Outcomes Of Deep Funding Governance Round 1
Over the past weeks, we have conducted our Community Governance experiment.
This experiment served 3 goals:
- Getting feedback from the community on what changes they would like to see in Deep Funding round 2.
- Taking a first concrete step towards calculating reputation ratings based on constructive interactions on our portal, and use these to give extra weight to the related votes.
- Compensate the contributors for their efforts by distributing the fixed amount of 100.000 AGIX based on their contributions in the program.
Much of this functionality was made possible by the new web3 login offered by our partner, Swae. This function enables us to connect portal activity with voting activity, without any need for KYC. All participants are primarily identified by their wallet address and any other information they choose to share voluntarily.
In this article, we will disclose the results of the vote, the reputation scores, and the distribution of the awards and share the details of the logic behind all this.
Spoiler / TLDR:
Goal 1: The results of the vote decide that in Round 2 we will have a special pool for existing services, ánd a pool for projects in the ideation phase.
Goals 2: Already at this very early stage, the collected reputation weight of some of the voters had a material impact on the outcome of question 2!
Goals 3: The rewards for contributing ranged from 1,583 to 10.500 AGIX! Three active individuals will all receive the highest rewards, and lower amounts are distributed to other contributors. Thank you all for participating!
Detailed Results Of The Vote
As a reminder, these were the questions and answers:
Question 1: Should we create a pool for searching and onboarding existing AI services?
Possible answers:
- Yes, it makes sense to have a specific pool for existing services as an alternative to envisioned services or complete solutions.
- Yes, but let’s not limit this to a pool, but dedicate a full round to this approach!
- No, although onboarding existing services is important, stimulating this should not be the purpose of the Deep Funding treasury.
Question 2: Should we have a small pool in round 2 just for projects in the ideation phase?
Possible answers:
- Yes, even though there is no immediate impact on the platform, it enables people to bring the best ideas to maturity and get early feedback.
- No, if the team is not able to create a proposal with sufficient detail, they are probably not ready to get funded at all.
The Results
After accounting for a quadratic rating of the votes and extra voting weight for high-reputation contributors, these are the results:
Visual Of The Voting Event:
The visual was generated by Robert Haas with his open-source graph visualization library [gravis].
The dots in the middle represent the wallets ordered by their voting weight (highest to the right). The orange dots represent wallets with added voting weight based on their reputation score.
The squares at the top represent the 3 possible answers to Question 1. The squares at the bottom represent the 2 answers to Question 2.
Click here for an interactive version that gives a very good intuitive representation of the voting results.
Engagement
Overall, counted as the sum of individual tokens, 19,222,490 votes have been cast over 2 questions. This has been executed by 84 unique wallets representing a total of 9,947,258 tokens. Compared to the 158 individual wallets, (representing 23,040,436.58 tokens) that voted in Round 1 of Deep Funding, we think this is a good score for the first governance round!
After transforming the token amounts by taking the square root per wallet, this resulted in a total of 18,332 calculated votes. After applying the multiplication rates for reputation scores, this resulted in a total of 25,715 qualified votes available per question. A small percentage of voters (6–8%) only voted for Question 1 or Question 2.
5 wallets that voted were also active on the portal and these wallets gained extra voting weight varying by a factor 1–5 based on their reputation score. During the governance experiment, the following content was created on the portal:
- 16 proposals
- 269 Comments
- 198 upvotes on other’s comments
- 3 downvotes
- 14 ‘icon-based’ positive reactions (smileys. hearts, etc)
- 0 Icon-based negative reactions (sad, angry)
Rewards Distribution
Based on the calculated reputation ratings, this is the distribution list of the 100,000 AGIX reward tokens:
Some remarks related to these numbers:
- As you can see, the 3 people that are professionally active on the SingularityNET Foundation will not get any rewards disbursed, to avoid any perception of unwanted bias in the process.
- There has been 1 person with a web3login that didn’t create any comment or proposal and therefore didn’t get a reputation score, but was still legible for a voting reward.
- There have been a number of people active on the platform, gaining reputation points, but not engaged in the voting event. In some (if not all) cases we attribute this to the limitation of the platform to voting with AGIX on ETH. (With some people from Cardano’s Catalyst being very supportive in our governance groups.) We hope and expect this limitation will be removed in the next voting round.
Reputation Scores Gained:
Scoring Rules Details
So how did we get to these results? Transparency is important. Especially at this early stage where we are still learning and adapting. These are the scoring rules we applied to this first round.
Ratings
We valued different interactions on the platform with the following rates:
Proposal score:
We attributed points to the act of creating a proposal and gave additional points for the rating this proposal received on the portal. The algorithm we used to define the reputation score is based on the product of [the average rating * the number of votes of each vote] (we ignored negative scores here). This gives every proposal a raw score.
In the next step, we presented this score as a percentage of the total raw score of all proposals. This percentage was multiplied by the number of proposals, to give it a value that makes it comparable to future rounds with more or fewer proposals. This is the base score for proposal ratings. This means that the points are proportional to the average rating height and the number of ratings given.
We added a multiplier to this base score. The multiplier has been configured to a number that ensures that +/- 20% of the total reputation points distributed are attributed to the creation and rating of proposals, with a smaller part of the points for the creation and a larger part for the rating given on the portal. This means that about 80% of the reputation points are the result of comments created and reactions given to those comments. So you don’t need to submit a proposal yourself to get a high reputation (but it does help).
This implies that, if we stick to the 20% points gained from making proposals, we will configure the proposal rating multiplier differently in each round, depending on the ratio between the number of proposals and the number of comments and feedback created.
Reputation rating and tiers:
We simply took the sum of all points scored (proposal creation, proposal rating, comments created, and feedback on comments).
Next, we flattened the outcomes by ordering all reputations into a tier system. Currently, we used only 4 tiers: A,B,C,D. In future iterations, as more people have a rating we can increase the number of tiers, and e.g. have a separate tier for that top 5% of users that are omnipresent and extremely valuable. The table below shows how we distributed the reputation scores to the 4 tiers:
As you can see we assigned a voting-weight factor to each of the tiers. This means that individuals in tier A have 5 times the voting weight as a voter without any reputation. This is the case for each AGIX they have in their wallet.
This may sound like a lot, but we believe it is a deserved difference and will enhance the quality of the voting outcomes. The individuals in Tier A created dozens of comments, read many or all proposals, and perhaps created 1 or 2 proposals themselves. So they attributed a lot of time and expertise to this experiment, and therefore it seems fair that their vote pulls more weight than any anonymous community member.
Impact Of Reputation Ratings
It was a very interesting surprise to see that the extra reputation-based weight assigned to some voters already had a material impact on the outcomes of this governance experiment! We didn’t expect this, but it appeared that the number of votes pro/against a special pool for projects in the ideation phase was very close.
Because of this, the weight of these votes was actually enough to flip the outcome!
See the results in the table below:
Impact Of Quadratic Voting
Before the start of the governance experiment, we stated that we would use quadratic voting.
The risk of wallet splitting seemed rather low for a governance vote, and the analysis of round 1 showed that a few high-value wallets can have a disproportionate impact on the outcome. Interestingly, applying quadratic voting did also change the outcome.
Without reputation weights and based on pure ‘1 token = 1 vote’, the outcome would have been this:
Conclusion
We feel that we can look back to a very successful Governance experiment!
Let’s have a look at each of the 3 specific goals of the experiment:
Goal 1: Getting feedback from the community on what changes they would like to see in Deep Funding round 2
We are very happy with the participation, especially in relation to the previous regular round of Deep Funding. We will of course commit ourselves to the outcomes, meaning that in round 2 of Deep funding we will see a special pool for existing services and a pool for projects in the ideation phase! The results are limited to round 2, so after round 2 we will evaluate both initiatives and decide how to proceed.
Goal 2: Taking a first concrete step towards calculating reputation ratings based on constructive interactions on our portal, and use these to give extra weight to the related votes.
We feel that this is also a workable and value-adding activity. We are surprised to see that even in this early stage, with relatively few voters that had a reputation weight, this was already enough to shift the result. Independent from our preference for any outcome, it is interesting to see that in a situation where the votes are pretty much in balance, the most active contributors on the platform were able to cast a decisive vote.
Goal 3: Compensate the contributors for their efforts by distributing a fixed amount of (100.000) AGIX based on their contributions to the program.
We notice that the contributors in this round have been very well compensated for their efforts. Perhaps even a bit too generously! But we also take this as a positive. We hope this will be an extra incentive for other community members to chip in and actively participate in the next round of Deep Funding!
What’s Next
We are preparing for the next round of Deep Funding!
We will include the 2 new pools that were voted on, but there are some details that still need to be decided, such as the total awarded amount for round 2 and its distribution over the different pools (interestingly, this topic was not put forward in the form of a proposal).
There are also some details to be defined and described regarding these 2 new pools, as well as other rules, as outlined in this previous blog.
In light of this successful experiment, we plan to use some of the same mechanisms for round 2 of Deep Funding: Reputation-based voting weights, Quadratic voting, and AGIX incentives for the most active and constructive participants. All this requires some careful definition and description of these rules on our website (Deepfunding.ai).
So while there is stuff to do, we are firmly heading towards round 2. We will keep you posted on all changes and resolutions, and be back soon with the starting date for Deep Funding Round 2!
So stay tuned and join the conversation!
We hope to meet you all in this mission on our Deep Funding portal.
In case you have any questions, feel free to reach out on our Deep Funding Discord channel or our Telegram channel.
