Epistemic Status: Early concept

The Gist of it: To support sane risktaking, an institution should generally grant career capital in accordance with a quantitative record of the failures and successes of a decisionmaker's past funding decisions. An impact market seems to be the best way of doing that. In addition to impact markets, I propose having impact investors work in a special currency. I don't think I straightforwardly justify this complication, but it seems like it might solve a couple of problems and be generally better for a couple of reasons.

Problems Addressed

Markets are good at rewarding successful risktaking, but there are many critical public goods that they reliably fail to fund in proportion to their importance. Examples include information sources, existential security, all non-excludable goods and unmeterable infrastructures, these are the holes left by the coordination problem of free-riding and they are generally left to governments to fill. Unfortunately, any large, long-lived institution, including most government funding bodies, tends to demand burdensome degree of legibility of their projects, they only allow simple, respectable-sounding projects, and they either forbid or disproportionately punish risktaking.

In most institutions, the downside for visibly taking a risk and failing is significant, while the upside of taking a risk and succeeding will be nowhere near proportionate to the downside. Sometimes the positive contributions you make wont protect you at all from being dismissed the moment you lapse, so institutions don't take risks, so nothing novel gets to happen, and most systemic change would be novel, there's a sense in which any sort of substantial improvement must be novel.

Being duly supportive of eccentrics is just naturally difficult for a large organization. Doing it without appropriate systems in place to protect disruptive decisionmakers from perverse political structures and collective loss aversion biases is like trying to hand-draw an even circle without a compass. It's easy to imagine that happening, but it doesn't really happen.

I have not seen a suitable compass, before finding this one.

Solution: Venture Granters

Venture Granters (VGs) act as participants in a public goods market, investing their marks to take stakes in various projects and companies. The investment of marks in a project corresponds fairly directly to allocating tax spending to that project.

While a venture capitalist's career capital comes to them in proportion to their holdings' captured profits, a Venture Granter's career capital comes in proportion to their holdings' expected captured and uncaptured value. What constitutes uncaptured value is defined and reported by the the principled, patient, methodical and often precisely operationalized judgements and metrics of the weighers.

As with private stock markets, money travels in correct proportion to the expected outcomes, there is no bias against risk, a bet can be leveraged, and stakes can be sold for marks to other VGs, providing exceptional VGs with early exits from their investments so that they can often go on to make other investments long before of the eventual arrival of the evaluations of the weighers.

Creative experimentation propels our culture forward. That our stories of innovation tend to glorify the breakthroughs and edit out all the experimental mistakes doesn't mean that mistakes play a trivial role. As any artist or scientist knows, without some protected, even sacred space for mistakes, innovation would cease.

~ technology critic, Evgeny Morozov

Leaders must foster an environment that appropriately tolerates mistakes, errors, and challenges to existing ideas.

~ US Marine Corps Doctrinal Publication 7 (2020)

Exploring the Details, Doubts and Warts

  • There is a low barrier to entry to becoming a VG, they're brought in from a diverse set of backgrounds, many are allowed to try, though not all will succeed.
    • It's also possible to be brought in by other VGs, as a project. Their investment in you will come back to them in proportion to your success, as naturally as it would with any other VG project.
      • A Project VG effectively has their returns garnished (in the sense of wage-garnishing), for a time. This is no less fair than the investment firm who takes most of their analysts' rewards and leaves them only a meager commission. The garnishing attenuates over time, though! Again due a general principle of weighing that applies to every VG project: The principle of fair attribution: Ultimately, the investor most responsible for their success will be themselves. Correspondingly, in the long run, their investors will stop receiving a cut of their returns.
  • Any citizen can make a project proposal.
    • No filing fee, any format, no bureaucratic requirements (those may come later if a VG likes their proposal).
    • Sorting through the proposal list is left to the VGs, or a VG-funded project curation agency.
      • It is exceptional that we can leave the design of that system to the VG market. Information markets usually don't work, because information is a type of good that a potential buyer can't tell the value of until after they have it, and at that point there's no way to get them to pay for it - so it's unmeterable. Today's governments try to partially patch that with intellectual property laws, which allow you to sort of receive a piece of information and evaluate its worth without being allowed to use it until you pay the fee, and well, that's not working out very well.

        The VG market is different, because it meters its success via the weighers rather than by captured profit, so you don't need a way of getting people to pay for information after they've already received it.

        For once, information curation and coordination is a problem that we can trust a group of competing actors to solve for themselves, and we certainly should let them solve it for themselves, because they will need the practice, before moving on to the task of building information tools for the rest of us.
  • If a VG believes in a project, they can lay out a weighing, an operationalization of the bet, a description of the proxy metrics that would most directly measure the project's generation of uncaptured value.
    • An overly simplistic example of a metric for a highway construction project would be "The new highway returns 1 util (standard unit of human flourishing) for ever mile driven times the speed being driven limited by the safe speed limit (measuring congestion prevention)"
    • A metric would have to provide a lower bound. Not all value can or will be captured by simple formal metrics. Room should be left for taste.
    • A good weighing is one that, if it went high, would indicate clearly, legibly, that value is being created, while also, it's not so obvious that it will go high, that a traditional, risk-averse organization would take the leap of faith and fund it. The VG then takes on the risk of funding it.
      • It's important that weighers never fall into a norm of rejecting weighings just because they consider them unlikely to go high. It isn't the weighers' responsibility to assess that. They must base their assessments solely on whether, if it did go high, no matter how surprising that would be, value would be produced.
        • For instance, if I propose the weighing "about 5 billion utils would be generated if Elizabeth Holmes' proposed next venture creates a universal flu vaccine in less than 5 years", the weigher has no business stopping me from giving my marks to Elizabeth Holmes! It is indeed true that if that vaccine were produced, regardless of who produced it, lots of people would benefit!
    • The weighers help the VG to develop the metric into something robustly true.
      • I'm afraid I'm going to have to give up on the goal of having minimal interaction between the VGs and the weighers, we'll have to find better ways of minimizing insider trading than that. Recording all conversations between weighers and VGs would be something to consider.
    • Once the metric is good enough, the weighers make a binding commitment to honor it, which lets the VG estimate the returns they expect to get, and when.
  • When a project is approved, its stakes are auctioned, usually with a reserve price (there will usually be a minimum amount of funding a project will need to be viable).
  • As the project's measurement events arrive, marks are paid out by the weighers in proportion to the measured social impact (accorded to the pre-agreed weighings, and additional judgements, less strict).
    • Generally, more are returned than were invested, because that is the nature of uncaptured value. It isn't conserved. The pie grows over time, the tide raises all boats. This means marks are inflationary. That seems good to me. A VG must keep running to stay in the race, complacency is punished.
  • A VG's salary (in real money) is a relatively modest livable wage plus a decelerating function of their mark earnings (a logarithm, perhaps).
    • Exceptional VGs don't earn as much as they would being private VCs. I don't think this will be a problem. In a VG job, the core human motivations are provided for in competitive quantities:
      • Security within their social group (job security) comes in proportion to a VG's marks. If you've got lots of marks then nobody can fire you, no different from the VC world.
      • VGs receive a lot of agency to have a beneficial impact on the world. VGs can make life lastingly better for the next generation. For a lot of people, that's everything. That's all they need.
        • Whether the public or private markets will give a person more of that seems to depend on the person. Some people only see captured value, for them the uncaptured value, the underfunded commons, the coordination problems, seem small or unimportant or possibly not arguably real. For them, being a VG would be unappealing. Myself, on the other hand, I've always been afflicted with a curse of sensitivity to uncaptured value, I see it everywhere and it is vast, and it's all I can think about, being a VG with a modest salary would be strongly preferable for me.
  • If relatively little value is created by a public project, its shareholders are unlikely to get back the marks they invested in it. Consistently mediocre investors will lose marks.
    • Once a VG's marks drop below total mark issuance/10^8 or something, they lose their wage. It's not obvious that it's necessary to take their marks away or prevent them from accessing the market, though? If they think they can climb back, why not let them try.
  • Open questions
    • Stopping VGs from selling their marks for real money?
      • Allowing this would exasperate corruption, though it's not clear how much. The activities through which marks are earned are all supposed to be beneficial, same as private trade. Good is done in the earning, not just the reinvestment.
      • If they can trade stakes for marks whenever they want:
        • (which would allow exceptionally bright VGs to rise faster. What is a "sure investment" (20% AROI) to others, is rocks in their boots, because they can see a better investment (50% AROI) that those others can't, so they should sell their shares to others and use the marks for the next thing)
        • , then selling marks would be very easy. A mark buyer "sells" useless stocks to the mark seller, while passing them real money in secret. If you can't see the real money moving, this could be somewhat difficult to distinguish from the mark seller having simply lost at the game.
        • resolutions:
          • Perhaps this could be resolved by requiring VGs to expose their personal finances to much more auditing than an average citizen would?
          • Marks aren't worth a whole lot of money to corrupt people. They only bring logarithmic salary increases and they're mostly only useful for generating public goods.
      • In other cases, cashing out would look like, investing their marks in a project of theirs or a crony, then that project spends it paying fat salaries rather than trying to do the thing? I dunno, that's just regular embezzlement, I guess, not that hard to spot. The ultimate judge of it is the weighers, who notice that nothing is really getting produced and downgrade the investors.
      • More subtle cases of cashing out, involving collusion with weighers, would often tend to look like a bunch of galaxy-brained capitulation about what constitutes the generation of value, you would see commitments to abominations like "utils are generated at a high rate in proportion to the number of abstract paintings that my daughter produces in one year". A self-serving narcissist could argue indefinitely that these paintings are disproportionately valuable, even if most people measurably don't derive any enjoyment from them, they could argue that the paintings have having some kind of subtle but crucial positive cultural impact. A weigher needs to be decisive when impacts are small and ambiguous, or else there is no way to limit the descent into indulgence and corruption. Work that seems like it would definitely help a much larger number of people than just the recipient of the grant (automation, medical research, existential risk reduction, etc) has to take priority.
    • Structures and cultures of succession are a big deal.
      • Many of the public projects this system needs to support will take longer than one human life to bear out, most novel projects take quite a while to fully reveal their fruits, the longer you can make investors' time horizon, the bigger the projects can be. Successions outlive any single member in them. They represent schools of thought, tested by experience and, if they prove successful, growing a branching legacy.
      • Rule? A VG, 30 years before their retirement, must take 2 to 4 mentees, and their mentees must inherit their marks and shares?
        • Creating a strong relationship with a set of people who know their techniques and business well enough to be able tell if they try to cash out. Your mentees are your most shrewd auditors.
      • (I wonder if a similar principle could be applied in private EA cultures... if a person doesn't have mentees, listed in their will, their interest in the long term can be questioned.)
      • Mentees should probably also generally have multiple mentors, it is not healthy to have just one.

The Weighers

The weighers soberly weigh the manifestation of value, once it's as obvious as it is ever going to be, in a somewhat predictable way. Their evaluations determine the payouts, so, the job of the VG is, approximately, predicting what the weighers are going to say (comparable to the concept that has been floating around the community of having lower courts just be prediction markets about the decisions of higher courts, with escalations random or in response to appeals.)

The weighing institution is a big, complicated, fallible part of the design that warrants as much and possibly more discussion as the VGs.

  • When should a public good be taken for granted, when should the mark rewards stop coming?
    • For instance, the public good of the existence of earth is of colossal value to every human alive, no one is paying for it, and I don't think anyone should be paying for it. It takes no work to sustain it (beyond meteor mitigation and AGI alignment (unaligned AGI may explosively shear apart the earth to get its materials out of the gravity well. Aligned AGI may also do this, but only if we agree to it.), but those only cost so much), we should be focusing our capital on other things. Imagine that a VG had funded the creation of the earth. Once the earth exists, it requires no upkeep to keep existing, so in a sense nobody should be paying for it any more, but the value being generated is constant, the earth is facilitating everything else that happens, so should the VG not keep receiving the majority of the mark issuance every year?...
    • Maybe??? If their wealth is divided among their successors is it possible you'd still end up with a functioning public funding ecosystem??
      • Yeah so that would mean that if you make a source of value that's more consistently instrumental in the production of value from then on, your succession displaces everyone who wasn't part of it or didn't buy into it. That seems about right?? Everyone who disbelieved it, or didn't make something comparably important, was infinitely worse at their job than you.
        • Eventually we should consider some sort of Rebalancer of Fortune, who identifies cases of gross luck and redistributes its spoils accordingly. For preliminary tests, though, I suspect that the VG market will already do a lot of that itself, luck will be anticipated in advance, and VGs will diversify their stakes accordingly, serving the same function. Additionally, the amount attributable to luck may turn out to be quite small, so never mind.
      • No because that would mean, they'd basically all be on a mark ubi so large that it would totally dilute any additional marks they could earn by actually doing their job. This wouldn't be such an issue if they were very sweaty expected mark optimizers, but humans aren't.
      • Maybe there should be a Red Queen who tweaks redistribution rates so that differences between VG performance always keep evolution rate or sustained diversity optimal, comfortably brutal, so that the complacent sink quickly, the exceptional rise quickly, and the decent are able to keep doing their job without being made nearsighted from stress.
  • How are factors contributing to the generation of value separated and rewarded in fair proportion to their contribution?
    • I don't think this issue is going to be worse than it is in the private sector, in which the surplus is split according to arbitrary things like the number of middlemen, the participants' moods in negotiation, inflation (wages and prices often stick, even when every party knows inflation is occurring), esteem, class organizational capacity, and weird norms against disclosing our pay to our coworkers.
    • While, the weighers are uniquely interested in making this fair and regular. They'll try earnestly to come up with a procedure for making these decisions from an omnibenevolent perspective, and also in a way that's predictable for VGs!
  • How do we keep the weighers as a separate, impartial institution from the VGs, despite the fact that VGs will often have some crucially insightful things to say about how the weighers should work and what information they should be paying close attention to?
    • If the barrier between VGs and the weighers weakens, VGs will try to push the weighers more and more towards enabling embezzlement? (marking up the useless projects of people who give them money under the table then pretending after the fact that those projects are helping people in some made up abstract way)
      • A healthy VG culture would have some barriers in the way of this: A community where people are accountable to their reputations for healthy VG virtues, everyone is looking for stock tips from others, and they're only interested in listening to VGs who have a genuinely good sense for value generation. It would be relatively easy to get VGs to police each other for rorts, because if you're a good VG, you don't get to exit unless other VGs are generally insightful and acting in good faith. Given that marks can be earned from policing against corruption, it might be an exceptionally effective self-policing culture.
    • A mildly competitive vibe will always separate weighers from VGs, their perspectives and character is very different, one group are mostly freelancers, the others institutional. Weighers are not dependent on VGs' approval in any way I can see, and almost every decision they make will piss a VG off who was invested in the opposite outcome. Weighers will get used to tuning out complaints and they can expect to be rewarded for evidencing attempts at bribery.
  • There's a lot to be said (or asked) about whose notion of goodness weighers are supposed to measure. Which culture's noble goals are they promoting? The broadest most encompassing kind of humanism? Or should they have an internal representation from all of the moral tribes of their society? Or should there be completely separate weighing institutions for each moral tribe, with their own budget sources, who cooperate with each other on shared interests in the same way nation-states often attempt to?

Concluding Advice

  • Civicists:
    • Assist in completing the design.
    • Bring in insights from the tokenomics world to narrow in on the best implementation for the conversion of marks into funding dollars.
    • Introduce investors you know to the category of public goods, and watch for a sparkle in their eye as you propose to them the idea of a market where total generated value is rewarded rather than just captured value.
  • Governments:
    • Just try this in fairly small scale and measure its cost-efficacy relative to more deeply hierarchical funding decisionmaking processes, and give us a report on where things go wrong so that we can patch it and try again a few more times. If the patches work, once we're confident of the model, scale up until it hits diminishing returns.


New Comment
8 comments, sorted by Click to highlight new comments since: Today at 5:26 AM

I would look into social impact bonds, impact certificates, and retroactive public goods funding. I think these are three different attempts to get at the same insight you've had here. There are incipient efforts to get some of them off the ground and I agree that would be great.

Interesting, I'll look for some of those. I guess prizes/bounties would be impact bonds, yeah? (Some recent examples: Musk's 100M USD xprize for carbon capture, or MIRI's 1.2M USD prize for generating a dataset associating sections of prose with the intentions of the author.)

I notice that there are sort of two ways of scaling down a public goods market for small-scale tests. We could call impact bonds horizontal down-scaling, narrowing it down to particular sectors or problems, while the VG system is a way of achieving vertical down-scaling, it's a way of letting the market decide what to do for itself while looking over every problem in the world, despite having funding sources that are much smaller than the world's needs, but without the funding being diluted away to a barely audible background noise, which is what I'd expect to happen with a lot of retroactive public goods funding?

And I think letting the public goods market decide for itself which problems to go after may actually be crucial! Most governments are not prioritizing the actual root causes (press, digital infrastructure and x-risk), unfortunately, good cause prioritization doesn't seem to be democratically legible, it is part of the illegible component of the problem that has to be left to VGs, with their special illegibility-compatible accountability mechanism.

On the other hand, if we're scaling down in order to run a demonstration, maybe fixating our systems onto very specific pre-determined goals would be preferable, the reality we live in is a crypt world where the past owns all of the foundations upon which the future can be built, the system has to be made convincing to these risk-averse organizations that do not like surprises. They do not want to find out that we should be pouring all of our money into some weird abstract indirect root cause, instead of the causes they were already invested in.
So maybe we should just keep doing horizontal stuff.

A paraphrasing of Venture Granters concept written for some friends working on an Impact Certificate market system:

The Venture Granters system would be a closed impact certificate market, fractional, with a special currency that seems like it would probably eliminate most losses to speculation.
That's important, as losses to speculation are otherwise potentially large enough that it could undermine the whole concept of an impact cert market and irrecoverably crash it: If there are too many losses, final buyers wont participate, the market wont be made

Venture Granters don't deal in real money[1], instead dealing in an inflationary currency that translates directly into real money for the worker (taken from the yearly national budget/funding stream) on the first sale of the impact cert.
This also solves the issue that there may be no practical way of taking enough taxes to get venture granters' relative career capitals to quickly line up with the value that they've produced for the world, as the uncaptured value produced by public goods, which we're trying to measure, can easily grow larger than the GDP. With an internal currency, any quantity can just be minted.

(I know that inflationary currencies are scary, a marketplace is always tempted to run out on them and crash them. I get the impression that this is totally answered by the fact that VGs are an altruistic captive market, they neither want to, nor, can abandon the currency. There's an extent to which they can: They can try to hold their wealth in impact certs rather than marks, but isn't that just investing? Isn't that what we want?)

[1]: An aside, can anyone tell me what "real money" actually means? I'm not sure I know.

If I'm really going to bite down on the "people who are good at this will be kind people", then is it really possible for "losses to speculation" to be a bad thing, given that the people catching the losses are so good?

While, the weighers are uniquely interested in making this fair and regular. They'll try earnestly to come up with a procedure for making these decisions from an omnibenevolent perspective, and also in a way that's predictable for VGs!

Why would they be uniquely interested in that? How do you incentivize them for that?

Internal legibility, rules, and democratic pressure. These incentives are good enough for some kinds of work.

They are not good enough for project prioritization and entrepreneurship, which can't be (reliably) internally legible, which follow no regular patterns that fixed rules can accommodate, and judging them takes lots of time and specialist knowledge. So we end up needing this special qualitative post-hoc accountability mechanism that is more suited to them.

But not everything requires that sort of mechanism.

What I think I might have done here is separated the work of public goods funding into two departments that run under different incentive systems, each suited to the job they're assigned.

We should consider the much simpler design of getting the regular VC market to care about public goods by having weighers throw money at whoever produces them.

I think it would be less cost effective, and would impel very little evolutionary pressure, when the budget is much smaller than the potential public goods market (which it will be in small scale tests), and it'd be a lot more difficult to sell politically, but maybe there could be a cryptoeconomy where none of those issues matter at all.

It does seem a lot more elegant.