My observation is that working groups tend to be the self-appointed board followed by co-opting (election/ratification) of “inner” circle. Whilst it gets shit done, it tends to overlook:
a) those busy and missed the window to object/vote
b) passive participants due to language, shunning or some other access barriers
c) people with key skills but without compensation don’t wish to volunteer (opportunity costs)
The nomination/ratification addresses some aspects but a competitive bidding / boasting round tends to elevate those with the most popularity or biggest voice. These may not be truly representative of the entire Uniswap ecosystem (welcome data to confirm/nullify).
LexDAO has been experimenting with sourceCred, a system that passively monitors discord activity. What it does is looks at activity and creates a graph measuring “engagement”.
Theoretically, if we run the right type of analysis we can look at the contrib/listen ratios eg (from another tool)
For those that want to shift away from a “board” style WG selection towards more do-ocracy WG, this may be a useful meta-gov technique to identify under-represented groups and deliberately engage them, either direct or via delegates to represent their interests. This may be considered when there are DAO-wide impacts like selection of legal entities (deliberately plural).
Questions to consider
-
is the current system of populating working groups good enough?
-
is there evidence one way or another that certain ecosystem profiles are not represented in particiating in governance?
-
should there be options to deliberately include contrary or minority opinions for potentially contentious decision making? cf 10th-man rule or concept of loyal opposition.
|| The reason why this is a sensitive topic is that to implement sourceCred, it needs access to the discord server and comes with privacy concerns. ||
2 Likes
Does your tool work with Discourse activity?
in theory yes (and not our tool, we’re just refactoring it) … this history of sourceCRED was that it was built several years ago on a grant, but after $$$ ran out, the team disbanded. They have a module for discourse but since LexDAO deprecated the use of that platform, we can’t say the most recent version works. Technical debt meant we had to spend some (scarce) paladin time to figure out why even the discord bot stopped working (incompatible upgrades). So if UniSwap wanted to put up a bounty on getting it working for discourse, we would spec the task (we’re scrapping the CRED token reward so its a custom vertical), and farm it out to say RaidGuild.
sourceCRED is still being used by some early web3 pioneers such as TokenEngineering Community and I’ve eyeballed AraCRED but those would be the old system which allocated rewards according to algorithm which hearsay could be gamed. You can even write new modules for other SNS but obviously doing it from scratch is more expensive. The challenging part (which is where privacy concern raises head) is that display handles on different SNS have different IDs so you can’t guarantee against identity theft.
The advantage I see to sourceCRED is that
- its is passive (so no intrusion beyond existing site policies),
- automated (great since legal engineering always short on dual-skill rogues+fighter) and
- the weighting can be tuned to encourage say … looking for long thoughtful pieces by delegates rather than snappy yes-yes-yes which may be evidence of lazy thinking.
MakerDAO used sourceCred as well on its forums. It created a lot of noise and not much value. There was a decent sized group who would post a lot and like each other’s comments, so it read as high engagement. That quickly disappeared when sourcecred was eliminated.
Maybe it can work, but its a big challenge to make a system that is rewarding to high value contributors, but not gamed by people in poor countries for whom SourceCred could be a significant source of income.
1 Like
BlockScience discovered the same with cliques on gitcoin grants. The problem you describe arises from original use of sourceCred to distribute rewards. But if the point is to detect long thoughtful posts then it becomes a threshold test, not allocation.
1 Like
