Predictions Affecting Proposal Passing Chances


#1

This is part of the Genesis Protocol V0.2 research umbrella.

The specific topic is predicting’s influence on passing proposals.

Here is the currently implemented protocol.

The Problem

It is possible to significantly influence the outcome of voting via predicting, because boosting a proposal drastically raises its chances of passing.

The chance of a non-boosted proposal passing thus far, and also in theory, is extremely low (0 have passed so far, I think?), because attaining an absolute majority is so difficult. The chance of a boosted proposal passing is dramatically higher — 50% or so in Genesis as of now (flying by the seat of my pants, but the point is there) — since the passing requirements are so much easier to meet. This striking difference means that if a staker can contribute significantly to a proposal becoming boosted or staying non-boosted, they are having a very strong, direct effect on that proposal’s passing chances.

(Notably, proposals obviously not aligned with the DAO will still overwhelmingly fail if boosted. Rather, this problem comes into play if two similar, reasonable proposals are submitted, and only one is boosted — that one will have decent passing chances, while the non-boosted one will have near 0.)

Staking is not meant to have a strong influence on proposals passing. It’s supposed to exert some influence, as the purpose of staking is to direct attention, and a DAO operating under any responsible protocol is unlikely to pass proposals that it isn’t paying attention to, but if that influence is too strong, the entire system falls apart. If staking and voting had equal influence on a proposal’s passing chances, staking becomes a form of voting, Reputation loses much of its meaning, the DAO becomes nearly open membership and risks losing its sense of “self” and mission.

Is the link between staking and a proposal’s passing chances too strong right now? I think it might be but haven’t put a ton of thought in yet.

Solutions?

The first thing that occurs to me as a solution is to reduce the effect of boosting on a proposal’s chance of passing. One way of doing this would be to switch to a continuous boosting protocol, where there is no huge protocol difference between a boosted proposal and a non-boosted proposal. Instead of going directly from a 30-day limit and a 51% quorum to a 3 day limit with no quorum, proposals would slowly ramp down the time limit and quorum as their staking score goes up (and staking would be prohibited whenever a certain proportion of the passing voting score is met). With a continuous boosting protocol, a stake’s influence on passing would be more related to only the size of the stake, rather than the size or the timing — only a huge stake could have a huge effect on the proposal’s passing chances (assuming the cost of staking is balanced well), whereas now a small stake can have a large effect if it pushes the proposal past the boosting threshold.

Another tool that might help is allowing proposals to be un-boosted, at least for a time after boosting. This isn’t a great solution, but it should help some, since other stakers would have more of a chance to “undo” some of the influence-on-passing that the initial stakers achieved.

Thoughts on Generalization

How much influence should staking have on a proposal’s chances of passing? It needs to have some, in order to facilitate value-aligned scalability, but I think we’d all agree that the option to half-way buy a proposal into passing is concerning.

Is this an inherent flaw of staking, even with continuous boosting? Is it not a problem? To me today, the answers to those questions depend a lot on the purpose of the DAO. If the DAO’s goals are easily measurable and the success of a given proposal can be tracked fairly empirically (e.g. a profit DAO), “paying” for votes via staking and similarly open policies might be okay or even beneficial.

In a DAO where the goals are not very measurable, and success is largely assessed through the opinions of the members (like a wellbeing-maximizing DAO or an art-creating DAO), buying votes or anything like it is potentially quite harmful—success for this type of DAO means the expression of reputation-holder (member) preferences, not the preferences of stakers.


#2

Another quick thought here: quadratic staking.

Quadratic voting is a fashionable idea where in voting for X, you commit y^2 “voting resources” to get y voting power. The relevant benefit to this thread is that this reduces the advantage of the wealthy if voting power is wealth, which it is in Alchemy’s staking system. This wouldn’t reduce the influence of staking on proposals passing globally, but it would make it much harder for any individual person to instantly boost specific proposals.

With quadratic staking, I would have to stake 100 GEN to get 10 “staking points” on a proposal, or I could stake 49 GEN on two different proposals to get 7 “staking points” on two different proposals. This means that for someone with 1 million GEN to fully exert their advantage over someone with 100 GEN, they’d have to stake 1 GEN on a million proposals – not particularly scary in terms of influencing the DAO.

Boosting and rewards would both be related to the “staking points” score. For example, if I upstake 9 GEN (2 points), you upstake 100 GEN (10 points), and the proposal passes, I receive 2/12 of the reward and you receive 10/12 (as opposed to 9/109 and 100/109 in the old system).