NNS MicroTasks for spam removal and general clean up

Governance
1

NNS MicroTasks DAO for spam removal and general NNS clean up

Proposal Cost
Proposals cost 2 ICP or a standard 30 dollars in ICP.

Microtasks DAO layer
Participating neurons are chosen based on voting power. The higher your voting power the higher the likely hood you will be chosen.

Neurons must opt in to participate and must log into the NNS and select an option to start reviewing proposals. These neurons are considered to be on stand by and waiting.

There is a count in the NNS dashboard which shows how many neuron’s are logged in and waiting to review proposals but it does not show who is participating.

When a proposal is initiated into the NNS

10 named neurons are randomly selected to participate.

Neurons are blind and cannot see anyone else’s votes…

Neuron’s vote if the proposal is spam or not.

A super majority must be met

All 10 Neurons must vote. assigned Neurons are given 2 hours to vote or they are removed and replaced.

Neurons that voted against the majority are not given a reward.

Neurons that are consistently against the majority get a reduced chance to be selected (to filter out anyone trying to bot this)

Rewards:
The Proposal cost is split and payed out among the number of reviewing neurons. For example a proposal cost of 2 ICP, with 10 neurons reviewing means each neuron would get 0.2 ICP if the proposal passes.

When someone pays to start a proposal, they are essentially paying the named neurons to review it. There is no extra out of pocket from Dfinity. There is no wasted ICP.

When people parties are activated we can let all neurons participate further achieving decentralization.

1 Like
2

There its is. Simple, Functional, affordable, no extra out of pocket cost and it forces people to make damn sure their proposal has at least some community support before submitting or risk losing their proposal cost.

3

@wpb

Hello, can I get your thoughts on this idea here. I am abandoning my previous concept its overcomplicates things. a rube goldburg machine to fry an egg. Let me know what you think of this. its much simpler and easier to deploy and does not affect neuron rewards in any way. The proposer pays for the service themselves.

Honestly, why don’t we just get a human moderator that gets paid from the proposal cost. Wouldn’t that just solve all our issues if we had a couple of trusted moderators who we hire to filter through governance proposals.

4

So what happens when a NNS proposal creator is intending to submits a spam proposal, and now there are 2 spam proposals?

This proposal feels like we’re testing that NNS voters are paying attention and penalizing them for ignorance. It would never fly in the US, and would be considered discrimination against people from different education levels or backgrounds (ability to read, language spoken, etc.).

1 Like
5

I think you misread this concept.

if the creator submits two spam proposals. There are two review groups. who are blind to which proposal they received. they will both believe they received the spam version and vote it as spam. half will get it right the other half will get it wrong. the proposal is removed. Now lets say that the proposer states publicly that he will be submitting both as spam as a way to allow everyone to earn rewards then we have the forum moderator. Its a captcha system based on proposals, that’s it. super simple click on which images are boats type system here. We could allow neurons to abstain from this if they really don’t feel like participating but everyone else who wants to can help the system self clean can participate and earn.

If spam does get through a forum moderator can then catch it as spam. the proposer loses the initial cost, and the reviewers lose their payment. Why would Someone purposely pass spam to lose their payment.

These types of double blind methods are a standard for reducing bias across the world. We are reducing user bias to voting randomly for rewards.

This system is meant to discriminate for those that don’t engage. Its an engage to earn system. However we could allow neurons to not participate if they wish.

I don’t understand how it discriminates anyone else could you explain this further. do you mean it discriminates people because they have to read proposals? I’m confused.

6

I personally don’t like this proposal, there are simpler and more elegant ways to solve the issue, asking people who create proposals to write a “fake” one is weird design.

1 Like
7

Yeah it is super weird but it would work. I thought it could be fun to do manually lol.

Putting the whole double blind thing aside would you be against using the initial cost of the proposal to fund proposal moderators. Like some sort of amazon micro turk system but for the NNS.

Propsers are already paying X ICP might as well put it to good use right?

Its pretty much what im trying to achieve here.

we cannot solve the spam solution without using actual human reviewers who are actually motivated to do it correctly. Machines are currently not capable of doing it. Humans will always take the path of least resistance for rewards so there should only be one path given. even if we go back to the previous non weighted proposal system spam can still be used as an attack method to bloat the NNS. We need a spam solution and it has to be done with live engaged people or at least until AI further develops.

by randomly assigning neurons I was hoping to give everyone a chance to participate but it might actually be better just to have trusted, vetted individuals work as proposal moderators that can be removed or hired like any normal business operation. Is this decentralized. no its not but it would work.

8

Would this not require a notification system in NNS in order for neurons to know they have something to review?

1 Like
9

Yes it could. A notification system would be prefferable. However for now i was thinking neurons log into the microtasks system manually and then are immediately assigned any pending proposals that need to be reviewed. They can basically begin working whenever they want. If the neuron is assigned a proposal and they are inactive for longer then 2 hours they are replaced.

10

This comment is referring to revision #14 of your post.

I’m against a forced 2-hour window for approving/denying a spam post. If this is rewarded, it now opens up another incentive for someone to build an automated “spam approval bot”, so I can sip Mai-Tais on the beach and if one of my neurons gets selected let my bot vote do the work of approving it - this doesn’t handle spam.

Doing this on a per-neuron basis opens up advantages for people who now split their neuron into many small neurons in order to have a better chance of being selected, and therefore collecting more rewards.

Having hundreds to thousands of neurons is not manageable by any one person, which means that developers who are able to build these types of automated tools would make out handsomely from these rewards with their spam approval bots → now we’ve replaced spam with voting bots.



Also, appreciate the enthusiasm, but in the future, if you’re going to make 14-15 edits to the original post, even changing the proposal mechanism it’s best to do this in a new post or refine your proposal a bit more before making an edit. It makes it hard to follow which comments and feedback apply to which version of your main topic post.

11

Understood.

Key points.

When a proposal is initiated named neurons are selected from a pool of logged-in neurons that are active. You are not logged out for waiting but if you are selected to review a proposal and you do not make a decision in two hours your are replaced. You must be logged in and engaged. Its pretty much like a moderation job where you work whenever you want. Look up amazon micro tasks. Its a clone of that system.

Only named neurons are allowed to participate until people parties have been developed. No random neurons are selected. Only trusted individuals. Maybe we can even have the named neurons have to go through an approval process with dfinity or attach a rating system/penalize consistently incorrect neurons.

Last point. It would be very hard to bot this, first you would have to get at the minimum one of your neurons named. Second, neurons vote blind, your bot wont have a clue how to vote. You could have it vote yes or no or random but you will always risk your bot not gaining any rewards if it votes incorrectly. A reward is only given when a majority decision is passed.

I 100 percent believe using human moderators is the only real solution at this time.

12

oh dear… did we get here already?

1 Like
13

Lol. Yeah i know. Its not the greatest but when deciding wether to allow al neurons to participate or only vetted neurons. Vetted neurons are just way easier to control for then opening it up to everyone (until people parties). However. We could use a ticket system. Where every neurons probability to get chosen is based on how much ICP they contain. This evens the playing field however large neurons would have greater advantage of awards. So its became a poor decentralization but fair rewards (basically trusted forum mods) vs fair decentralization but poor reward balancing ( favoritism to large neurons). I chose to bite the bullet for now while we wait for people parties.

14

For me it’s a step too far.

One compromise is to balance the probability as (voting power / neuron size), meaning it is only affected by lock-up time and age, giving priority to 8 year gang and anyone who hangs around long enough.

1 Like
15

I think your right. Voting power as the probability marker would work. Might as well reward those who are most fully invested.

Thank you!

1 Like