Post ideas about a reasonably fair algorithm to identify contributors and measure current contribution to Ethereum-related work of all kinds.
This algorithm could be run in an oracle and used to allocate community contribution rewards. It would be able to access data from the web or Ethereum mainnet, and track contributors and types of work.
Please don’t argue about which is good or bad, just reply with ideas. I will collect related ideas together for focused evaluation.
Spitballing/brainstorming:
I mean, I just posted on the ol’ Twitter machine, but I’ll post here too, 'cos why not?
For reputation you could do some kind of Colony-like peer-assigned and review task-based system. The benefit would be accounting of non code-related tasks. It isnt sybil resistant, but you could potentially mitigate attacks/noise if there were recognized stakeholder groups that managed and reported contributions within their groups, and were then probably audited by other peer stakeholder groups. ex: magicians, cat herders, EF, Parity, Geth, maybe a union of open source contributors, etc.
maybe a way of coming to this would be to start with an open doc that allowed people to list all of the ways in which a person can “contribute to Ethereum.” And then set out prioritizing that list somehow. fuck it, make it a TCR. Why not? and then come up with ways to measure contribution around all of the significant identified contribution types. but the point would be to start with what we consider valuable contribution. I think we too often just measure the things that seem most easy to measure, regardless of the comparative value we place on those contributions, or the relevance of those contributions to a person’s ability to make certain decisions (assuming that youre looking for contribution measurement approaches to identify decision-makers).
Edited to include my brainstormed list of contribution types:
base level protocol coding
dapp development
ethereum research
cat herders
event organizers
magicians
lobbying groups
dapp users
conference attendees
conference speakers
people who write articles explaining things
working for an ethereum company
buying ether
auditing code
contributing to twitter/reddit/other forums
educating people
We should definitely not be afraid to try our best to measure all the types of contributions that are relevant.
Perhaps those in each field of Ethereum could maintain their own measuring systems, updating their part of the algo, much in the way that Google is constantly updating PageRank.
sure - then you would all have to have some way to universalize metrics across and within groups, or it’d get difficult to measure contributing groups against each other. maybe you could identify some meta-level Ethereum purpose/outcomes and prioritize them in some way (TCR or EIP-like process, like Aragon’s AGP process) so that they could be assigned value, and then each contributor group could self organize around how they can contribute to reaching those outcomes. Then individuals within that group would make a claim about their contribution totals which could be audited by the rest of the groups in some way.
Inputs, things to measure, things to track over time:
Things to counteract:
If we’re allowed to use an oracle, then I would be quite interested in incorporating a PageRank-style reputation system in which community members can “endorse” other community members and these endorsements are used to generate contribution scores – an endorsement from someone who is also endorsed is worth a lot, while an endorsement from someone who is unendorsed is worth little. The subjectivity gets around the Goodhart-related problems which we get into when we use objective metrics.
It can also be made somewhat sybil-resistant by limiting who gets a contribution score “to start” (i.e. we use objective github data to determine the “starting contribution distribution” and then use the subjective endorsements to distribute the contribution scores beyond what the data themselves can capture). If someone creates a fake account, their endorsements have zero weight until someone endorses them. So not sybil-resistant per-se, but not trivially sybillable either.
I like this idea because it is grounded in something objective (i.e. github contributions) but incorporates subjective information (endorsements) beyond what the objective can provide. It also means that we don’t have to worry about capturing every objective signal as the subjective assessments will “smooth out” the gaps in what the metrics show.