Our official application is contained in this link:
The following text is the transcription of our aplication.
We are publicly declaring our intent to join the W3F’s Decentralized Voices program as part of the fifth cohort. The proposed design is the first expected evolution of our governance approach.
The current design is composed of two parts:
- The Model
As others in Polkadot and outside of Polkadot [ Crypto Governance Is Broken. AI Can Fix It. ][ https://polkadot.subsquare.io/referenda/1679 ][https://www.cyphertux.net/articles/en/projects/creating-polkadot-ai-agent-openai-sdk\\\\\\\\\\\\\\\][ Polkassembly - Referendum #1463 ] have accurately pointed out already, AI agents (AI to DAO), AI aid (AI to human voter) and exploratory analysis (descriptive statistics), are important tools against the risks of DAO governance over a long time. Namely, scalability, voter exhaustion, anti-competitive behavior, fund drainage among others that OpenGov participants probably are aware of already.
With that in mind, Saxemberg govgineers, treated it as an engineering problem since the early days of DAO governance prior to OpenGov so we have continuously developed models aiming to take over many of the governance duties. The idea is to reduce the FTE of “experts” in favor of a statistical model which performs the decision making in a close manner as if humans had done the voting. The core idea is to minimize the error between the predicted votes and the actual votes of alive entities.
The current model has been trained with the following targets ( GovernanceDocuments/Data/PolkadotScores.csv at 6ea2ea6623145528895fc80a708ada41508cec2c · Saxemberg/GovernanceDocuments · GitHub ) which should be aligned with the current Guidelines and Directives outlined by Saxemberg ( Treasury Guidelines and Directives | Saxemberg ), in addition to the same goals presented in the applications of cohort 1 through 4 and biased towards negative as the current price of DOT remains depressed (lower price means a negative bias higher price means positive bias).
The model result will be executed by the following address:
DOT
11fx8xKPNd4zVSBxkpN8qhhaGEmNJvPgKqwhDATZQXs7dkM
KSM
HcEbeTviCK33EddVN3mfJ6WymWLyKfFuekjhjn5PFirjJ5F
In this way, the “expert” model is executed manually every Sunday and votes are cast accordingly.
The model will still be executed manually with the resulting score in case of unforeseen or damaging results so that results can be revisited. The second issue are the flip-flop referenda which change to “vote NAY” for different reasons, which a simple abstention doesn’t work well enough (more on this later). The emphasis at this stage is the second part which should be the element that is capable of damage control, decision making, bribe resistance, etc. in the long term.
The prime directive is to increase governance scalability while reducing FTE. This is what has been identified as the best approach towards scalability in onchain governance. We estimated that we reduced the FTE was from ~7 to <1 with this pipeline.
- The Sax Guild
This will be the emphasis of this cohort.
This group of individuals will have the capacity to change the vote to ABSTAIN through a secondary vote. It should include all individuals who are willing to engage with the voting process who participated in the governance experiments conducted earlier this year as well as all the other agents who were interested but were not able to enter due to the size limit and are still willing to commit their time to OpenGov. Other overlapping members from other voting groups or guardians will be excluded from the list.
The Sax Guild space on Subsquare, governed by the SAX token on AssetHub with ID # 48 will be the one responsible for the vote. The Sax Guild voting is an off-chain process in a simplified manner. To only vote AYE if you want the model’s vote to be recast to ABSTAIN, NAY otherwise. In addition, the DV Web3 rules have included the one-person one-DAO rule so distribution of this token will be limited to the extent possible to those who don’t belong to other voting DAOs as well as light DVs individuals despite the participation in previous experiments (many who participated were members of KusDAO).
This is the way to:
1. Control any potentially harmful vote presented by the model.
2. Be the ones responsible for any desired change vote cast by the model.
As such, the Sax guild acts as a counterbalance to any vote that can be legitimately harmful as we had some in the past.
In addition as other DAOs focused on bribes and PolitiFi (https://lobbyfi.xyz/) have already shown, bribes presented onchain either as incentives, direct payments, distributed payments or alternatively punitive measures can only be balanced equally by this party that’s capable of nullifying the vote, therefore, making this model a potentially bribe resistant / bullying resistant approach to onchain voting.
Finally, regular conviction voting is encouraged as it’s also the purpose of this guild. If there is enough conviction in voters to nullify the vote of the model then it’s a perfectly acceptable behavior.
Details about the SAX token
This token will be initially distributed at the minimum amount possible and issued without any cap so that future distributions can cancel out voting influence of the current participants for whatever unforeseen reasons.
This is a work in progress, so future token distribution will depend on past work as well as changes in voting entity membership. Increase in economic alignment will be heavily sought so other distribution methods will be tried as well as different thresholds. Considering that the number of voting agents at this moment is rather small, a manual initial distribution looks like the best approach while we continue testing the unknown unknowns of this approach.
Token transfer dynamics are to be explored too in case membership is willing to be sold by the tokenholder.
Future
More data. We have already processed the Kusama and Cosmos governance datasets which also contain rich data. The Internet Computer governance contains a gargantuan amount of referenda with superficial information, however the inclusion of its features might prove beneficial as more data equals better model. Other onchain governance systems like Arbitrum and NEAR will be included in the future.
Release of raw & processed data. An always challenging approach as newcomers will have no way to build the ingestion pipelines right off the bat. However, this data will prove beneficial for other applications inside and outside the Polkadot ecosystem, most likely exploratory data analysis.
Expansion to diverse models and architectures. This approach is one of many possible so different approaches might be tried in parallel, afterwards or merged together as time progresses depending on results.
To deploy a similar solution in other chains to test the validity of this approach.
Pseudonymous identity as the default. Ideally speaking, the identities of the members will be known. The identities in the conversation and votes will not be known. Privacy when it comes to decision making helps foster conviction in individuals. See discussion: https://github.com/opensquare-network/collaboration/discussions/1117
The list of the tokenholders is being worked as we speak so that they do not overlap with other voting DAOs (i.e. excluding members from ChaosDAO, KusDAO, Nexus, TruthDAO, as well as guardian candidates, etc.)
F.A.Q
-Will the Sax Guild be able to change votes to AYE or NAY at some point?
To allow others a vote overturn is one of the objectives, however, it might not be done by this token or it will be done by other guild. The idea is that economic alignment is likely the best way forward with this situation.
-I liked the way Saxemberg voted before, can we have that instead?
The current approach was already in full swing during the last stage of cohort 2 (the heaviest in terms of number of referenda) and during the entirety of cohort 3 & 4 so the votes will likely be similar to those without many noticeable changes.
-I liked the feedback and guidance that Saxemberg was able to provide, is that going to continue?
Unfortunately that is not going to be in the covered areas of the cohort application anymore. There already exist entities that fulfill that role and they are already incentivized by OpenGov to do it. Feel free to reach them and engage with them. The idea is to create an issue if heavy mistakes in the training targets have been made or to create your own models with your own targets so that it resembles the way you’d like to vote. The appropriate way to disable the vote is to engage with the members of the Sax Guild. Rationale and other light engagement based on previous experiences will be available however. Direct contact, DMs, calls, meetings will be discouraged for now.
-Do the tokens have any value?
The transactability and impact in governance is something that is going to be tested by this token in this cohort. That is the main objective. Token membership and transactions derived from it will be observed to see how to strengthen this approach.
-Can I buy the token from you?
No. The only ones that have the token are the members of the Sax Guild and the minter address is able to mint more.
-In the future SAX tokenholders can take membership on other voting DAOs but they will keep the tokens. How to approach this challenge.
The idea is to keep increasing the number of tokens in future distributions so that the influence from early members doesn’t get embedded forever. This can become more ossified as time progresses.
-SAX appeared on the EVMs. Is that the real deal?
If Hyperbridge was used, then it is the real deal but it’s just tests done by our team.
