Holographic Consensus — Part 2

This post was collaboratively written with Ezra Weller.

This is the second post in the series on holographic consensus (HC). In the first one, we discussed the scalability problem of large DAOs and introduced holographic consensus as a conceptual solution. In this post, we dive into more details with an end-to-end specification of a minimally viable HC scheme, the Genesis v0.2 protocol.

The discussion thus far can be summarized in the following steps:

1. Large DAOs with large numbers of active members may need to make large numbers of decisions at a high frequency.
2. For the sake of resilience, all those decisions need to be representative of the DAO’s global opinion, i.e. be in line with the majority of the DAO’s voting power.
3. On the other hand, since attention is scarce and decisions so frequent, only a small fraction of the DAO’s voting power can observe each decision.
4. We thus reach a conflict: naive governance protocols cannot allow for frequent and aligned decisions at the same time.
5. Holographic consensus (HC) provides a solution to this tension by allowing local decisions, hence scalability, that are guaranteed to represent the entire group.¹

The proposed implementation below is not perfect, but we believe it to be free of major failure modes and good enough to go live with DAOs launching on Alchemy Earth. A complete technical specification of the protocol can be found here.

The Genesis Protocol

The goal of the Genesis v0.2 protocol is to achieve a minimally viable version of holographic consensus that extends the basic governance throughput of crowd organizations to a large degree, while avoiding major resilience issues. We’ll begin by describing the protocol’s basic mechanics.

The DAO system consists of smart contracts that can control internal or external assets and rights. It is activated by agents, identified by Ethereum addresses, which interact with the DAO by submitting proposals for its actions, then voting and staking on them:

Importantly, the DAO itself downstakes a specific amount of GEN on each and every proposal to help incentivize initial predictors to search for good proposals.² Notice that the DAO (since it is downstaking) pays only to successful predictors of proposals that are approved via the holographic mechanism (see below). This payment reflects the cost of outsourcing the navigation of the collective attention to an efficient market, and hence the effective cost of scalable and resilient governance. Even with this outsourcing, the decisions are nevertheless made by the DAO’s reputation holders alone.

The decision-making flow in the DAO goes as follows:

Boosted proposals allow for greater scalability; they still some voter support for approval, but they are exempt from any voting-quorum requirement (relative majority). The boosting condition is designed to also maintain representativity of outcomes and thus resilience of decision-making:

• Every proposal has its market Confidence, equal to its total upstakes divided by total downstakes, C=S+/S-. The Confidence represents how likely predictors believe it to pass. For example, a Confidence of four means that predictors think the proposal has a four-to-one chance to be approved by voters. Predictors who disagree with the market, and think a proposal’s Confidence is higher (lower) with respect to its actual chance of getting approved by voters, have the incentive to place a downstake (upstake) until predictions finally reach saturation.
• A proposal becomes boosted if its Confidence score is above the boosting threshold continuously for some time, similarly to the quiet-ending feature. This extra-time requirement gives dissenting predictors a chance to place their stakes by preventing instant-boosting (which opens up an attack surface) and allows the saturation of predictions for the sake of maximizing security.
• The boosting threshold we use in Genesis v0.2,

grows exponentially with the number of currently boosted proposals, N_B, with α >1 being the boosting difficulty parameter. The dependency on N_B ensures defensibility against the dilution of voters’ attention.

Although the Genesis protocol may seem complicated at first sight, after dozens of protocol explorations we have found it to be the simplest HC implementation that properly mitigates the tension between scalability and resilience in large DAOs. In forthcoming posts, we will explain in more detail how and to what degree the Genesis protocol resolves this tension, and provide a lengthier discussion of the protocol’s possible failure modes.

Learn More and Get Involved

• Dive into Alchemy and on-board yourself to the Genesis DAO.
• Governance researcher or decentralist advocate? Join the conversation on DAOtalk.
• Join the DAOstack community on Discord, and follow on Twitter.

¹ Local decisions are made to contain the whole group’s preference information, resembling a hologram.

² With the DAO downstake, predictors have the incentive to search out good proposals. Once a proposal gets upstaked, predictors also have the incentive to downstake it, if they think it’s going to fail.

³ If time outs allowed the prediction game to resolve normally, whales would be economically incentivized to try a censorship attack: putting downstakes on proposals so large that the proposals will never become boosted, making it very likely for those proposals to time out (and give the whales back their stake plus a share of the upstake, if timeouts resolved as “fails”). If it is profitable, the ‘whale censorship’ attack is dangerous. A non-profitable whale censorship attack, on the other hand, is risky enough for the attacker to be non-viable. We’ll cover a detailed analysis of the protocol’s failure modes elsewhere.