notes-groupDecisionMaking-futarchy

"The idea behind futarchy was originally proposed by economist Robin Hanson as a futuristic form of government, following the slogan: vote values, but bet beliefs. ... individuals would vote not on whether or not to implement particular policies, but rather on a metric to determine how well their country (or charity or company) is doing ... Given a proposal... prediction markets would be created each containing one asset, one market corresponding to ((each choice)).

((eg)) Suppose that the success metric chosen is GDP in trillions of dollars, with a time delay of ten years, and there exists a proposed policy: “bail out the banks”. Two assets are released, each of which promises to pay $1 per token per trillion dollars of GDP after ten years. The markets might be allowed to run for two weeks, during which the “yes” token fetches an average price of $24.94 (meaning that the market thinks that the GDP after ten years will be $24.94 trillion) and the “no” token fetches an average price of $26.20. The banks are not bailed out. All trades on the “yes” market are reverted, and after ten years everyone holding the asset on the “no” market gets (($1 per trillion of GDP at that time)). " -- https://blog.ethereum.org/2014/08/21/introduction-futarchy/

"futarchy fixes the “voter apathy” and “rational irrationality” problem in democracy, where individuals do not have enough incentive to even learn about potentially harmful policies because the probability that their vote will have an effect is insignificant..in futarchy, if you have or obtain information that others do not have, you can personally substantially profit from it" -- https://blog.ethereum.org/2014/08/21/introduction-futarchy/

In practice, how should a trader bet on a policy? It's more complicated than it may seem. Let's say that there are two policy choices, A and B, and you think A will yield a higher success metric than B, but the market currently thinks that B will yield a higher success metric than A. This means that the current market price for B is higher than A. So you sell B and buy A. However, since the market currently thinks B is better, there is a higher chance of B being chosen than A. So if you simply buy a quantity of B equal to the quantity of A that you sell, then since it is more likely that B will be chosen, you end up 'net long' w/r/t the success metric. This means that if other, unrelated things change the expectation of the success metric, you are exposed; for example, if something bad (but unrelated to this policy choice) happens, this will reduce the expectation of the future value of the success metric, and you will lose money. What you would prefer is to only make or lose money as the market changes its mind about how good the policy is. The way to do this is to sell a larger quantity of B than you buy A, in proportion to the probability that A will be chosen. A third 'prediction market'-style asset could be created to track the market's estimate of this probability, and a fourth asset could be created which is a composite of sells of B and buys of A in proportion to this tracker, in order to make it simpler to bet on A vs. B while hedging out the risk of the underlying success metric.

The fourth or 'composite asset' mentioned above which buys/sells both A and B in proportion to the asset tracking the market's estimate of the decision is somewhat tricky to realize, as these proportions will be changing over time. Worse, as the decision time gets arbitrarily near, the probability of the expected choice flipping becomes arbitrarily small, in theory necessitating an arbitrarily large amount of the probably-wont-be-chosen asset being purchased in order to remain net flat. The former difficulty is similar to the problem of writing call and put options without exposing oneself via delta-hedging, and can be taken care of by professional traders (who will charge a premium because they remain somewhat exposed during rapid changes due to the fact that actually delta hedging via trading is not instantaneous, so their position will always lag the optimal position a little). The latter can be taken care of by giving up on the goal of having a perfectly ideal composite asset, and instead having a composite asset that assumes that the probability of the less probable choice never goes below some small floor (such as 1 in 1000).

This system allows traders to avoid exposure to the underlying and still directly profit when they realize how good a policy is before the market does, but the market figures this out before the decision is made. However, if the market does not come to agree with them before the decision is made, then, since the effects of the unchosen decision are never realize and the market for the unchosen decision is erased after the decision is made, in order to profit they must hold an exposure to the underlying until the success metric measurement is made a long time in the future. During this time, they are fully exposed to the underlying, and for most decisions it's likely that unrelated fluctuations in the underlying will swamp the profit related to their rightness on this particular decision. This is because, after the decision has been made, their belief that the decision was mispriced translates into a belief that the underlying is mispriced.

Clearly, the power to set the futarchy agenda is important. But there is a proposal to decentralize this; have new questions at a fixed temporal frequency, and let anyone bid for the right to post the next proposal; they get a prize (their bid back, plus more) if the proposal is adopted.

Problems:

example of manipulation:

"Suppose P[1] is "send tokens to Wei Dai to stock his wine cellar" and P[2] is "send tokens to development team to fix bugs and add new features". P[1] is obviously bad so if the system operates as intended, it should be predictable from the start that the probability of P[1] being adopted is very low. But in that case, the prices of R[1] and S[1] denominated in external currency must be very low (since everyone expects they'll become worthless at the end), which means I can buy them up very cheaply. Once I control all or most of R[1] and S[1], I can then trade with myself in M[1], jack up the price of S[1] denominated in R[1], and make P[1] pass. This shows "everyone expects P[1] won't pass with high probability" is not an equilibrium of this system, which means the system must quite often pass obviously bad proposals." -- https://blog.ethereum.org/2014/08/21/introduction-futarchy/#comment-1554195290

Links: