Technical Working Group: Governance

Governance
22

My Introduction and Governance Interests

Hello everyone. This is my very first post on the DFinity forum. I’ve done my best to find the appropriate place for this post, so please redirect me if I didn’t hit the right category and thread.

I am quite interested in contributing and collaborating to improve NNS/SNS governance, so I would like to be a part of this technical working group. I am not a software developer, but I have other technical skill sets. I am currently scoping out an experimental project for my PhD starting next year. One focus of the project would be improving the state of the art with respect to the nascent field of computational social choice. This research could help fill a huge gap in current NNS governance, where prioritizing alternatives, optimizing resource allocations, and incorporating confidence/uncertainty levels are all currently not possible. This is obviously due to the extremely limited information that approval voting can provide (as used in all DAOs), where only one proposal at a time can be considered in isolation, and only a binary “yes” or “no” response is permitted.

Computational Social Choice as a Path to Address the “Impossible” Prioritization Problem

At first pass, this sounds like an easy problem to fix. Why can’t we just use a different voting method that works consistently well when prioritization is required? Unfortunately, it is not so easy, as nearly 75 years of modern social choice research has shown, along with lively academic debates before then going back to the time of the French Revolution and beyond. Surprisingly, no reliably sound prioritization method currently exists. In fact, a Nobel Prize was awarded to Kenneth Arrow in 1972 for his axiomatic proof in 1950 that a dictatorship is essentially the only social welfare function that meets some minimal set of aggregation criteria when >2 alternatives are involved. Normally, the Nobel Prize is awarded for achieving what was previously impossible. However, in Arrow’s unique case, it was basically awarded for proving what is shockingly unachievable, per his “Impossibility Theorem”.

Although a perfect preference aggregation method may never exist, there are still many ways to improve and optimize how social preferences are aggregated. This is where computational social choice (aka “COMSOC”) comes into play, which combines the independent fields of computer science and social choice. In effect, COMSOC algorithms can objectively aggregate individual priorities into the most consistently rational (non-cyclic) social prioritization that meets as many key aggregation criteria as possible. These criteria, along with the best computational social choice algorithms for ICP governance, could ultimately be approved and implemented by the NNS, assuming pilot prioritization projects prove to be successful.

How We Could Collaborate on NNS Governance and Why it is Urgent

I can obviously build these aggregation processes and models on existing cloud infrastructure – and that effort is already well underway – just like I could easily capture individual evaluations using existing survey applications. However, my preferred approach would be to build all of this out more permanently on the Internet Computer (i.e., both the front end capture of individual evaluations and the back end aggregation of those evaluations into a single social preference). The ICP blockchain is where I see this computational social choice functionality being most useful in the longer term given the pressing need for all DAOs to have full prioritization (and resource allocation) capability on the blockchain. Approval voting is simply unfit for this purpose and grossly inadequate.

It is a critical job responsibility of everyone in traditional management today that they be fully capable of prioritizing alternatives with consistent rationality. Yet DAOs have no such critical capability, even though their most important role is to make these managers obsolete. To put this in perspective, just imagine how unacceptably handicapped CEOs would be if they were forced to consider every alternative (or investment allocation) in complete “yes/no” isolation, without any relative prioritization, comparative evaluations, or confidence/uncertainty levels factored into their decision process. That is exactly the type of stripped-down, serial decision making that all DAOs – including the NNS – are currently stuck with. Such a blunt decision-making process should be considered just as unacceptable for DAOs to use as it would be for any CEO running a large organization.

Anyway, I will try to cut short my rambling exhortations on this point for now. If anyone is interested in helping to prototype a more robust evaluation and preference aggregation process on the ICP blockchain, then please let me know. I can handle the design and algorithm parts, but I have no expertise to migrate or build this functionality on the ICP blockchain. In fact, I have no idea if it is even possible on the blockchain yet. Alternatively, if someone is just interested in academic brainstorming or experimental ideation around this DAO governance topic, I would greatly appreciate any feedback or input from others.

Finally, it is remotely possible that I could get some funding for this development effort. However, I can’t commit to anything unless the blockchain development is scoped out a bit, which I don’t have the expertise to do.

This Make-or-Break Urgency Applies to the Future Viability of All DAOs

From both my perspective and DFinity’s, governance is truly the “crown jewel” that will ultimately make or break the entire ICP project. It is also what will determine the future viability of all DAOs – running as an SNS or otherwise – since they won’t be able to evolve into larger organizations without socially scalable prioritization processes. Getting governance right is truly that important. Therefore, my hope is that others might grasp this monumental urgency too and will want to collaborate on a blockchain-based solution.

8 Likes
23

@brutoshi, it seems I’m not the only one to make a similar point, copied below from your document:

Besides Yes/No, NNS should leave room for more varieties of questions, such as multiple choices, or re-ordering/prioritization of certain options (roadmap milestones). (BlockPunk)

Please see my other post from today related to the urgent need to incorporate prioritization capability into the NNS/SNS decision-making process. There is a compelling academic reason why all DAOs still rely on the grossly inadequate, least-granular method of approval voting for all decisions, which I describe in that post. However, this extremely restrictive method will not be sufficient once DAOs grow beyond a single dictatorial group effectively dominating all prioritization decisions.

The dirty hidden secret of all DAOs today is that the most important organizational prioritizations and resource allocations are effectively discussed and finalized offline. They are then serially implemented as proposals via a majority coalition or delegated voting control over the DAO. What is especially troubling is that these offline prioritization decisions tend to have the highest strategic impact and involve the greatest risks and complexity. By contrast, isolated, single-issue proposals on-chain tend to be the most routine and involve the least risk and complexity, requiring nothing more than a yes/no approval vote.

In other words, decentralized organizational power is largely an illusion in today’s DAOs. Unfortunately, that illusion cannot scale indefinitely, nor should we really want it to. All of these “democratically-empowered” DAO members will eventually take notice and angrily object once they collectively realize that they have no clothes on, just like the management emperor that they thought they deposed long ago. If we don’t have a solution in place before then that is truly empowering democratically, the entire DAO governance process could ultimately implode.

6 Likes
24

The next meeting is in like 3.5 hours, correct? That’s midnight my time zone, but I’m willing to stay up for this.

3 Likes
25

@Arthur my email is: hello@isaacvaladez.com

26

Hi All,

Is there a central place where the time / zoom link is posted for the governance working group?

27

Hi all, here is a quick summary of short catch-up on the governance working group earlier today

  • Admin: We should announce governance working groups sessions a few days in advance in the IC forum and DSCVR/Discord/Twitter.
  • Good candidates for the next sessions: Guiding principles/Ethos, NNS treasury, …
  • Before a meeting in a big round, sub-groups should work out concrete options/proposals. @Arthur & @aiv offered to facilitate a sub-group on guiding principles/Ethos. Please let them know if you would like to join.
  • Once we have a concrete options/proposal for a topic we can schedule the next session of the governance working group.
6 Likes
28

This idea occurred to me, I don’t know if it could be debated.
I am not an expert on the subject and excuse me if I mess up due to ignorance, but if we want to increase decentralization and participation in governance, why not widely reward the neuron that proposes a proposal that is finally accepted and implemented by the NNS, those rewards can then be divided evenly by the followers’ ceded voting power (voting power) given to that neuron through liquid democracy.

The extraordinary reward, being amount pre-established by the NNS (from unassigned ICPs), would encourage not only the appointment of new neurons, but also their active participation in governance. In addition, since the reward is a fixed amount to be distributed, the neurons with less voting power (for different reasons) would be more attractive to the followers since, in the case of proposing and implementing a proposal, there would be more rewards to be distributed discouraging the accumulation of power and encouraging decentralization.

Obviously, this at the same time would encourage the active action of the community to look for the right partner periodically, discouraging passive voting and the stagnation of power that leads to apathy and corruption.

for example, before creating a given Governance proposal, a contest of proposals will be held with a certain duration so that the community has time to develop on a certain convenient topic. The winner decided by Nns would upload their proposal to vote to be implemented. if it is implemented then there is a prize. in this way we would squeeze the greatest potential of governance and community intelligence

@Sabr thoughs?

29

@esquivada, I’m not sure if I understand your proposed concept correctly. However, the key problem with almost any fixed incentive system applied to subjective personal contributions is that it can be gamed or manipulated based on the known levers/variables to increase rewards. Incentive systems that are less vulnerable to this risk tend to be too convoluted or complex for participants to understand, which would obviously undermine the core motivational purpose of the incentive. Moreover, some proposals are clearly much more valuable (or impactful) than others with respect to the social welfare of the community, so this subjectivity should be factored into the reward process somehow if the incentives are going to be perceived as fair.

That said, incentives that must always apply equally to everyone, regardless of any subjective personal contribution, can of course be fixed and formulaic. A good example would be the scaled inflation rewards for dissolve and holding periods. Although one could argue that these rewards are suboptimal at any particular point in time and should be tweaked (which the NNS can do), at least they can’t be gamed or manipulated by anyone, since subjective personal contributions are never factored in.

Given that DFinity likely initiates (directly or indirectly) a greater proportion of all proposals compared to its ownership percentage of circulating ICP, I’m not sure how rewarding the neuron that makes a successful proposal could help decentralize power. Even if DFinity’s followers could share in the success rewards of DFinity’s proposals somehow, I see an increase in centralized power as being more likely with this suggestion, as opposed to a decrease. Moreover, in a hypothetically paranoid scenario, if DFinity wanted to manipulate more rewards in its favor, it could simply increase the number of proposals it generates and approves via its coalitional voting power.

Keep in mind that the key problem embedded in your suggestion is how to allocate resources (rewards) to participants based on subjective personal contributions. As a result, the only potentially fair solution that I see working in these scenarios is a voter-driven incentive based on prioritized input data - in other words, cardinal or ordinal preference orderings from each voter as opposed to approval votes.

Naturally, that brings us back once again to what I called the “prioritization chasm”, which we first need to cross to make any collective prioritization process feasible and fair. No organization can ever be meaningfully decentralized until it develops a consistently rational capability to prioritize multiple alternatives without dictators. Using approval voting combined with coalitional voting power to rubber stamp single-issue proposals that stem from prioritizations made by shadow dictators accomplishes almost nothing of substance. That’s just decentralization window dressing – an illusion of participatory control that only makes us feel empowered when we are unable to see the reality.

4 Likes
30

I understand your point. but if we were able to channel the suggestions of proposals at the objective convenience of the governance through public contests and voting through the nns. we would solve opportunism. on the other hand to those rewards can then be divided evenly by the followers’ ceded voting power (voting power) given to that neuron through liquid democracy.

The extraordinary reward, being amount pre-established by the NNS (from unassigned ICPs), would encourage not only the appointment of new neurons, but also their active participation in governance. In addition, since the reward is a fixed amount to be distributed, the neurons with less voting power (for different reasons) would be more attractive to the followers since, in the case of proposing and implementing a proposal, there would be more rewards to be distributed discouraging the accumulation of power and encouraging decentralization.

31

@Arthur @aiv I would be interested in participating in this. You are probably not surprised, so I thought I’d offer my help. How can I help?

2 Likes
32

@esquivada, the “public contests” among multiple proposals ultimately just pushes the collective prioritization problem back one step before the NNS vote. However, that key problem still remains.

I have no major issue with your suggestions, including using unassigned ICPs, which I agree might incentivize those token holders to start participating and help with some decentralization. I only have an issue with how collective prioritization could be implemented to make the suggestions work as intended.

I still don’t fully understand what you mean by the rewards being “divided evenly by the followers’ ceded voting power”. If passive neurons continue to follow DFinity blindly, then how could it be that they should receive the proposal success reward? Wouldn’t this just increase their incentive not to participate in voting, which could only decentralize power further? Also, are you saying that popular followees (like DFinity) would get either a reduced reward or no reward at all just because others are following them? Keep in mind that any reverse discrimination against DFinity as a followee would have to be imposed upon all neurons, any one of which could also become a followee by someone else’s choice.

Perhaps you could provide a simple example to explain the logic. As one possible example, pehapswhere neuron A is following the DFinity neuron (“D”), who follows no one; and neuron B is following neuron C, who is also following no one. How would a reward of 100 ICP for a successful proposal be divided amongst them, and how would this reward depend on who submitted the proposal?

2 Likes
33

I’m going to sleep because my wife is going to be angry. tomorrow I’ll give you an example

1 Like
34

just added you to our email thread diego

1 Like
35

@Sabr why should there be a reward for those who pass a successful proposal?
Wouldn’t something like
[ the proposer must stake X amount of ICP to put forward a proposal, if the proposal is successful then they receive their funds back, if its unsuccessful then the staked ICP gets burned to prevent proposal spam ]
be better?
I mean I guess I can understand rewards as an incentive to promote network development, I’m not really sure how I feel about proposal rewards and don’t really see how its necessary.

36

@jacquesSinclair, that was a suggestion from @esquivada, not me. I don’t have an issue with someone making either argument (for or against), but I would not personally support rewards only for successful proposals either. This could promote some group think, where only obvious or very popular proposals get pushed forward instead of taking the risk on a healthy debate or discussion making a controversial proposal more widely supported after being submitted.

As a better alternative, I would support rewards for community contribution value, as evaluated by the community, since relative contribution/participation value by each community member is necessarily a very subjective assessment. I would also support member reputation scores (in multiple categories) driving voting power instead of ICP voting power, at least for certain proposals related to core values / ethics, technical R&D decisions, etc.

1 Like
37

I think I didn’t explain myself well. I want to make it clear that I would not touch the current system of governance.

In the interests of not increasing inflation with the NNS minting rewards from neurons that have not voted, all the ICP of rejected proposals coming from Spam, opportunism or proposals rejected by the community could go to a pot of rewards for the proposals that are accepted by the community.
This would not change the current ROI of rewards nor would the NNS increase the inflation from coining “unassigned” ICP from rewards from neurons that have not voted.

What continue with this?

on the one hand discouraging absurd or spam proposals and encouraging active participation in convenient smart solutions at any given time, and decentralization on the other.

Suppose that the pot of ICP minted from the proposals not accepted by the NNS is 1000 ICP at a given time, then conveniently a neuron X finds an improvement or solves a problem, the neurone X proposes to the NNS and its proposal is approved.
Now suppose that Neuron X has a voting power of 100 ICPs, of which 10 ICPs are from itself and the other 90 ICPs come from 9 other neurons that follow it with 10 ICPs each.
Those 1000 Reward ICPs from the pot would be assigned to the maturity of all those neurons via Neuron X with a 10% weighting for each of them, that is, 100 ICPs for each one.

what can we expect with this?

The extraordinary reward, being amount pre-established by the NNS (from rejected proposals), would encourage not only the appointment of new neurons, but also their active participation in governance. In addition, since the reward is a fixed amount to be distributed, the neurons with less voting power (for different reasons) would be more attractive to the followers since, in the case proposing and implementing a proposal, there would be more rewards to be distributed discouraging the accumulation of power and encouraging decentralization.

Obviously, this at the same time would encourage the active action of the community to look for the right partner periodically, discouraging passive voting and the stagnation of power that leads to apathy and corruption.

for example, before creating a given Governance proposal, a contest of proposals will be held with a certain duration so that the community has time to develop on a certain convenient topic.
To avoid the opportunistic rush of followers, once the proposal is uploaded by the neuron to the NNS, it will no longer be possible to follow that neuron during the voting process.

The winner decided by Nns would upload their proposal to vote to be implemented. if it is implemented then there is a prize. in this way we would squeeze the greatest potential of governance and community intelligence

Make it clear that those neurons that vote independently following or not a winning neuron of the proposal, will also take their reward +~ predictable by the current tokenomics @Sabr

38

Fundamentally, I cannot support rewarding only successful proposals, since that would promote not just group think, but an amplified bias towards the status quo, since that bias naturally already exists. That said, I do recognize that spam proposals can create a lot of soft costs that we should try to avoid.

I would prefer to filter out spam proposals with a relatively low bar applied to reputation scores, where everyone could start with a neutral reputation score/weighting of 1.0. Poor reputations (e.g., from a history of spamming the NNS with frivolous proposals, as judged by all neurons) would lower that score or weighting factor below 1.0. Once someone’s reputation score gets chopped down to a low enough threshold (say, 0.2), then they would be temporarily ineligible to submit more proposals. A slightly higher score (e.g., just above 0.2), might get their proposals down-weighted or subjected to some intermediary approval process before being voted upon by all neurons. In any group decision-making process that continues indefinitely, reputation should matter, and it should be tracked.

Also, I still don’t see how this suggestion from @esquivada would promote decentralization. Since DFinity’s proposals would always tend to have the highest success rate due to their expertise reputation and coalition voting power, everyone would be even more inclined to blindly follow DFinity to cash in on those success rewards. Moreover, any increase in voting power that they get from success rewards would be matched by DFinity’s increase in voting power on those same rewards.

3 Likes
39

I’m not sure how I feel about a social credit score. It could be a good way of filtering out spam proposals, but I dont know if having more moving parts is good.
I think the raw approach of just burning the 10 icp for the failed proposal is a good enough way to disincentivize spam.

In another thread I saw someone talking about implementing ranked proposal, where the proposal would go through different levels of maturity [ idea>motion>implementation] and required icp to submit would be in proportion to it.
All of the spam would settle down at the tier 1 proposals so it would be easily filtered, and it wouldnt disincentivize good proposals from being submitted because of low entry cost

40

Who’s reputation? neuron? an individual can have many neurons.

41

@jacquesSinclair, thanks for helping to clarify my thoughts and how to communicate them. I agree that a generalized “social credit score” for a person overall on all contributions would likely not be appropriate. At a minimum, we would obviously need to differentiate proposal reputations from voting reputations. “Social credit score” is also a very loaded concept, with some social banishment complications that we would want to avoid.

As for proposal reputation scores, which could ultimately get so low to temporarily ban proposals from certain individuals (or named neurons), I agree that these may not even be necessary if we can find an easier way to filter out spam or low quality proposals. This is just one additional suggestion to consider.

As for the voting reputation scores that I referred to, which would be separate, these were more in the context of domain-specific areas. They would generally range from a neutral (1.0) level to higher levels. When certain individuals acquire or repeatedly demonstrate clear expertise in certain domains, such as cryptography, their votes should be given higher weight in those domains (per group consensus) vs. others’ votes in those same domains.

That said, a more general reputation score might also be appropriate, when the topic is something as broad as deciding on a non-technical matter that is very important to the future of the IC. Someone who is demonstrably committed and has contributed consistently for years towards building this IC future in a positive way should obviously have a higher reputation score than someone who is just along for the ride with short-term investment concerns. Again, this score would be ranging mostly from a neutral (1.0) level to higher reputation levels, rather than be used as a social exclusion score.

On your other topic about implementing staged proposals, I agree 100%, since this is how proposals are most commonly done, via nomination first, then voting. Only after proposals are sufficiently nominated (via positive indication only), either by a minimal number or % of votes, could they then proceed to an actual vote. This alone could be sufficient to eliminate almost all spam and low quality proposals.