptions are added to the thought experiment) then N=M. If, however, the buyers prefer switching to pushing, then M<N. 

An alternative hypothesis for what is going on is that these "impact certificates" will value the expected choiceworthiness of the action, rather than just its consequentialist impact. But I think this is not quite right either. Consider the following two cases:

  • An artist who lives off a $10M trust fund sells an "impact certificate" for their upcoming painting for $A. Without such funding, they would make the painting anyway.
  • A broke artist sells an "impact certificate" for their upcoming painting for $B. Without such funding, they would not be able to make the painting.

Most of us would expect the price for $B to be greater than that of $A. I think one big reason for this is that in the latter case, the patron has had a larger impact on the creation of the art - without them, it would not happen. That is, merely saying that the certificate represents the totality of the artist's creation does not make it true. People will intuitively understand that the certificate represents the act of being a patron for the art, and that for different artworks, being a patron can be a larger or smaller part of creating the value.

This suggests that an "impact certificate" might serve to represent the value of being a patron to a certain project. We have several more reasons to believe this hypothesis:

  1. If a researcher sells all of the impact for some work, then receives a prize for the work, we would generally not think that the prize ought to go to the owner of the impact certificates. This suggests that we do not realistically think that all of the responsibility for an action is transferred to the purchaser of its impact certificate.
  2. Impact purchases have previously been capped at 50%, to follow the convention that the person who does the work ought to retain some of its impact
  3. Impact purchases have been thin on buyers. Partly, this may be because the prices suggested they would receive all of the impact, whereas the social convention would only accord them part of it (the part corresponding to that of a patron).
  4. If an immoral and unpopular person sells an impact certificate for a rare positive act, people would be less likely to buy that certificate. This is partly because buyers will take into account how the seller's reputation could reflect on them, i.e. they will take into account the fact that they are acting as the seller's patron.

So let us treat as our working hypothesis that an "impact certificate" for some project will confer to its holder the responsibility for being a patron to that project. Note that this won't be particularly sensitive to whether the certificate is called an "impact certificate", a "responsibility certificate", a "patronage certificate", or something else. When you put a certificate to the market, what that certificate signifies is in the eye of the beholder, not the namer!

The "patronage" hypothesis has many interesting implications.

Certificates that signify patronage could exist not just for utilitarians, but also for deontologists, art afficionados, basically anyone. This is good news for fans of "impact certificates" because it means non-utilitarians can help them to reach a critical mass. And indeed, many kinds of patronage certificates have already done so a long time ago. For example, when a plaque outside a public designates the sponsor of that building, this is a patronage certificate in its essence (albeit a non-transferable one). When a new way of issuing patronage certificates arises, such as Patreon, or NFTs for charitable acts, we can rest assured that others will use them. That is, creating certificates for patronage is relatively easy.

A trickier question is as follows: if the certificates will be taken to designate patronage, then how can we still establish an impact economy? Insofar as the buyers are relatively utilitarian, they can still value projects proportionally to their impact. So certain sectors of the economy could behave as an impact economy, at least in that they could give equal values to M and N in the first example. But it seems harder to give equal values to A and B in the second example. This means that these certificates could not be taken to directly indicate the value of a project. Instead, they would at best value the grantmaker's contribution to that project. This suggests that impact valuations might be less informative for people running projects, relative to what we previously thought, but more useful for grantmakers.

It also offers a hint regarding how to discover more buyers in the impact economy: don't try to sell all the impact of a project. Rather, just sell the impact of patronage. If the impact of being a patron is 20% of the impact of the total project, then this would suggest that buyers could cause a project to happen for 1/5 of the price, and so the market ought to become more thick with buyers.

Fractions of Certificates. The existing convention was that patrons would purchase the impact of a project, up to 50%. What we really want is for them to purchase the impact of being a patron, which may be more or less than 100%. But how could we implement this? One convention would be to sell "the impact of being a patron". On this scheme, the first 1/2 of the certificate for being a patron of a project would be the same as the 2nd. If the patron deserves 60% of the impact of the project, then every fraction of the impact certificate would be worth 60% of its fraction of the impact. An alternative scheme would be to sell fractions of a certificate with differing prices, so that if the patron has 60% of the responsibility of a project, then the first 60% of the certificate are valued at full-value, whereas the remaining fraction of the certificate is worth nothing. The latter scheme is better for giving an insight into the value of a project (not just the value contributed by patrons), because this is encoded in the value of the first fraction of the certificate. But the former scheme is a bit simpler. I'm not sure which scheme folks will settle on, and whether we can influence it, given that ultimately the market will decide how these certificates are valued.

Directness of patronage. I have said that the "patronage" hypothesis might explain why people aren't prepared to buy all of the impact of a project. But it also might suggest that people should not by all the impact of being a patron. Consider the following three cases:

  1. A patron funds a project before it happens, paying $D for the "impact certificate"
  2. A patron funds a project after it happens, paying $E for the "impact certificate"
  3. A patron compensates someone who compensated someone who retrospectively funded a project, paying $F for the impact certificate.

I think intuitively, most of us would expect the prices to line up as F<E<D, as in case (1) the patron had the most direct effect on causing the project to happen, and in case (3), they had the least. Also, I think that I don't think it's psychologically realistic for a patron three-times removed to be able to take all of the credit for supporting a project, as would be implied by any valuation where E=F (3). After all, we will always know who really discovered it - the direct patron. In some regard, we need the price to decay with each step that we take away from the original patron. I think there are two ways to do this. Either we: (a) expect that each patron will only sell some fraction of their impact certificates on to the next, or (b) rather than allowing patrons to sell impact certificates themselves, we allow them to mint new impact certificates for their act of being a patron. Then, they can sell all of these. Either solution will create room for price decay mechanics. 


In conclusion, I think the "patronage" hypothesis explain more of what we've seen so far, and if you minted some certificates and priced them according to the impact of being a patron, rather than the total impact, you might see a much better-functioning impact economy.

NB: this post is a work in progress!


15 comments, sorted by Click to highlight new comments since: Today at 1:52 PM
New Comment

I agree with you, both that an impact-certificate market would likely end up a little decoupled from pure utilitarian impact, and that it isn't necessarily a bad thing, since there are lots of forms of altruistic (or semi-altruistic) patronage that can help the world, and non-utilitarian uses of impact markets don't prevent utilitarians from using the markets how they want.

But there are a lot of tricky edge cases when it comes to impact certificates!  I recently made this list of five "sketchy certificates" exploring various potential dilemmas: 



Here is some other recent conversation about impact certificates and various ways we might try to structure the market to avoid these and other problems.

Regarding the AMF case, your point (2) seems to be the central motivation for impact certificates, right? To reward people for having done impactful altruistic acts, so I’m not sure why you think it’s weird. Because the altruistic action is a donation instead of something more direct on the object level?

And at some future point malaria and co will hopefully be solved and the value of the 2015 donation will approach the much higher value of a life at this future timepoint (minus the contribution that belongs to others involved in the success of AMF).

I proposed renaming them to "Transferable Attribution", usually shortened to just "Attribution". I like this, because the point of selling these things really is to transfer credit, and attribute the act to the buyer.

This would make it less awkward to refer to fractions of the tokens than "impact certificates", as "attribution" is a quantity word, rather than a discrete object, you can have amounts of it. Sort of gets rid of the speedbump of having to explain how it's possible to own a portion of a token or a certificate. (I guess "patronage" does this too, though)

People seemed to like it. I don't know what we'd need to do to actually standardize on this name, though.

It wouldn't be too catastrophic if we just ended up with a decontextualized "cert".

"Transferable attribution" is exactly the right sort of idea! Or what about "responsibility certificates" or "Recs" for short? Arguably that is even better because 1) it keeps "certs", 2) "responsibility" has the connotation of being divisible, even moreso that "attribution", 3) "responsibility" gets closer to the relevant concept of "causal blame"/ "the cause of an effect" by Joe Halpern 4) doesn't have as many connotations of intellectual property as "attribution'.

Other ideas: acclaim, approval, credit, blame, or attribution certificate.

Regarding intellectual property connotations, it seems to me, at this point, that all of the systems we're making for trading impact certs would also be useful for trading intellectual property, so we might have to tolerate that.

I like "credit"... It might introduce too many ambiguities. I initially overlooked it because it also means "money"... it's also used in "carbon credits", which would exist in the impact cert system.. but another ambiguity is introduced there in that the impact cert for a carbon capture job and a "carbon credit" would be subtly different, the capture cert wont be unitized or fungible, while a "carbon credit" would be expected to be, to represent one tonne of capture, or something.

I'm resistant to "responsibility", because that represents this idea that is pretty important and fragile to the anglosphere, the intersection of authority and legitimacy, and I wouldn't want to overload that. I like the idea of using "blame" because unlike "responsibility", our culture's conception of blame is extremely unhealthy, and needs to be fucked with. Useful to promote the idea that blame can oscillate between being good or bad (asset or liability), or can be shared between many parties, and that anyone who does things in the world will tend to accumulate blame.

Interesting points. I agree that impact certs differ from carbon credits by being by corresponding to a fraction of the impact of a whole project, or at least an amount of work (inputs), rather than a quanta of impact (outputs). But it does strike me that carbon credits still might be the most closely related among well-known existing concepts. This could suggest "project credits". If you say - on your resume for example - that you sold "project credits" for a company, or a research project, it seems this would give a naive reader more of an idea of what has gone on than many other terms - they are a way of assigning credit to patrons of the project. The main downside, as you allude to is that they sort-sound like someone might be owed something. But if talking to a naive outsider, you can just say that the credit is a certificate that commemorates their patronage of the project, similar to a carbon credit, which would seem to be clear enough...

I think the problem with blame is that it sounds too negative - you won't want to write that on your resume. And if the term isn't used by recipients, then it's unlikely to catch on.

Re "transferable attribution", to be a bit more concrete, if I say that I have sold the attribution for a paper I wrote, it sounds a bit like I am giving away the authorship to a funder, which would be some kind of academic malpractice. Since that's not always the case, it seems like we don't want the general term to sound like it is...

I like it! What do you think about “(Transferable) attribution contract”? “Contract” clarifies that it’s an agreement rather than something nature-given. On the downside, it’s not as much of a quantity word. Then again I’ve been imagining certificates (just like contracts) also as the thing that lists the fractions rather than something that gets cut up.

Well I think that would be an extremely uninformative and fairly confusing thing to call it. It's only an agreement insofar as any exchange of anything is an agreement, the class of agreement is uncharacteristically open-ended relative to most contracts, and reducible to a transfer of ownership.
I'm supportive of Ryan's suggestion of "credit" at this point. The difference between "an amount of credit" and "an amount of credits" might resolve the ambiguity it might have had with carbon credits.

I don’t understand the critique of “attribution contract.” Could you try rephrasing the second sentence?

I’ve seen “credit” interpreted as “this person is to be (morally) credited,” which leads to all sorts of complications in thinking about impact markets. I suppose no one makes the mental gymnastics anymore to remember that their credit card provider is to be credited temporarily for the food one buys with the credit card, but impact markets are not as commonplace yet, and it’s something I’ve come across already. So I prefer terms that simply have something like “certificate” or “contract” in them.

A contract is also not usually transferable. It doesn't really have an owner.

But the implied analogy about the owner of the impact cert being morally credited for the work, is actually good, and clarifying. If you buy the credit for the act then you get credit for the act. Yes. That's how it's supposed to work. And if the worker wants to retain credit then they should retain a fraction of the impact cert, because selling the entire thing is genuinely supposed to mean that they relinquish all right to be retroactively rewarded to the buyer.

Contract: I was thinking of the contract as specifying the ownership, sort of like a page in the land register or the list of owners in a company that uses a paper list of the purpose. So you don’t really care about owning the piece of paper or the abstract idea of the list, but you care about owning the part of the company or the plot of land that is specified there.

Credit: It’s interesting that you find it clarifying. If it has that effect, then I suppose that works? But in my experience people who encounter the term “moral credit” or just “credit” in this context start to think about these contracts in mystical terms: “Why would I value buying and holding impact certs [gold, euros, Bitcoin, Apple shares] if I don’t inherently value impact certs [gold, euros, Bitcoin, Apple shares]?” “Are other people going to respect me for buying and holding impact certs [gold, euros, Bitcoin, Apple shares]?” These questions probably seem weird and besides the point to anyone with any of the bracketed items, but I’ve heard them repeatedly when it comes to impact certs.

One of the tests I want to do is to run the system while making the concept of the impact cert very nonobvious in the UI. I’m hoping that that’ll make it easier for people to grasp the idea of the impact market without being distracted by these mystical thoughts. But maybe it’ll be confusing again in some other way…

Can you imagine a way to get a person to engage well with an impact market (or any market) when they don't understand money/beneficial self-reifying games or whatever?

I replied to this in private, but maybe it’s helpful to put it here too:

> Dann suggested to just see to it that investing is really profitable in the beginning. :-3 That’s a bit in tension with my hope to limit various risks by [initially, for the first experiments] not attracting non-altruists to the market, but I think experiments on the EA Forum can safely be made quite profitable.

I think your post raises two points: (1) There are a lot of moral dimensions, and we need to grapple with that somehow if we want to project them onto a single price dimension, and (2) the name “impact certificate” may be confusing.

I’ll be happy with whatever name people find most intuitive, and we’ve been doing okay as a species despite confusing names such as “koala bear” (not a bear) or “wombat” (not a bat). So I’ll ignore 2.

But 1 is crucially important in my opinion. The failure mode is the following: A charity offers universal basic meat deliveries to all households in Country. They’re funded mostly from people who speculate on their impact certificates (one per batch of national deliveries), and have even sold some to retroactive funders already. The valuation is enormous. They can scale up the amount of meat they can deliver.

But that is just because the nonhuman animal suffering is not priced in. The animals don’t participate in the market, and if they could, they wouldn’t be able to profit off any short positions as much as the humans can profit off their longs (esp. after adjusting for the liquidation risk).

I see two big categories of responses to this: (1) force people to put collateral at stake when buying any (including long) positions in the market or (2) commit everyone to norms for the valuation of impact that make moral defection like that unattractive for issuers and speculators.

I don’t have much hope for the first because the amounts of collateral would have to be enormous. But in this section of a big draft of a future EA Forum article I propose a definition that I hope will have the right self-reinforcing feedback loops built in that it can become a Schelling point for how impact is valued on an impact market – especially once the biggest retroactive funders adopt it. I also explain it in this recent talk at the Funding the Commons II conference.

I think this is a really big deal, so I’d be delighted if anyone wants to red-team my proposed solution!

I see your point from one perspective. I agree that people don't just look at impact.

But neither does EA more generally, and still we say things like that EA is about doing the most good or impact-maximisation. We do take constraints into account, but maximise impact given those constraints. Therefore it iss in my view reasonable to say that EA wants to maximise impact.

Analogously I think it's reasonable to use the term "impact certificiate" even if impact isn't, in one sense, literally your only criterion. (Also you could of course argue that taking, e.g. reputational concerns into account is actually impact-maximising if you factor in indirect effects. Cf. the debate on naive vs sophisticated consequentialism.)

To me, "patronage certificate" - while logical from one point of view - doesn't sound particularly inspiring. It makes the patron very salient which I worry leads thoughts in the wrong way.

By contrast "impact certificate" sounds more inspiring, and more in line with effective altruism's core values. I don't think it is as utilitarian a term as you seem to suggest.