A couple of my recent posts (1,2) have mentioned (at least in footnotes) that judges in prediction markets can instead be modeled as market participants with large wealth relative to other market participants, who use their wealth to effectively price-fix.

One might reply: "That doesn't make sense. No one would do that. It won't make any money." I've argued against this view twice recently (1,2).

An analogue is impact certificates. Here, let me set up the analogy:

I've picked up the idea of impact certificates organically from being around the Bay Area, but over the past year I've run into several Bay Area types who didn't know what they are, so I'll explain. It will be easier if I first explain Decision Auctions.

Decision Auctions

Caspar Oesterheld described decision auctions in Decision Scoring Rules and A Theory of Bounded Inductive Rationality. A simple decision auction works as follows:

• You're trying to decide between options A and B.
• You auction off one share of A and one share of B.
• You do whichever has the higher price (let's say A), and refund the other auction-winner (B).
• Later, you decide how well the decision went and award money to the person who still holds a share (A), perhaps based on a pre-announced rule such as percentage of revenue.

This incentivizes accuracy with respect to the best option and its price, since under-bidding the value of an option can get you out-bid, and over-bidding loses you money.

This is better than prediction markets for decision-making, because prediction markets don't incentivize honest bets anymore if they're tied to decisions. Options that very probably will not be chosen are unprofitable to bet about even if they're actually high-quality options. Prediction markets used for decision-making can therefore get locked into suboptimal policies.

Decision auctions can be won by the biggest optimist (underestimate downside risk), but they give that optimist a chance to lose their money or prove themselves right, which gives them good learning-theoretic properties. Prediction markets might be too cautious and not learn; decision auctions might be too experimental, but they'll learn.

Notice how the above reasoning relies on money-maximalist ideas, however. The bidders are imagined to be money-maximizers, which smuggles in an assumption that they're disinterested in the actual consequences of the decision (they only care about the payout).

This is like a prediction market with judges. At some specific point the judge declares the contest to be over, and awards prizes. A prediction market, at least, can be money-neutral (the winners get money from the losers). Decision auctions, however, require a payout.

Impact certificates decentralize everything. There's no specific judge, just the next person who buys the share.

Impact Certificates / Impact Markets

Suppose I do some work that benefits the world, such as help distribute mosquito nets. I can then issue an impact certificate, possession of which implies a specific sort of bragging rights about the beneficial work. A philanthropic investor comes along and assesses the impact certificate, decides that it is worth a specific amount, and buys it from me for that price.

This is similar to a simple money award for good deeds. The advantage is supposed to be that the philanthropic-investor can now sell the impact certificate onwards to other philanthropic-investors. This means a philanthropic-investor doesn't need to commit fully to the role of philanthropist; they have a chance to sell impact certificates later, perhaps for an improved price if they chose wisely, which makes the philanthropy less of a big up-front commitment, thereby attracting more money to philanthropy.

My favorite version of this also gives some percentage of the price difference back to the original impact-certificate issuer every time the certificate changes hands, so impact-certificate issuers don't have to worry as much about getting a good initial price for their impact.

Like decision auctions, impact certificates do seem biased towards optimism. I don't know of a way for impact certificates to account for actions with net-negative value, which means philanthropist-investors could over-value things by ignoring risks (and face less economic downside than they should). 

The big advantages of all this is supposed to result from a robust philanthropic-investor economy, with big established philanthropic-investment firms, small philanthropic-investor startups, widely used impact-certificate marketplaces, etc.

This is very similar to my idea for judgeless prediction markets. Call it belief certificates. You buy things because you think they are the wrong price; you buy what you believe. If enough people are buying/selling in this way, then you don't have to question too often whether you're doing it as an investment or because you intrinsically value correcting the market (maintaining the truthfulness of the shared map). (Although, I tend to think this works best as a play-money mechanism.)

This is the same idea I was trying to get across in Judgements: Merging Prediction and Evidence. Showing the 2x2 grid illustrates a connection between the moral (1,2,3) and epistemic (1,2,3,4) threads in my recent writing. 

But does it work?

All three people I can recall describing impact certificates to over the past year have expressed skepticism of the money-maximalist kind. Where does the value come from? Aren't you just hoping for a bigger sucker to come along? Financial instruments with no underlying value will be low-volume. Why buy an impact certificate when you can buy stocks that earn money, to then buy bread instead?

Bread is great, but it is not the sum total of my values. More charitably: it is not actually the case that I can best serve my values by totally factoring my strategy into [best way to earn money] x [best way to spend that money to get things I want]. 

People buy things for all sorts of reasons. I don't believe these mechanisms are against human nature or against the nature of economics. Setting aside potential legal difficulties, 

However, cultural acceptance is needed for robust markets to emerge. It might possibly be too far from the prevailing money-maximalist mindset (or face other hurdles).

I do think all these ideas nicely illustrate the connection between theories of rationality and institution design.