search

GovXS / Evaluating-Voting-Design-Tradeoffs-for-Retro-Funding

Open-source simulation framework to measure how different voting designs perform against a number of typical retro funding design goals. Achieved by simulating voter behavior and applying formal, axiomatic reasoning.
MIT License
3 stars 0 forks source link

Robustness / VEV / Skewing Results (Metrics discussion) #24

Closed AngelaKTE closed 2 months ago

AngelaKTE commented 2 months ago

Description

Hey Measuring to what extent a single voter is able to skew the voting result by assigning the max amount of tokens available to a single project sounds like a nice thing to have.. @Idrees535 I guess this means that when given an instance and a voting rule: for all projects, we test for each voter what happens to the outcome (the L1 distance from the original final allocation) if he allocates 90%-99% of his funding and the remaining to all other projects (since there is an obligation to vote on all projects in round 5? if not we can just make him allocate 100% of the funds to one project). Then we take the maximal L1 distance from all projects and voters that represents the maximal skewness that can occur for that voting rule and instance, and then we can just find the average maximal skewness.

Why here?

I went over Jonas comments again. Reviewing our conversations, I can see that Optimism main concern is that voters assign the max. amount of funding to a particular project (VEV - Voter Extractable Value).

The votes will use the max available voting power within the round to try to increase the OP amount for the projects. For example, in round 3 the max vote was 5m OP.

I wonder if we could adapt our "Robustness" metric. Could we measure to what extent a single voter is able to skew the voting result by assigning the max amount of tokens available (c) to a single project? Also, what (similar) setup would work in round 5 where voters HAVE to vote on all projects? Could we use a scenario where they assign small numbers (e.g. below 10K to other projects, and the main junk to their favorite project?) And finally, could we put it in context to different voting scenarios, looking at small numbers of voters and projects (relevant for round 5). WDYT?

What we measure:

a) we take the maximal L1 distance from all projects and voters that represents the maximal skewness that can occur for that voting rule and instance, and then we can just find the average maximal skewness.

b) TBD we measure how much funding this single voter is able to extract (in absolute value, or in % of total funding)

linear[bot] commented 2 months ago
GOV-32 Robustness / Skewing Results (Metrics discussion)

#### Description Hey Measuring to what extent a single voter is able to skew the voting result by assigning the max amount of tokens available to a single project sounds like a nice thing to have.. **@Idrees535** I guess this means that when given an instance and a voting rule: for all projects, we test for each voter what happens to the outcome (the L1 distance from the original final allocation) if he allocates 90%-99% of his funding and the remaining to all other projects (since there is an obligation to vote on all projects in round 5? if not we can just make him allocate 100% of the funds to one project). Then we take the maximal L1 distance from all projects and voters that represents the maximal skewness that can occur for that voting rule and instance, and then we can just find the average maximal skewness. #### Why here? I went over Jonas comments again. Reviewing our conversations, I can see that Optimism main concern is that voters assign the max. amount of funding to a particular project (VEV). > The votes will use the max available voting power within the round to try to increase the OP amount for the projects. For example, in round 3 the max vote was 5m OP. I wonder if we could adapt our "Robustness" metric. Could we measure to what extent a single voter is able to skew the voting result by assigning the max amount of tokens available (c) to a single project? Also, what (similar) setup would work in round 5 where voters HAVE to vote on all projects? Could we use a scenario where they assign small numbers (e.g. below 10K to other projects, and the main junk to their favorite project?) And finally, could we put it in context to different voting scenarios, looking at small numbers of voters and projects (relevant for round 5). WDYT? #### What we measure: a) we take the maximal L1 distance from all projects and voters that represents the maximal skewness that can occur for that voting rule and instance, and then we can just find the average maximal skewness. b) TBD we measure how much funding this single voter is able to extract (in absolute value, or in % of total funding)