Impact Certificates on a Blockchain

by RowanBDonovan33 min read3rd Aug 202126 comments

33

Certificate of impactBlockchain
Frontpage

Summary

We can create liquid markets for public goods by porting impact certificates to a blockchain. The decentralized exchange Serum appears to be well suited for the purpose. I can see no technical barriers.

But the legal situation in the US and UK (where most EAs are) seems complicated and risky. Some tweaks to the idea of the impact certificate might enable someone in a country with more advantageous legislation (like Switzerland) to operate the system. I’m based in Switzerland, conveniently.

Investigations of the legal situation and the building of various software tools that I envision will take a lot of effort, time, and a bit of money. So before I go forward, I want to check that (1) there’s sufficient demand for the system to warrant the costs, and (2) I’m not making any deleterious, irreversible mistakes.

I Need Your Help

  1. Are you planning to use these tokenized impact certificates? If so:
    1. Do you want to issue certificates? For what project?
    2. Do you want to buy certificates? What would be your yearly budget?
    3. Do you want to trade or arbitrage certificates or use them in other creative ways? Can you guess some metric, such as daily volume, to give me an idea of the magnitude of the engagement?
  2. You can create a token for your charity or project right now. (Please read the section on legal risks first.) If you’ve considered it but you’re still hesitant, what are the reasons? Risks, missing features, or something else?
  3. If there were a certificate that represents the impact of the whole field that you work in, would you accept it as a donation? Would you hold it or would you sell it within months?
  4. Which whole field of altruistic work are you most excited about? If there were a certificate that represents the impact of that whole field, would you buy it? How much would you invest?
  5. Do you notice mistakes in my thinking or gaps in my knowledge that you can correct/fill?
    1. In particular, do you see downside risks that I’ve overlooked, i.e. risks that are not merely like a failure of the project but create net harm?
    2. Do you think the better approach would be for an economist to first publish something like a whitepaper on the risks and opportunities from tradable public goods?

It'd be wonderful to get frank feedback in the comments so I can decide how to prioritize this project among some other options that I’m considering![1] Thank you!

Problems

Public Goods are Illiquid

Public goods are currently very illiquid investments: Donating not only comes at the cost of the money you donate but also at the cost of the option value to donate it elsewhere later.

I’d guess that I’ve improved my giving by a factor of about 100 between 2012 and 2015.[2] Most other investments of mine are liquid enough that it takes me just seconds or days to sell them in such a case.[3] But in 2015, I had no way to sell out of my investments from 2012–14 anymore to shift my impact portfolio toward my new preferred allocation.[4]

This illiquidity of the market for public goods is probably part of the reason that people, even altruistic people, are much more hesitant to invest the same amounts of money into charities as into publicly traded companies. The first is final; the second is more a form of parking money with minor risks owed to price volatility.

As a result, large funders can’t park their money in conservative assets like GiveDirectly, Oxfam, or the Global Fund until they find a better opportunity; they need to park it in for-profit companies, most of which probably perform worse on the bottom line that the funders care most about.

Public Goods are Treated Like Consumables

A reason for this illiquidity may be that there isn’t a general understanding that you can resell investments into public goods or that there’ll be buyers. It’s as if they were consumables that the donor eats right after buying them. Yet some interventions that we can support have lasting effects – be it cash transfers for just a few years, a brief period of reduced catastrophe risk that has an effect on the aggregate catastrophe risk in any period that contains it, or changes to knowledge and culture that may even compound over time.

Before I buy a share in a public good from someone, I’ll want to know that there is no one who can produce or draw on an arbitrary amount of it at a smaller cost than what I’d pay, that I can prove that I own the share in the public good, and that there’s a somewhat liquid market for shares in this public good. If some of these are only somewhat lacking, I might trade that off against any additional utility that the proof of ownership might have – dividends, voting rights, access, etc. These desiderata are not currently fulfilled except in unusual cases.

Public Goods Fail to Reward Price Discovery

If I do research and thereby receive private information that a for-profit venture promises to be very profitable, I can invest in it (or found it) and get rewarded for my prescience when my shares appreciate in value. Therefore I’m incentivized to do such research and can refinance the cost of the research. The same is true if I’m ahead of the curve in discovering that an existing project is overrated and I short-sell it. This mechanism gradually increases the power of those who can make good predictions and decreases the power of those who don’t.

This is very hard to do with public goods at the moment. You can invest into some plant-based and cell meat companies (or some people can), but if this were true across the board, for example with wild animal welfare and insect welfare, then people like Brian Tomasik would be much richer than they are. Shorting playpumps and Scared Straight might’ve also been profitable. (Make-A-Wish might’ve been a less profitable short since its share price would’ve been held up by people with somewhat different values.)

A central limit order book is also anonymous. If you think a project is badly run or follows a bad strategy, you can short it, and no one will know. But everyone will see the decline in the share price. This avoids all the social awkwardness of gossiping about it in a way that warns them away from it without making you seem cynical, jealous, or vindictive. If you are, and that’s the only reason you short it, chances are you’ll lose money.

Other Problems

  1. The benefits of working for vs. donating to charities are often nonobvious. You can’t easily see at what size of a donation a charity would value your work for them equal to the donation. This depends on the supply they have access to for the particular type of work and the degree to which they are funding constrained, both of which are not routinely published in job ads or donation appeals. There is also no marketplace where you can track these prices in real time and compare them between charities.[5]

  2. Public goods can evade democratic oversight. Some charities have a lot of leverage over the future. (80,000 Hours has a list of pertinent risk factors.) They are mostly using it thoughtfully at the moment, or that is my impression, but the risks remain. My plan is to empower charities much further, so it’ll become even more important that they serve some sort of causal or acausal compromise morality rather than burn resources in Pareto-inefficient battles. Democratic oversight will also be important if charities start to act like decentralized governments to provide government services to people worldwide, especially those who don’t receive them from their country’s government.

  3. Altruists often feel bad about reinvesting. Many altruists are hesitant to reinvest in themselves rather than “generating impact” right away, so they neglect their mental health or are impatient with their learning or training progress. But if profits are the for-profit equivalent of impact, then it stands to reason that if many companies reinvest all their profits to grow faster for many years, it’s probably a perfectly good approach that altruists should consider too. Companies have other metrics, like revenue, to track their progress, but altruists may lack them, which can feel like stagnation.

  4. Impact investing is often not very impactful. Impact investing is held back by the inability of fund managers to invest into the organizations that have the biggest altruistic bottom lines because those tend to be nonprofits. Just like it’s hard to create a low-volatility cryptocurrency fund that mixes cryptocurrencies or tokens that are correlated with Bitcoin with such that are anticorrelated (because they don’t exist), it is hard for managers of impact investment funds to put together a fund that is reasonably strong on the profit and the impact bottom line.

  5. Funding is difficult to scale down. Funders seem to be doing a good job funding big and highly promising projects. But it seems plausible to me that there’s a long tail of small and mediocrely promising projects that yet are unlikely to do much harm if they fail. At the same time some funders have trouble allocating their funding because the big and highly promising projects are already well funded.

Solution Concepts

I want to draw on the idea of impact certificates to create a form of certificate that does not represent any sort of asset or monetary flow (e.g., a hotel building or fundraising revenue) of a charitable project but its all-things-considered impact on the world.

That’s impossible to measure with certainty, we don’t know if it’s defined at all, always infinite, or whether the first single-celled organism is to thank and blame for everything, but if thoughtful altruists are involved in the market and either own a lot of certificates or want to buy a lot of certificates, other traders will want to predict what these “impact whales” think is impactful. That can be as easy as reading about it on their websites. Hence the price will hopefully roughly track the enthusiasm that thoughtful altruists feel for a given charitable project, which is probably as good as it gets at the moment when it comes to approximating whatever it is that we care about.

These certificates can come with additional perks so that they’re also interesting for people who don't just want to speculate on their price. Owners of the certificates might have a say in certain decisions, especially those that concern moral preferences rather than predictions, because the experts of the charity are usually better at making predictions in their area of expertise. (Without certificates, the charities would likely run polls to get a feel for such preferences.)

The marketplace itself could also reward people for buying and holding certificates. It might have its own certificate and pay a slowly decreasing fraction of it to people who hold other certificates.

There are two practical hurdles: Clarifying the legal situation and creating a liquid market. I have three ideas for addressing them.

Charity Shares

Here it is a charity that creates its own certificate and offers a fraction of it for sale. Say, they might create 100,000 shares, offer 10,000 for sale, give 20,000 to employees with a vesting period, reserve 20,000 for contingencies, and lock the other half up for several years.

That probably works fine in some countries, but without professional legal advice (for at least the US, UK, and Switzerland) it is, at this point, hard for me to tell what registration requirements a charity would have to satisfy to be allowed to issue a certificate.

US federal law draws on the Howey test to determine whether something is a security. Four factors need to all be met for something to be a security: It is “(i) an investment of money, (ii) in a common enterprise, (iii) with an expectation of profit, (iv) solely from the effort of others.” (Source) (Some states, notably California, apply different tests.)

That implies to me that charities would have to register their impact certificates with the SEC regardless whether it represents a share in something monetary or in their impact. Such a registration is said to take some 1–1.5 years, so charities are unlikely to do it. Some nonprofits are exempt, but they adhere to requirements that prevent most forms of trade of the shares.

These legal constraints can make charity shares unappealing, at least in some countries, or they might give rise to a new type of charity, a decentralized autonomous charity, that is not geographically located, is run by a fully anonymous team, and is governed by smart contracts.

Project Shares

In the case of what I tentatively call “project shares,” it’s not the charity that issues the certificate, but an individual employee or several employees who work together on a project (an unincorporated project). Say, Ajeya Cotra’s project to investigate AI timelines. A number of charities are distributed, so these employees may be in countries where it is easier to issue certificates than in the country where the charity is registered.

The project will also be smaller than the whole charity. Some countries have thresholds in terms of the revenue from the sale of securities below which no registration is needed. The project could be sized such that these requirements are met. If they even count as regulated securities in that country that is.

Intervention Shares

Another option is to issue certificates not for the impact of one charity or project but for the impact of a whole “space,” that is, of a whole group of charities that work toward a common goal or even work toward that goal using similar methods. An “AI safety” certificate might be conceivable or a “technical AI safety” one. The boundaries of the “spaces” would be fairly arbitrary.

Anyone could issue such a certificate. That person or organization could be based in a country with laws that make it easy to issue it. (I could be that person if Switzerland turns out to be a good location for that.)

The creator would put together a committee of trusted, thoughtful altruists to decide on the initial allocation of the shares. Half of them may be retained and the other half split among trustworthy organizations within the space, which they’ll be donated to.

There could be, e.g., quarterly events where a new organization can receive a fraction (which decreases every time) of the remaining shares. The shareholders would vote on who the new awardee will be, and the issuing organization or a smart contract will donate the shares accordingly.

I will need professional legal advice to be more certain about this, but it seems to me that this solution has the advantage that it’s legally unproblematic for the charities to receive the shares donated to them. (They’d behave like any other crypto donation.) There is also the additional perk that the value of the shares each charity holds will depend on the success of all other shareholders, so that charities are incentivized to support each other and not fall prey to the “narcissism of small differences.”[6]

Implementation

I prefer to follow a “lean” strategy where I start with a minimal viable product (MVP) that is hopefully just good enough that it can tell me whether people will want to use it, i.e. whether there is a market for it and whether the product market fit is at least on the right track, or whether I should change something about my approach fundamentally or prioritize another project entirely.

Proto–Minimal Viable Product

My first attempt at an implementation was based on a Google Doc that stipulated the terms of the certificate and a spreadsheet that tracked the ownership of the shares. The version history of the documents made it transparent what changes have been made to the allocation. Monetary transactions (bank transfers) happened based on trust.

Minimal Viable Product

That system can easily be improved upon. Blockchains are designed to serve the purpose of the version history, and in recent years tokens (called, for example, ERC-20 tokens or SPL tokens depending on the blockchain technology) have become a popular way of representing securities, credits, or votes. They are perfect for our purposes. Creating SPL tokens takes a few minutes and costs mere cents.

Furthermore, Serum (based on the Solana blockchain) is an exchange that supports order books like a normal stock exchange but is completely decentralized, i.e. there is no central company that decides whether you get to list your asset on Serum. It currently costs around $100–200 to create a new market on the exchange. Then a founder can set up limit orders to implement a bonding curve auction or get in touch with Raydium, a Serum-based automated market maker. (A specialized auction platform is also in the making.)

So not only can we easily create tokens to represent our shares, we also have the technology at our fingertips that lets us create a streamlined market to trade them. (And the project is managed or advised by Sam Bankman-Fried, an EA who started out earning to give at Jane Street and has since become a major crypto entrepreneur.)

Here is a fairly nontechnical guide to creating and listing your own token. Be sure to assess your legal situation before offering it for sale.

Future Plans

Incentives

It’s imaginable that those who would typically donate to charities would also be ready to buy shares in the charities’ impact from the charities. But why would anyone buy shares from anyone other than the issuing charity? And why would purely profit-oriented investors buy impact shares? Arguably most of the best-case impact of the project will be in attracting enormous amounts of for-profit funding for charities, so that’s important.

All of this becomes a nonissue once there is an established, liquid market, because then people will want to profit from increasing prices. So it’s a bit of a chicken and egg problem.

I’m hoping that tokens will eventually find use for all sorts of purposes that are interesting for holders: Charities can use them for governance, i.e. let impact shareholders decide on matters that are relevant to them and that they know more about than the charity, such as where and how to hold the next general assembly. Intervention tokens in particular can govern their own distribution. I’d also be interested in a platform that allows for some form of “staking” of the impact tokens and pays out a reward for that. Finally, one could develop a mechanism whereby holders can quasi-burn their shares in a verifiable way while still being identifiable as holders; they could thereby forfeit their ability to ever sell the share again and decrease the future supply that others expect.

I’m hoping that such immediate uses will make the tokens interesting for enough buyers that we can establish markets that, in turn, make the tokens interesting for pure for-profit investors.

If no such market wants to emerge, buy-and-burns from platform fees may help, though I haven’t thought much about the precise mechanics. A platform that runs a Serum instance could generate fees and use these fees to buy back impact tokens, specifically intervention tokens. It could, for example, buy them back at the market price proportional to grants from the EA Funds, the Open Philanthropy Project, or some other sophisticated funder.

If the funder makes grants of $1 million to A, $2 million to B, $4 million to C, $8 million to D, and $16 million to E, and A and B are charities within the ABBA intervention ecosystem, and A, C, and B are within the ACDC intervention ecosystem, and there are no other intervention tokens, then $(1 + 2 + 4 + 8) million = $15 million are going to charities within defined intervention ecosystems. But for the sake of the calculation we need to double-count A, because it is in both ecosystems, so the new, artificial total is $16 million. ABBA accounts for 3/16th of it; ACDC accounts for 13/16th. Hence, the platform would allocate 3/16th of the fees to buying back ABBA tokens and 13/16th to buying back ACDC tokens.[7]

It would be straightforward to then burn these tokens, but alternatively they could also be locked up to be allocated to new charities that get voted into the ecosystem.

Ecosystem

If this project gains further traction and the legal questions can be resolved, I’m planning to create, or encourage others to create, a series of tools:

  1. A curated gallery of certificates that details the properties of the token, explains the purview of the impact it represents, and lists the current owners. This gallery helps to make impact certificates differentially more interesting for value-aligned altruists as opposed to, for example, terrorists. I’ll be sure to interpret “value-aligned” to mean “morally cooperative,” “thoughtful,” and “prosocial” and not “agrees with me.” I’d also be happy to delegate the curation to an elected team of experts.
    1. An MVP of this gallery would probably take under a week to implement. It could be more fully featured within a few months.
  2. A program to manage the governance of the “intervention shares” so that no one needs to be trusted with up to 50% of the total supply of the tokens until they are distributed. The program code could be audited so that even investors without a technical background can trust the program. No one but the program would have access to the tokens. 2. This program would be a larger project because I’d have to learn a lot more about Solana and Rust, but I can probably realize an MVP within a few months. Such a program should be audited, which may be costly.
  3. A platform to allow owners of tokens to “stake” them and receive platform tokens as reward. An added incentive to buy and hold tokens. The staking rewards could be influenced by the votes of the owners of the platform tokens. 3. This platform will take longer to implement than both previous steps, though an MVP might also only take a few months. I don’t know whether I can build this on top of audited code or whether I’d have to get it audited too.

Risks

There are a number of risks inherent in this project beyond failing to find traction. I detail specific risks below.

Generally, I could test the whole concept and system on lower-risk spaces, such as athletic disciplines or music genres. Some people care deeply about barefoot bouldering or extratone music and are probably not earning a lot of money with it, so they may welcome some donated tokens. And if the idea backfires there, at least it’s not the whole future that’s at risk. But I don’t have the knowledge or connections to (reliably) identify the experts in these fields.

A compromise solution might be to test it on mental health. People with mental health problems in legislatively advantageous jurisdictions, for example, may appreciate the ability to not have to beg for comped therapy but to be able to sell their own mental health progress to people who value it directly. An added perk is that all transactions are anonymous so that no one needs to fear any reputational damage or stigma that they might be afraid may arise from revealing their mental health status.

Even legal experts will have difficulty interpreting the applicable law with confidence, and the various differences internationally complicate the situation further. It seems to me that the maximum penalty for making a mistake here is imprisonment for up to five years in the US and up to two years in the UK. I don’t know what the sentences tend to be in practice in such cases (problems with the SEC that don’t involve fraud) or whether fines are more common, but I’ve heard that settlements with the SEC are common.

In any case, it would be devastating for (especially but not only) AI safety if key people had to go to prison for a significant fraction of the time that I think we have left to solve the problem, so even a small risk of that is likely going to be too much.

The legal situation in Switzerland seems more relaxed, which makes me optimistic about the intervention shares. There doesn’t seem to be a registration requirement for securities, and the penalties for failing to disclose information are mere fines, and even those should be easily avoidable because they are penalties for failures to comply with a request, not mistakes in proactive disclosures. Meanwhile there is even an exemption from the requirement to apply for a banking license so long as you accept less than CHF 1 million for token sales. Since I’m only ever planning to donate the intervention shares, that revenue should be 0, and eventually I want the funds to be managed programmatically anyway.

Another option is to use the relaxed legal environment in Switzerland to issue charity or project shares on behalf of others and then donate them to the respective charity or project. But I don’t know whether that even makes a difference legally.

Meanwhile, confusingly, there are a lot of crypto projects that issue security tokens seemingly with very little regard for the legal frameworks. In some cases the founders are not even anonymous,[8] and the majority of these projects are probably not intended as scams. It might be that they’re all based in countries with permissive securities law, but that seems doubtful in view of the strong representation of US Americans in the cryptocurrency space.

I work only part-time and so only have a budget of about $4,000 this year for legal costs. That’s a fraction of the retainer of a US lawyer. But I don’t know how much research it takes a lawyer to answer these questions. With most EAs (according to the 2019 survey) living in the US and UK, I want to focus on these countries first. (In addition to Switzerland, where I live.)

Opportunity Costs

I originally worked on a different project when the arguments of Ajeya Cotra and Daniel Kokotajlo gradually convinced me that transformative AI was likely closer than I thought by a factor of 4 or so. I reprioritized and found that the project I had been working on was indeed not maximally pressing in view of this urgency.

But the project (this one) that came out on top, largely carried by its apparent traction, doesn’t promise to do much to avert existential and especially suffering risks from transformative AI either. At least not at the margin because additional liquidity is not often cited as a major bottleneck that is holding back AI safety research.

Do you think I’m wasting my and others’ time with this project, and that I should be doing something else?

General Use

One red flag for me is the general usefulness of impact certificates and the technology around them. If token sales become a common fundraising mechanism for nonprofits, they can also be used by terrorists or, less egregiously, by charities whose methods are likely counterproductive.

I’m not considering this a prohibitively bad problem since (1) the technology already exists, (2) my additions to it will be tied closely to a particular community of thoughtful altruists, and (3) I’m hoping to counterweight the uncontrollable nature of the decentralized finance ecosystem that I’m building upon with a curated, centralized gallery of recommended projects.

Taken together, these factors will hopefully prevent my work from appearing attractive to people whose projects are recognizably counter to cooperative, prosocial ideals.

Speculation

I’ve gotten the impression that the cryptocurrency space is dominated or at least strongly influenced by people who were selected for being early adopters (and “hodlers”) of Bitcoin. Being an early adopter is likely uncorrelated or imperfectly correlated with a commitment to value investing. That appears to me as a possible explanation for why the price of Bitcoin reflects the games of price speculators much more than its intrinsic value, and why cryptocurrencies with (in my layperson’s opinion) much greater intrinsic value, different purposes, and different implementation – such as ETH, SOL, ADA, DOT, etc. – are correlated with it even on the timescale of minutes.[9]

The art market might be even more rife with price manipulation and speculation to the point where the price is wholly divorced from the quality of the piece (as it would be assessed by an efficient quality-of-art market).

That could happen to impact certificates too, in which case they’d be worthless for thoughtful altruists (except if they trade them as a form of earning to give). The price might then be dominated by Keynesian beauty contest–like dynamics or Fibonacci retracements regardless of whether the certificate was meant to represent the impact of playpumps or CEA. This will most likely just vitiate my vision and not cause net harm beyond opportunity costs, but it might be that an attacker with a lot of capital or flash loans or good timing can use market manipulation to usurp power over a space that uses intervention shares, and then vote to have each future token airdrop funneled to their wallets. That may well end up being net negative.

I’m hoping to address this by making those who I consider experts most powerful at the start by giving them the bulk of the shares (in the case of the intervention shares) or heeding their opinion when it comes to listing a project in the gallery. That will hopefully ground the price in a thoughtful assessment of the impact and limit the extent to which price manipulation can be successful.

Appendices

I’ve moved a few details here that seem too specific for the main text. Feel free to skip.

Terminology

Impact certificate, certificate, stock, token: A few synonymous ways in which I refer to a contract that represents the impact of a project. The term token is only applicable in the context of blockchains.

Shares, tokens: A small fraction of the certificate. Yes, I find it confusing, too, that token refers to both the whole and the part. The whole is often used in the singular and the parts often in the plural.

Charity shares, project shares, intervention shares: I expect these terms to change. I find the analogy of “like a share in regular exchange-traded stock but of impact” very helpful. The term “certificate” doesn’t evoke the same associations. Then again it may be necessary for legal reasons to avoid terms that are used for regular securities.

Peer to Peer Funding or Vetting

There’s the idea[10] that we could cut down on vetting overhead by trusting trusted people to regrant grants to people they trust. In the simplest form, we’d have a central grantor, let’s call it the regranting bureau, that distributes delegatable votes to the people it trusts. These people delegate some of their votes and use the rest to make concrete grant recommendations. The central grantor then just carries them out.[11] It would coordinate money transfers with recipients, do all the accounting, etc., and it could also add some bells and whistles, such as weighing the grants by how often someone has been recommended and by who (e.g., using the Page Rank algorithm), and checking for cycles.

Blockchain-based impact certificates might streamline this process, but they’d also come with side-effects whose desirability will need to be assessed.

Instead of having a regranting bureau, you’d have only (1) an initial set of highly trusted regrantors, say, Alice and a few others; (2) public good tokens such as “Alice’s Small Cap AI Safety Impact” (ASCAISI) that is initially owned only by the respective regrantor, Alice in this case; (3) funders who place buy limit orders on the ASCAISI/USDC market and all the other *SCAISI/USDC markets on Serum at whatever prices they think are fair, (4) random EAs who analyze the regranting networks to see what happened and how successful they think they were. Grantors would be held to keep a small share – Justin suggests the square root – of their tokens for themselves for their efforts. (Others could base airdrops on a Page Rank–like analysis of the regranting network since it’ll be public.)

That would (1) save the overhead of setting up a central regranting bureau, (2) outsource the auditing to people who have no reason to be biased about the system, and (3) give grantors an incentive to optimize hard for the tradeoff between making many grants and making good grants so the funders are happy to pay more for the tokens either because successful “hits-based” grants or because of successful conservative grants, (4) maybe “gamify” the process for regrantors in that they can compete to regrant such that they have the most highly-valued token, (5) keep the system open to any small donors who want to contribute by also buying some tokens too, (6) maybe even keep the option open that the system will attract speculators whose funding counterfactual is much lower than that of the funders.

But that would come at the price of a last-minute veto option of the regranting bureau. Funders would have no way to selectively not make certain grants other than to withdraw their orders from the market of the original regrantor.

I’ve briefly looked into the legal constraints on charities and individuals with a bit of a focus on the US. I’m not confident at all in these findings.

Charities

US nonprofits can issue securities and may be able to apply for an exception from a SEC registration, but there are various requirements that they have to fulfill at the federal and the state level. (One source even said there may be requirements at the municipal level.)

Some of these requirements only restrict stock buybacks, which may be avoided by asking a big donor to do them instead or burning tokens the charity still holds and thereby reducing the fully diluted supply instead of the circulating supply, which could also have an effect on the price.

But other requirements seem to restrict transferability (same source): “In numerous subsequent no-action letters, the SEC staff has consistently agreed that memberships that entitle the members to no share in the profits and are not transferable do not constitute securities under the Securities Act.” (Italics mine.) I see no way to realize charity shares while restricting transferability.

The “Regulation Crowdfunding” legislation is probably not applicable because it requires that the sale goes through a registered intermediary, which Serum is not.

Projects

By projects I mean such things as Ajeya Cotra’s work on AI timelines. They cost money that needs to come from somewhere but they’re unincorporated.

One entity that can automatically come into being by operation of the law and that is common in common law systems is the trust. This should be relevant to a number of countries, the US and parts of the UK among them. I can’t quite pin down who the settlor, trustee, and beneficiaries are in our case because it depends on whether the corpus of the trust is the money used to purchase the share in the impact, the share in the impact, or the material form of the impact.

Another possibility is that the sale forms an unincorporated association because the terminal purpose of the project is to generate positive impact and not profit. However, individual investors may invest into the project for the purpose of generating a profit. I don’t know whether that prevents the whole of the project from being an unincorporated association.

It seems less likely to me that the sale forms a general partnership or joint venture because the investors are generally anonymous and there is no expectation that they agree to be liable for any debts of the project.

If future impact can be considered property, project shares may also simply be a case of co-ownership. And finally, all these legal entities seem to be particular, somewhat standardized forms of contracts, so there may not be any way to narrow down what is happening legally further than to call it a contract sui generis.

In each of these cases the question becomes whether such a project trust or project association or project partnership requires registration in a given jurisdiction and is subject to any relevant restrictions.

Investors

As mentioned above, the “Regulation Crowdfunding” legislation is probably not applicable because it requires that the sale goes through a registered intermediary, which Serum is not. Outside of this exceptional case, it seems that only “accredited investors” are allowed to buy securities, i.e. millionaires, some businesses, people with a high income, or those who have particular licenses.

One option might be to start or partner with such an intermediary which could custody shares on behalf of US citizens, but it is unrealistic for me to attempt that, since I’m not a US citizen or resident.


  1. For example collecting, prioritizing, and roadmapping of longtermist project ideas to make them “shovel-ready” for suitable entrepreneurs. Or creating systems for better Bayesian models for futures studies and priorities research. But don’t glorify tokenized impact certificates only to prevent me from working on these. If you think they’re bad ideas, just tell me! . ↩︎

  2. By the lights of a proxy metric I favored in 2015. The earlier donations may have compounded faster than the money would have, but probably not at a rate that would invalidate the argument. ↩︎

  3. That’s different for people who own the majority of the shares in something because they would greatly reduce the share price if they sold much of it, but that is not the case here. I was rarely among the largest donors of any particular charity. ↩︎

  4. Some of my 2012–14 investments even increased in popularity, so I could’ve made a profit selling them. ↩︎

  5. Owen Cotton-Barratt inspired this thought. He already has quite a detailed solution concept for organizations that want to issue impact certificates. ↩︎

  6. I don’t observe animosity as a problem in our community, but I think some organizations could profit from talking more with other organizations, be it just so they don’t duplicate effort or, conversely, steer clear of an important intervention because they falsely assume that otherwise they’d be duplicating effort. ↩︎

  7. The crude design would also limit us to occasional (e.g., yearly) big purchases of tokens. It would be nicer to spread these out over the year, say, in hourly intervals. It might be possible to use a prediction market for this. This way, people would try to predict the total funding allocation from a particular funder to an intervention ecosystem. This market could be renewed every year or it could work more like a perpetual future with funding payments. Funding payments are usually much more frequent, though. What would be even nicer are predictions based on a metric that removes the influence of funding gaps. Sophisticated funders usually only grant to organizations that they think are impactful, and exceptional cases, such as when a funder wants to buy influence over a potentially dangerous organization, are probably already weeded out through the ecosystem voting process. But within that elite set, the sizes of the grants are probably as much or more informative of funding gaps than expected impact. If funders instead estimated the relative impact of the first dollar anyone ever donated to an organization using all the information they now have thanks to partial hindsight, funding gaps would probably skew the picture only very slightly. There might be even better operationalizations. ↩︎

  8. Rowan Donovan is a pseudonym. ↩︎

  9. It seems implausible to me that this correlation reflects the value of the whole cryptocurrency space, or that it reflects it well, because minutes seem like too short of a timeframe in comparison to something so vast and vague. But what do I know. ↩︎

  10. I’ve come up with a similar idea, and Justin Shovelain and Matt Goldenberg have, also independently, come up with a more detailed concept. ↩︎

  11. I know of no jurisdiction where this could be done while maintaining tax-deductibility, except maybe with extreme bureaucratic overhead, so I’ll ignore tax-deductibility in the following. ↩︎

33

26 comments, sorted by Highlighting new comments since Today at 4:34 PM
New Comment

In particular, do you see downside risks that I’ve overlooked, i.e. risks that are not merely like a failure of the project but create net harm?

Related to what you wrote under the "General Use" section, I think we should consider the risks from funding "very risky altruistic projects" that are actually net-negative, even though they have a chance of ending up being extremely beneficial. The root of the problem here is that certificate shares can never have a negative market price, even if the underlying charity/project/intervention ends up being extremely harmful. So from the perspective of a certificate trader, their financial risk from their purchase is limited to the amount they invest, while their upside is unlimited. In other words, the expected future price of a certificate share (and thus its price today) can be high even if everyone thinks that the underlying charity/project/intervention has a very negative expected value.

Is it possible to make it so that the estimation of the share value, from the perspective of certificate traders, will somehow account for the historical downside risks of the charity/project/intervention? (Even if by now the downside risks no longer exist and the charity/project/intervention ended up being extremely beneficial.)

Hmm, this seems tricky. In principle, not only should the price be allowed to go negative, it should not be bounded below at all, and certificate holders could have a debt for holding certificates with negative value.

Is it possible to create certificates for the negative of their expected impact, too, and tie them together in the right way?

I'm thinking like in some betting markets, if you hold 1 share for each of the outcomes, you can destroy them and get their total value, e.g. $1 by design in some markets. Or like perpetual futures in crypto.

Maybe you would also need to put up collateral to participate in these markets, and holders of the regular certificates could be forced to pay interest to holders of the negative certificates when the (net) price is negative or get margin called (and vice versa).

Token lending seems like a really good thing here since it would allow people to short and it would generate passive income for hodlers. Since the tokens will be fairly obscure compared to BTC or USDC, the interest will be high, provided there is any interest in shorting in the first place, right? So the shares of controversial projects like playpumps will yield high interest but will remain cheap while, say, AMF shares will yield low interest but appreciate. I’ll need to think about whether that’s good. And what the interest currency should be.

Perpetual futures would be another proven solution to this problem. Bonfida already offers three of them based on Serum, so they seem like an obvious partner to try to get on board with this. Moët is another one but hasn’t launched yet. Considering that Bonfida only has three and they have low volume, I suspect that it’ll be hard to get this off the ground…

A completely out-there idea: Instead of just having one dimension – price up or down, crudely representing positive impact – you could have a multidimensional market. A charity could get their token listed on markets of happiness/USDC, suffering-reduction/USDC, eudaimonia/USDC, edification/USDC, progress/USDC, existential-safety/USDC, experience/USDC, etc., so a token/USDC market with many dimensions. Then traders could buy or sell arbitrary vectors on that market, and their PnL could be something like the cosine similarity to the current multidimensional market price times their investment. But that would require a lot of liquidity since an order can’t go through if even along one dimension there are not enough offers at the price. In any case, just a random phantasy, not actionable for me. xD

That's an interesting line of thought. Some potential problems:

  1. There may be a wealthy actor that doesn't like a certain net-positive intervention (e.g. because they are a company that tries to avoid the regulation that the intervention aims to impose). Such an actor can attack the "positive shares" by buying all the "negative shares" and then artificially making their price arbitrarily high (by trading with themselves).

  2. A more speculatively concern (not specific to your idea): Suppose that most traders would believe that: conditional on an existential catastrophe happening, owning the right certificate shares is less important. This may cause traders to give less weight to downside risks when making their decisions. (RowanBDonovan mentioned this issue here.)

Could 1 already be done? They could short the impact certificates, and even inflate the certificate price for oppositional work by trading it with themselves. An efficient impact certificate market should have the latter cause downward price pressure on the certificates they want to short, right?

This leads to an interesting question: how should certificates/organizations working exactly against each other be priced together in an efficient market? Presumably their market caps should be negatively correlated. Would an efficient market ensure one has a market cap near 0 (and so be drained of resources from the market)?

And then maybe this gives us a solution to the original problem without the need for artificially introducing something like negative impact certificates: if the expected value were negative, negative certificates can be introduced naturally and someone can start an oppositional organization or otherwise start doing oppositional work.

There might be reasons to think one side is more efficient than the other at achieving their desired outcome, though. I'm not sure what implications this would have.

EDIT: Previously brought up here.

I think "oppositional work" can't always serve as a way to mitigate the harm of a net-negative projects (e.g. it doesn't seem obvious what the "oppositional work" is for a net-negative outreach intervention).

Simply shorting shares doesn't seem to me like a solution either. Suppose traders anticipate that the price of the share will be very high at some point in the future (due to the chance that the project ends up being very beneficial). Shorting the share will not substantially affect its price if the amount of money that participating traders can invest is sufficiently large.

Oh wow, yes, thank you! That disconnect between distributions – one symmetrical, the other roughly log-normal – strikes me as very important and dangerous!

This “ex post penalty” that you suggest would be great… I’ll think about things like offering standard (non-inverted) perpetual futures to make it easy to make a lot of money by shorting risky projects; investing through shorting tokenized public bads or world suck; making robustness against downside risks part of the intervention share governance; and whatever else one might do against this problem.

Re the shorting related ideas: maybe you're thinking about mechanisms that I'm not familiar with, but I don't currently see how these approaches can help here. Certificate shares for a risky, net-negative intervention can have a very high value according to a correct fundamental analysis (due to the chance that the intervention will end up being very beneficial). In such cases traders who would "bet against the certificate" will lose money in expectation.

Yes. I think we have different types of asymmetries in mind here. I can see three types at the moment, but maybe there are more that I’m overlooking? How do you define “net-negative” if not in terms of expected value? Stochastic dominance? Or do you mean that the ex ante expected value of an intervention can be great even though its value is net-negative ex post?

Different asymmetries that come to mind:

  1. Investors should be able to short just as easily as they can long, and their profits from correctly predicting downside should be just as unbounded as their profits from correctly predicting upside. This can be approximated with borrowing and lending or with a perpetual future. The first approach requires that someone offers the tokens for lending and that the short trader is ready to pay interest on them. The second only has minor issues that I’m aware of (e.g., see Calstud29’s comment), so that, I think, would be a great feature. Then again the ability to lend tokens that you have would create an incentive to buy and hold.

  2. Profit-oriented investors only care about profits that they make in futures in which they can spend them. So if a trader thinks that a project is vastly net negative, shorts it, then the news spreads (maybe aided by the proof that traders are putting their money where their mouth is, namely in a short), and the project is discontinued before the risk materializes, then everything is fine. But if a trader shorts the project, few others follow the lead, and then we go extinct because of the project, the mechanism failed. So in particular traders who don’t have a big audience will the uninterested in shorting bad projects. I don’t know how to solve that elegantly but getting sophisticated altruists to vote on what projects to include in intervention shares can serve to include only fairly robust projects (to the best of our current knowledge).

  3. Various ethical asymmetries, such as how to weight suffering vs. happiness, making happy people vs. making people happy, maybe stuff like how to think about logical counterfactual, etc. I have various opinions and intuitions about these, but I (tentatively) feel like if I built nudges into the marketplace that push in one direction or another, half the potential user base would perceive the market as unfair and acausally somewhere in the universe someone very much like me would build a marketplace where the nudges push in the opposite direction. So I feel like it’s more fair to resolve these through discussions and trade than through marketplace mechanisms.

If you were just referring to the ex ante vs. ex post distinction, then I think it’s fair that people can get lucky by betting on risky projects (risky in the sense of downside risks) because (1) they need to bet on a lot of them to get lucky, so their contribution to each is smaller, and (2) they can get equally lucky by betting against risky projects. I don’t want people to support risky projects, but so long as we can figure out problems 1 and 2 above, the share prices of risky projects should remain quite low.

Or are you thinking of problems along the lines of the St. Petersburg game? I.e. the expected value is unbounded so the share price of a St. Petersburg game charity should go to infinity to the degree the available capital allows? I don’t think that will happen. The upside of various technology companies is virtually unbounded due to transformative AI and yet the share price remains finite and the market capitalization well below the capital that is available in the world. Admittedly, it makes me uncomfortable to rely on “It’s not happening now, so it’s unlikely to happen in the future,” but in this case it seems like a pretty strong reason to me…

Here's a concrete example: Suppose there's 50% chance that next month a certain certificate share will be worth $10, because the project turns out to be beneficial; and there's 50% chance that the share will be worth $0, because the project turns out to be extremely harmful. The price of the share today would be ~$5, even though the EV of the underlying project is negative. The market treats the possibility that "the project turns out to be extremely harmful" as if it was "the project turns out to be neutral".

How do you define “net-negative” if not in terms of expected value? Stochastic dominance? Or do you mean that the ex ante expected value of an intervention can be great even though its value is net-negative ex post?

These seem like very important questions. I guess the concern I raised is an argument in favor of using ex ante expected value. (Though I don't know with respect to what exact point in time. The "IPO" of the project?). And then there can indeed be a situation where shares of a project, that is already known to be a failure that caused harm, are traded at a high price (because the project was really a good idea ex ante).

  1. Investors should be able to short just as easily as they can long, and their profits from correctly predicting downside should be just as unbounded as their profits from correctly predicting upside. This can be approximated with borrowing and lending or with a perpetual future. The first approach requires that someone offers the tokens for lending and that the short trader is ready to pay interest on them. The second only has minor issues that I’m aware of (e.g., see Calstud29’s comment), so that, I think, would be a great feature. Then again the ability to lend tokens that you have would create an incentive to buy and hold.

I don't see how letting traders bet against a certificate by shorting its shares solves the issue that I raised. Re-using my example above: Suppose there's 50% chance that next month a share will be worth $10 (after the project turns out to be beneficial) and 50% chance that the share will be worth $0 (after the project turns out to be extremely harmful). The price of the share today would be ~$5. Why would anyone short these shares if they are currently trade at $5? Doing so will result in losing money in expectation.

Perhaps the mechanism you have in mind here is more like the one suggested by MichaelStJules (see my reply to his comment).

  1. Profit-oriented investors only care about profits that they make in futures in which they can spend them.

That's a great point. This also applies to traders who go long on a share (potentially making them give less weight to the downside risks of the project).

getting sophisticated altruists to vote on what projects to include in intervention shares can serve to include only fairly robust projects (to the best of our current knowledge).

I think something like this can potentially be a great solution. Though there may be a risk that such a market will cause other crypto enthusiasts to create competing markets that don't have this mechanism ("our market is truly decentralized, not like that other one!").

If you were just referring to the ex ante vs. ex post distinction, then I think it’s fair that people can get lucky by betting on risky projects (risky in the sense of downside risks)

My concern here is about net-negative projects, not risky projects in general (risky projects can be net-positive).

Suppose there's 50% chance that next month a share will be worth $10 (after the project turns out to be beneficial) and 50% chance that the share will be worth $0 (after the project turns out to be extremely harmful). The price of the share today would be ~$5. Why would anyone short these shares if they are currently trade at $5? Doing so will result in losing money in expectation.

Ah, you’re right. Not sure why I was so confused about this before.

I might’ve implicitly been thinking about the case where the bad news about the intervention gradually comes to light (in the worlds where it turns out to be bad) and the shorter regularly increases their short to maintain the same leverage while the market drops. Would that work?

I’ve been under the impression that that’s how the HEDGE tokens work, specifically with a rebalancing interval of one day or less (extra rebalancing during the day in case of high volatility), but the result is weird… DMG dropped 77% all at once on Feb. 5, 2021, so I would’ve expected DMGHEDGE to go up 77%? Instead it also dropped 55%. Maybe there was not enough buy-side liquidity to rebalance properly because everyone just wanted to sell?

I wonder if negative impact tokens would work or would fall short for similar reasons. Is there some sort of software for simulating markets? Otherwise I might just try to put together a little custom agent-based model for this.

I wonder, what would happen if we created a USDC/TOKEN market instead of a TOKEN/USDC market?

That's a great point. This also applies to traders who go long on a share (potentially making them give less weight to the downside risks of the project).

Yeah, sadly.

I think something like this can potentially be a great solution. Though there may be a risk that such a market will cause other crypto enthusiasts to create competing markets that don't have this mechanism ("our market is truly decentralized, not like that other one!").

Hmm, it would be completely decentralized. Only the “ICO” would consist of giving the token to specific charities and funds, which would thereby obtain particular voting rights. If a project wanted to avoid that, they could exclude the funds, but they can’t exclude the charities as that would defeat the purpose of the token…

But that’s not much consolation. A lot of the charities I do want to see supported run interventions with potentially vast downside risks, in my opinion. I don’t know if they really have vast or rather limited downside risks, so I would like the market to know more than me about that, not less.

Glad you're thinking about this!

I've never had much luck myself trying to fundraise just by posting to the forum. Just in case you're not already, I'd suggest trying to approach some potential purchasers in the $1-$10m range directly via email.

Thanks! Yeah, and charities whose buy-/sell-in would be important. I’ll start tracking my leads more systematically.

I'll try to directly answer some of the questions raised.

I'm generally interested in this project. If such a system existed, I'd probably issue certificates for research artifacts (papers, blog posts, software, datasets, etc.) and would advocate for the usage of impact certificates more broadly. 

If I were able to reliably buy arbitrary fractions of certificates on an open market, I'd probably do so somewhat often (every several weeks) in order to send signals of value. My personal expenditures would be very small (a few hundreds of dollars per year probably unless something significantly changes), but I'd also try to influence others to get involved similarly. 

As for concerns, I'm very uncertain about my position on the diverging concerns raised and argued by RyanCarey and gwern in this thread. As a creator, I can imagine wanting access to the entirety (or at least the majority) of the value of certificates attached to my work. As an observer of a market, I'd like for it to generally be open for speculation and revaluation, etc. Perhaps I'd be in favor of a system that splits the difference somehow, perhaps via smart contracts that enforce a split of resale royalties (most going to the creator, some going to the prior owner)? 

Relatedly, I'd love to see a workable / understandable / intuitive system for revaluation of a certificate as various parties end up owning various parts of it, bought at differing prices (if such a thing is possible). I can imagine myself wanting to send a signal that a cert should be valued more highly by buying a small fraction of it for higher than the going rate. I may also just be unfamiliar with pricing schemes for fractional ownership and prices like this.

Very interesting!

I'm generally interested in this project. If such a system existed, I'd probably issue certificates for research artifacts (papers, blog posts, software, datasets, etc.) and would advocate for the usage of impact certificates more broadly.

Well, it does. :-) If you haven’t created tokens yet, is it because of the two concerns you listed? I.e. (1) that the creator is not privileged over later buyers of the shares and (2) that the auction mechanism of the exchange always gives you the best price even when you’d like to pay a higher price to signal your support?

The first is something I haven’t thought about enough because I considered that the demand side, especially from profit oriented investors, would be the bottleneck. But especially when I want to attract the attention of well-funded, time-constrained, popular charities, it may be necessary to be able to offer more benefits to the suppliers too. Is there some equivalent in the regular market? Maybe a company offering shares that do pay a dividend and shares that don’t, and issuing the first to founders, early employees, particularly value-aligned partners, etc.?

Otherwise, one mechanism that comes to mind that also has the advantage of being continuous and adjustable is to cap the fully diluted supply at a very high level and put only a small fraction of the tokens up for sale. Most of the value will then be in the hands of the creator but the high fully diluted market cap will make the token relatively unattractive for buyers. The creator can then burn more and more of the tokens in their wallet until an optimum is reached where they still have a lot but the token has also become attractive for buyers. Would that work for you? Or would you prefer for such a thing to be automated, like a bonding curve smart contract except that it burns tokens instead of buying them? Then you wouldn’t have to worry about the particular shape of the curve or about what potential buyers who are still waiting will expect you to do next.

(One could also argue that the creator is necessarily privileged over later buyers by being first to own the tokens, beating all future buyers to it, and having access to most information about the project.)

The diverging concerns you’re referring to are the idea of creating a market as prediction market vs. a normal market for trading tokenized impact? My focus is clearly on the second. I’ve very freshly starting to think about whether there’s some angle of tying in prediction markets, but so far I have no idea how that could be done. Ozzie’s approach here seems very promising to me, but to realize that I’d first need so much buy-in from lots of smart people – sophisticated evaluators like Open Phil and others – that it doesn’t seem to me like it has enough traction at this stage.

I’m also wary of tying in too many (more than one) unusual concepts at the same. “Like the stock market but for impact” already ties in the novel aspect of impact. If that were then also based on Harberger taxes, except prepaid, and stochastically resolved prediction markets, it might get too complicated for users and surely too complicated for me to get all the incentives right…

The second is difficult because being able to buy at something other than the best price runs counter to the basic workings of the (Serum) exchange. Would you be happy to buy for a fixed amount of USDC and offer the surplus of tokens you’ve bought (because of the, in your eyes, discounted price) for sale at the price you deem to be fair?

I imagine other issues are maybe:

  1. That it is still hard to market the tokens? Maybe the EA Forum can pin a post where all top-level comments are token offerings.
  2. The legal questions?
  3. The lack of automated market making? Raydium might be able to help here.
  4. The $100+ for creating a market?
  5. The lack of a mechanism to generate passive income through holding tokens? Raydium can perhaps again help with that.

The way I see Raydium is that they’re still offering their services to a small set of carefully selected projects, so that it’ll probably be hard to convince them to set something up for a token representing one blog post and easier to convince them to set something up for a token representing an established charity. But I’ve been in touch with them only very briefly so far.

Thanks again for your input! :-D

Interesting, thanks for the reply! Let me unpack what I'm thinking of when I say "if such a system existed". Here are some things I'm imagining in such a scenario:

Ideally, there is a market already (not just the potential for one, as that link indicates), or there is a clear plan and a number of EAs that I know the names of who have said that they will participate. I'm willing to be an early adopter, but I'm not in a position where I can vet the fundamentals of the project. For example, I'd like to see people who were involved in the prior attempts to do Certificates of Impact endorsing a plan. Similarly, I'd like to see analysis from a different and identifiable person who is an expert with crypto. I'm just conversant in crypto, and I find most of the writing here to be very somewhat inaccessible due to its length and complexity.

The above is currently my main set of cruxes, but here are a few expanded thoughts on things I'd like to see:

  • What is the precise plan (not just a discussion of tradeoffs and technical possibilities). 
    • Example: A blog post (or several) detailing exactly how the Certificates of Impact system works, how each type of individual can interact with it in all the expected ways, and what its constraints are. After reading such documentation, someone with close to zero crypto knowledge should be able to participate.
    • Nice to have: A community-vetted website or portal hosted on a reliable domain that simplifies the interactions in the market so all unneeded and removed complexity and terminology is hidden.
  • Public scrutiny of the system by identifiable crypto people in the community. At my level of knowledge, the only way that I can feasibly be relatively certain that the system would likely be sane is that I've seen it publicly scrutinized by people who are extremely skilled in this sort of thing.

I realize that what I'm asking for is costly. From my perspective, these requirements seem to be pretty fundamental for us actually kickstarting  a vibrant impact cert market. 

On the flip side, I think there's a lot of potential for such a system, so I'd see this work as quite plausibly very high impact and thus hopefully a mini-cause around which folks can coordinate. Personally, I can try to rally support once a system exists (see above), but I'm not currently in a position to rally community leaders nor get crypto experts to scrutinize the plan.

Awesome, thank you!

  1. Public endorsements from trusted experts in relevant domains, such as crypto, econ, impact certificates, EA-style prioritization. Yes, that’d be great! I’m closest to knowledgeable on the last one, and I’m very concerned about Ofer’s concern above. So I want to come up with a promising solution to that before I move forward at all.
  2. A series of blog posts detailing how different actors can use the system. Yes, very good. I’d also explain the reasoning behind various design decisions in these posts. Something that irks me about a lot of crypto projects is that they have a very specific fine-tuned byzantine architecture that they explain in detail but never mention why they made all these design decisions.
  3. A software that abstracts away all the unnecessary complexity. That is probably something that I’ll need to build gradually as I see signs of adoption. It’d be too easy to waste a year on building a software that no one uses. (I’ve worked for a startup that did that.) So I hope I can do it in a way where the complexity doesn’t itself impede the adoption too much…

On the flip side, I think there's a lot of potential for such a system, so I'd see this work as quite plausibly very high impact and thus hopefully a mini-cause around which folks can coordinate. Personally, I can try to rally support once a system exists (see above), but I'm not currently in a position to rally community leaders nor get crypto experts to scrutinize the plan.

Thank you! 😊

There is a lot of information in your post to digest, but I'll focus on what seems to me your biggest point: "public goods are treated like consumables and not investments"

The way I see it, people can consume private and public goods, just like they can invest in providers of private or public goods.  If you want to invest in providers whose main purpose is to provide public goods, there are already a number of possibilities:

  • Buying government bonds
  • Buying shares of Decentralized Autonomous Organizations / cryptocurrencies

Perhaps you can blur the legal lines and find ways in which non-profits can reward shareholders, but I feel that is no different from:

  • Buying shares of impactful/responsible companies which plausibly make some public goods, namely (quadratic funding) crowdfunding platforms, as their main aim is to let groups/communities coordinate to fund (public) goods.

As for whether the main mechanism you introduce holds up in practice:

> If thoughtful altruists are involved in the market and either own a lot of certificates or want to buy a lot of certificates, other traders will want to predict what these “impact whales” think is impactful.

I recommend chatting with https://www.linkedin.com/in/raphael-mazet/ !

Government bonds, of course! Yes, I should mention those as a precedent. :-)

DAOs with a promising mission: Yeah, but those don’t seem so different from social enterprises or even most companies that provide an important product.

Raphaël Mazet: Interesting, thanks! It’d be great to get input on that question.

To begin with, let me just reiterate a terminological remark I made elsewhere, that might help a bit with conceptual clarity: the certificates don't really denote an increment of impact. They denote being a "patron" for an impactful action. So they're really patronage certificates (Or you can choose a similar name). If the buyer is an EA, it is still an impact purchase, because buyers are simply valuing the certificates based on their impact.

Now let me say something more novel. To decide what the impact certificates should be for, it seems like discreteness is a key desideratum. An organisation is relatively discrete, so it's easier to say whether a charity did/didn't do something, than to evaluate smaller objects (like a project) or larger objects, like intervention areas. Instinctively, I'd think that intervention shares are a non-starter. Because it's so unclear who is allowed to sell them. It would seem to me for an intervention share to be built out of charity shares, similar to how an ETF is built out of stocks, or how a mortgage-backed security is built out of home loans. Out of the other two, I don't have as strong of an opinion. 

If you do "charity shares", then you'd probably want to sell shares corresponding to activities that are restricted to a particular year, or at least those that have already happened in the past. Otherwise, the charity could just sell shares corresponding to large projected future impacts, and then shut down. Once you sell the charity shares, the buyers would need to be able to split those shares up, and sell only that portion of the patronage corresponding to their preferred projects.

If you do "project shares", then there's a bit more overhead for the charity, in selling these separately, but then the buyers can just buy their favourite projects directly. Or if they want to buy patronage for all the charity's activities, they can buy a full set, or bundle them together.

So not sure what's better.

Nomenclature: I’ll need to think about that at some point… I particularly like the analogy with for-profit shares, so having a name with “shares” in it would be useful. Not if it creates legal problems though. I also like “public good” more than “impact” because it sounds more reputable to me and makes clear that we’re talking about positive impact and not random perturbations or negative impact. “Public good patronage share” is getting a bit wordy though… It seems too early to think about this in earnest though.

Discreteness: That is indeed desirable, and charity shares are probably top on that metric, but I don’t see the differences between the models as particularly great. Intervention shares are almost as discrete as charity shares in that they are simply a few charities pooled together with the option to add more, sort of like the Serum Ecosystem Token. At every point, it’ll be complete transparent who has been voted into the pool and how much tokens their wallets still hold. Projects are a bit more vague in that no one goes out of their way to write binding by-laws for a research project and lock themselves in to some particular mode of operation through expensive marketing – but then again charities can gradually change their complete staff. A project will rather tend to be discontinued when a key person (maybe the only person) stops working on it. But in the end, I think charity shares still win by a small margin.

But charity shares are probably a nonstarter for different reasons: (1) So long as the benefits of fundraising through impact stock issues is avant-garde, speculative, and legally risky, charities will be loath to invest any time into it. That might be avoided by me issuing the token for them and donating them all to them, but I don’t know if that really makes a difference legally. Besides, they need to be auctioned off in some fashion, which will again cost staff time. (2) The legal situation will probably be determined by the location where the charity is registered, so it would be infeasible for a third party to do the legal research on behalf of all the charities. Even if most of them are in the US, the laws there vary a lot between states. (3) Liquidity will be a huge bottleneck so long as the markets are not yet well established. Even the liquidity on the SECO/USD market I mentioned above is rather meh, and that’s listed on big exchanges. This will be more manageable when there are fewer markets. That may also make it easier to get someone like Raydium on board with the project. (4) To be able to short bad charities, there needs to be lending. The smart contract that could manage the remaining intervention shares could, for a long time, see to it that these can be borrowed at high interest rates (e.g., > 500% APY, so that lending still incentivizes hodling). (5) Intervention shares have a nice build-in decentralized mechanism for keeping bad actors out. For charity and project shares, only centralized mechanisms come to mind.

So for better or worse, I don’t think anything other than intervention shares will be feasible at all at first. But if the system catches on, charities and projects will want to jump on the fundraising train, and then the incentives would be right, so that they’d be happy to put some work into issuing and auctioning off their shares. Note that the idea behind project shares is to remove the legal risks from the charity, so the charity would never touch them or do anything with them; that would all be managed by the employee who runs the project and is in the right jurisdiction to issue the shares safely.

Past activities: Only selling impact in past activities would seem odd to me. Of course there are a lot of established companies that have their quarterly earnings calls where they report on various fundamental and how they changed over the past quarter, and then shareholders react to that. But most EA charities are more like startups. If we want to grow the community, then hopefully most EA charities are entirely in the future. That may be analogous to how one can base the valuation of a profitable company on EBITDA but has to come up with other methods for early-stage startups. AMF might have quarterly net calls, but the valuation of most charity startups will probably also be based on such criteria as the soundness of the strategic plan or path to impact, the team, the network, the marketing, etc. There might of course be Safemoon-type charities in the end, but I don’t see how that could be avoided. All we’ll be able to about it is education, centralized vetting, and using the classic legal system against them. Besides, we at least already sort of know how to do that, to some extend, since we’re currently allocating our illiquid donations according to similar criteria.

Agree that discussing terminology is not yet useful in and of itself. Though I'm intending it for the purpose of idea clarification.

Re charity Vs intervention shares, my thinking was just that it would be more transparent for intervention shares to be constituted of charity shares, and for such shares to be issued by charities. Based on reading your comment, I'm not sure whether you agree?

As for your arguments: I find myself not so convinced by (1-2). I think the process of issuing charity shares could be automated for the charities. If desired, it seems not out of the question that these entities could even run as for-profits - given that you are proposing a revolution of the NGO sector, it seems weird to restrict yourself to the most common current legal setup (although I agree that tax deductability is nice to have).

I can see that (3) pushes weakly toward impact certs, but not strongly because ideally you also want to have specific markets, and the benefits of liquidity and specificity trade off against one another (in terms of the information that readers can gain). And even if resale markets are fairly dormant, I don't think it's a disaster - it should still be at least as good as the status quo (donations), and in many ways better (valuation is done retrospectively).

Re (4), why can't charity shares be bought/sold? Re (5), what is the built-in mechanism?

Re past/future shares, on further thought, even if you only allow patronage certs to be sold for past events on the "bottom layer", there are ways to route around this: you can sell shares in the company itself, or you can sell the rights to any future patronage shares. I'm certain this is a good thing, because it allows people to invest in orgs that will have large future impact, similar to investing in an org that you think will win an x-prize. The real question is just whether you should allow this "natively", i.e. whether you should.be able to sell patronage of future activities. If you think of normal stocks, they do confer an ownership of the company into the indefinite future. Stocks can also have the problem where people make a company and make a bunch of promises about what it will do, sell it, then reneg on those promises - they call it securities fraud, and have a lot of defenses built up against it. If you want to piggyback on that, maybe you would want to only allow sale of past activities "natively", and then for sale of future impact to be done only by sale of regular stocks in the company itself. That's my initial instinct, although there may be a lot of other considerations.

I think the process of issuing charity shares could be automated for the charities.

Yes, that’d be awesome!

  1. Creating a wallet is so easy, there is virtually no need to automate anything further.
  2. Creating a bonding curve auction on Serum is something that would be valuable to automate.
  3. Raydium integration, lending, etc. are slightly less essential, but that would also make sense to automate at some point.
  4. But researching the legal environment in the state where the charity is registered and coming up with creative ways around local regulations or going through complicated registration procedures such as the ~ 1.5 year long one with the SEC are not things that I can automate (I think).

So it varies – a lot of things can be automated but what remains is probably still prohibitively complicated. There are also more things that charities can do with their shares to make them more attractive, such as governance mechanisms. Finally, I don’t want to put all the work into automating the system so long as its demand and product-market fit is unproven.

If desired, it seems not out of the question that these entities could even run as for-profits - given that you are proposing a revolution of the NGO sector, it seems weird to restrict yourself to the most common current legal setup.

I’m taking two different perspectives in the comment based on the following steps: (1) What is realistic to realize now to get the idea off the ground, and (2) what is realistic to expect to happen in 5–10 years assuming that step 1 has succeeded. Now that we’re at step 1, I think it’s unrealistic to think that almost any charity will be ready to run these risks and invest any time given the unknown value of the fundraising system for them. But once we’re at step 2, the value of the system will be proven (or else we won’t reach step 2), in which case charities will hope to raise tens of millions or more through the system, and it will be worth going to great lengths for them, e.g., founding a for-profit, hiring lawyers, and going through SEC registration processes.

US for profits of course also have to register their securities. They can do a lot more with them, so doing this as a for-profit is probably necessary, but it’s still very costly. Founding a for-profit branch in another country may be an option, but I don’t know enough about that to tell.

although I agree that tax deductability is nice to have

I don’t see a way to get that unfortunately, but then again the money with the least valuable counterfactuals comes from for-profit investors who don’t expect deductibility anyway. The legal risks I’m referring to are not simply that it might not be possible to get tax deductibility. It’s rather that in the worst case the responsible people at the charities may need to pay 8–9 digit settlements to the SEC or go to prison for up to five years for issuing unregistered securities. Especially the second would be a tremendous risk to the whole AI safety ecosystem. Even if the risk of that happening is small because they’re likely to be able to reach a settlement, it may still be too great of a risk for any AI safety charity to touch tokenized impact at all.

I can see that (3) pushes weakly toward impact certs, but not strongly because ideally you also want to have specific markets, and the benefits of liquidity and specificity trade off against one another (in terms of the information that readers can gain). And even if resale markets are fairly dormant, I don't think it's a disaster - it should still be at least as good as the status quo (donations), and in many ways better (valuation is done retrospectively).

Yeah, makes sense.

Re charity Vs intervention shares, my thinking was just that it would be more transparent for intervention shares to be constituted of charity shares, and for such shares to be issued by charities. Based on reading your comment, I'm not sure whether you agree?

Ah, you actually intended it exactly like SECO, okay. :-)

My thinking has gone through the following steps: (1) I want to create charity shares. (2) Oops, charity shares are prohibitively difficult to do because of legal risks, effort, and hence very low chance of getting the buy-in from all the US- and UK-based EA charities. (3) So I need to come up with something that is almost as good but is more achievable: Project shares, because individuals are more likely to be outside the US or UK, and intervention shares because I can do them from Switzerland with minimal buy-in (just an okay) from charities.

So once we are in a position where it becomes realistic to expect charities to issue their own shares, we don’t need intervention shares anymore. They may still have their various benefits, like governance and maybe higher liquidity, so they may continue to be developed, but at that point they can become an afterthought. And maybe there’s a way to convert them into a pool of charity shares too.

Re (4), why can't charity shares be bought/sold?

Ah, my point here was more that an evil charity that is afraid that it’ll get shorted can decide not to offer the (say) 99% of its shares that it still holds for borrowing. The only shares shorts can borrow will then be those that some others have, directly or indirectly, bought from the charity, so relatively few. Conversely, the smart contract that manages the intervention shares could just be developed such that it automatically puts up all the shares it still holds onto a lending platform. It’ll probably take a few years before that proportion is down to 1% and by that point the shares will be distributed across a number of charities of which hopefully very few are evil. ;-)

Re (5), what is the built-in mechanism?

I’m referring to this one:

There could be, e.g., quarterly events where a new organization can receive a fraction (which decreases every time) of the remaining shares. The shareholders would vote on who the new awardee will be, and the issuing organization or a smart contract will donate the shares accordingly.

I’m currently very concerned about prices not reflecting downside risks, and this mechanism is the only one that may be able to keep risky charities out, so it seems very important to me but unfortunately also rather unsatisfactory.

researching the legal environment in the state where the charity is registered and coming up with creative ways around local regulations or going through complicated registration procedures such as the ~ 1.5 year long one with the SEC are not things that I can automate... I’m taking two different perspectives in the comment based on the following steps: (1) What is realistic to realize now to get the idea off the ground, and (2) what is realistic to expect to happen in 5–10 years assuming that step 1 has succeeded... I don’t see a way to get [tax deductability] unfortunately... The legal risks I’m referring to are not simply that it might not be possible to get tax deductibility. It’s rather that in the worst case the responsible people at the charities may need to pay 8–9 digit settlements to the SEC or go to prison for up to five years for issuing unregistered securities.

VCs often manage to buy stakes in companies privately. Wouldn't it be natural to sidestep that issue by copying what VCs do (and staying off the blockchain)? i.e. step (1) is privately traded patronage certificates, then step (2) is public ones? If so, then one could imagine a scenario where all you need for now is to do some research, and write up a pro forma contract?

Ah, my point here was more that an evil charity that is afraid that it’ll get shorted can decide not to offer the (say) 99% of its shares that it still holds for borrowing...

I can envisage a lot of ways to ensure some lending, so this seems like a small advantage.

I’m currently very concerned about prices not reflecting downside risks, and this mechanism is the only one that may be able to keep risky charities out...

Yes, having the ability to short companies is quite a weak method for punishing companies, because they can just stop selling patronage certs if they go negative. It would be better if we could get charities to pay for their negative impact somehow. An ``absolving'' certificate, of  sorts. Maybe the people who would want to sell these ``absolving'' certificates are similar to the ones who look to buy ``patronage''...

My thinking has gone through the following steps: (1) I want to create charity shares. (2) Oops, charity shares are prohibitively difficult to do because of legal risks, effort, and hence very low chance of getting the buy-in from all the US- and UK-based EA charities. (3) So I need to come up with something that is almost as good but is more achievable: Project shares... and intervention shares...

Ahhhh, OK! I must say though, it rewards and punishes orgs for the performance of other orgs in their area. You portray this as a positive, but it seems like a big negative to me. It incentivises people to start new incompetent orgs in an intervention area, (or to keep incompetent orgs running) just because there are existing competent orgs. Conversely, it punishes competent orgs for the presence of harmful orgs implementing their same intervention. It's quite messy to require an external panel to divide up the tokens between orgs. Frankly, given the fact that it's a bit inelegant, I would bet that other problems will arise.

I can't promise I'll have much more to say in this thread, but in case I don't, let me say that I've found this an illuminating discussion. Thanks!

VCs often manage to buy stakes in companies privately. Wouldn't it be natural to sidestep that issue by copying what VCs do (and staying off the blockchain)? i.e. step (1) is privately traded patronage certificates, then step (2) is public ones? If so, then one could imagine a scenario where all you need for now is to do some research, and write up a pro forma contract?

Heh, yes. That’s an option. But I don’t suppose having a contract template has been the bottleneck why this hasn’t happened over the past years? I made a Google Doc for this for one impact purchase, and it worked just fine. They even have a version history that’s a bit like a blockchain. Writing contracts is rather far from my absolute advantages, so maybe it’s also far from my relative advantages…

I can envisage a lot of ways to ensure some lending, so this seems like a small advantage.

Yeah.

It would be better if we could get charities to pay for their negative impact somehow.

No mechanism comes to mind, but that general problem is one I want to think about more.

It's quite messy to require an external panel to divide up the tokens between orgs.

That’s just for the initial allocation. So long as no one (other than me) has the tokens, no one (other than me) can vote, which is boring. But once I’ve allocated some 50% of them to MIRI, FHI, CLR, AISS, et al., the actual voting can start. I haven’t researched how that is usually done on Solana, but surely there’s some elegant mechanism for it.

It incentivises people to start new incompetent orgs in an intervention area, (or to keep incompetent orgs running) just because there are existing competent orgs. Conversely, it punishes competent orgs for the presence of harmful orgs implementing their same intervention.

If the existing shareholders – MIRI et al. – initially vote to get Evil Org accepted into their ranks, but then it turns out that Evil Org is evil and they regret the decision, then yes, that’s a problem. But if they set a high bar and only vote in orgs that have proven over many years to do high-quality and conscientious work, the risk from that failure mode can be minimized. So for this to become a problem, a new org would first have to impress MIRI et al. a lot, but then completely fall short of their expectations after all.

And if the problem with the disappointing Evil Org gets too bad, someone (me) will deploy an alternative intervention token, and the market will decide which one people like more. There might even be a market of one token against the other.

I can't promise I'll have much more to say in this thread, but in case I don't, let me say that I've found this an illuminating discussion. Thanks!

Awesome! Thank you for your input! :-D