Updates to SourceCred

Updates to SourceCred

Over the past few weeks, @befitsandpiper & the Fauna swarm have been making extensive adjustments to sourcecred to make it more fair for everyone, and to reduce the effectiveness of people trying to game the system. The following are the changes that we’ve made to last distribution, the changes that are now live on sourcecred but did not affect last weeks distribution, and the upcoming changes that we are proposing. As any changes or updates happen, the post will be edited to include new info. We would love your feedback on how sourcecred has been for you so far, and potential changes you would be interested in seeing. Thanks for your input as we work hard to make this as effective as possible for everyone.

Changes to Last Distribution:

  • Discord: ‘Super emojis’ now give the proper amount of cred (2x not 1x)
  • Discord: Unverified users no longer mint cred (their emojis do not give any cred)
  • Discord: Fauna/seeds mint cred 2x over a verified user (a fauna/seed emoji will count for twice as much as a regular verified user)

Changes on the Live Instance:

These changes were not included in last weeks distribution

  • Discourse (Forums): likes on posts has been reduced from 8x (relative to discord standard emojis) to 4x (50% decrease)
  • Distributions will be switched from Friday’s to Monday’s to fix the 1 week delay on SourceCred as SourceCred switches over Sunday night.

Upcoming Changes:

  • Role-based cred minting
    • Fauna/seeds mint cred 1x, verified users 0.5x, unverified users 0x
    • This means a fauna/seed emoji will count for twice as much as a verified user
    • Implementation of a new ‘trusted’ role called ‘Bee’ that will mint cred at either 1x or 0.75x
    • Currently working on implementing this now
  • Channel weighted cred
    • meme channel getting less cred (0.25x or 0.5x normal channel)
    • possibility of higher impact channels getting higher weights (maybe check-in?)
    • 0 weights for information channels
    • lower weights for exclusive channels like fauna - to not advantage members with privileged roles
  • Review data on Github, may need its cred doubled
    • 2 potential options here:
      • double PR weights AND double commit weights(this disproportionately advantages seeds)
      • quadruple PR weights, keep commits same weight(this incentivizes PRs as our preferred workflow, and improves the cred UI)
  • Cred budgeting
    • As proposed by mZ, cred budgeting partitions the total new minted cred into plugin level buckets. The following is an example of weights, not necessarily the values we will use.
      • X = 50% to github plugin
      • Y = 30% to forum plugin
      • Z = 20% to discord plugin
  • Asymptotic discord reaction weights
    • each additional react on a message counts for half the previous react
    • spamming infinite emojis on a message has a capped cred value per person reacting
  • Automated onboarding for pollen

List of Additional Concerns:

  • Discourse: Member trust levels 3 & 4 currently have 1.25x & 1.5x cred minting (they give additional cred to others with their likes)
  • How are these changes affecting past and future source cred? Screenshots & data on these changes would help to gauge effectiveness

Current Configuration:

Discourse (Forum)

  • Post Likes: 2x * trust modifier
  • Post reply: 1x * trust modifier
  • Connections to a post: ~1x

Discord:

  • Reaction (emoji): 1x * trust modifier
  • Reaction (super emoji): 2x * trust modifier
  • Mentions: ~1x

Github:

  • Pull request: 2x
  • Commit: 1x
23 Likes

Upcoming Changes:

“possibility of higher impact channels getting higher weights (maybe check-in?)”

Does this run the risk of a double-whammy, whereby checkins are usually liked by a lot of the seed/fauna members. Because they are minting higher cred, and the channel also gives higher cred, it may give higher than the intention?

Or perhaps the checkins/contributions are so undervalued right now that this change would bring them in line with expectation/the desired outcome - just something to consider and monitor

/edit Apart from that it looks really good and considered, and nice to have a simple/concise list of current config/weights at the end. Appreciate the time you have taken to write this up

6 Likes

It’s possible. We could leave the channel with standard weighting

2 Likes

This was my concern also and why I’d rather only decrease specific channel weights and not increase any. If channels the community value less are decreased then all other channels have a slight relative increase anyway, which I think is good enough.

2 Likes

Meme is a good way to make something go viral. Have you ever considered why some project get a huge strong community just thanks to meme ?

3 Likes

Would it be possible for users to verify their social media identities so that 1hive can survey social media platforms such as Reddit or Twitter for non-Discord contributions by users? That way those of us who are very active on Reddit, for example, can still be acknowledged for our contributions to community growth and education?

1 Like

Do we have a way to measure the impact of SC changes compared to the previous configuration?
It would be good to have some metrics to learn the effects so we know the best way to implement new changes in the future to accomplish the results we want.

3 Likes

i have ask to sourcecred developper and it’s seems for now, it’s not possible to connect facebook, twitter and reddit to the system for adding cred.
Maybe in some days, they will develop it?

1 Like

Excellent work guys. I really like the pollen initiative.
Thanks @Blazar and @befitsandpiper for making the details of the process clear to everyone.

Excellent news Blazar, i guess there will be always need for tweeking till we will find the right balance for the community which is ever evolving.

My question is which are the super emojis ?

2 Likes

Those four are the super emojis:

emojis

5 Likes

Thanks i did not knew that … I guess we learn something new all the time.

5 Likes

Yes. Use those power emojis wisely.
Please don’t ever react to posts with 2+ different emojis, nor click on every type of emoji that might already be there.

If you think someone was very helpful, you can give them one of the super emojis. If you just like a message, you can use any of the regular ones.

4 Likes

We’re not gonna ban people for using emojis, come on. Also there are plenty of reasonable times to add 2+ emojis to people’s responses. And anyway, we may have some sort of asymptotic cred minting that caps the total amount of cred people can mint anyway.

The issues that we’re seeing I think are more around people trying to sybil, or collude in ways that don’t add value to the community. But it is not so simple to say this is a ban-able offense.

11 Likes

Of course memes are important!

Giving a meme of the week bump was suggested on Discord.

We don’t want people just spamming memes.

A meme limit per person was also suggested, but I’d hate to see that. It’s hard to know which meme will take off sometimes.

I think the meme of the week, paying to several places, will produce better results. People will put effort into being their best (tbh I’ve seen some sweet sweet work on the Discord), and once a person has published they won’t feel a need to create more to harvest more likes, which will also give them time to contribute more on other servers.

I haven’t thought this through, but it may also be possible to award HNY based on votes with everyone getting one vote?
Even better if it can be split.

So instead of picking winners just everyone vote and each meme that got a vote gets that vote’s share of the HNY.

This is all so much work to figure out.
Love watching governance in action.

2 Likes

Hey @Santigs, we currently kind of eyeball it. When @befitsandpiper runs his local instance with the changes that week we have a look at the distribution to see if the changes make sense. There’s no real method to this other than us judging whether we think the distribution looks better or worse. Usually this is done over the course of a few days so we can get it as right as possible. Hopefully we’ll have better tools at a later date.

4 Likes

That is great. At least you test locally to have an idea of how the changes are going to impact. I am working on getting the info from discourse right now and trying to get the first graphs out of it, from there we all will start getting graphical info from different angles and have a chance to see how our involvement + any changes are affecting the distributions.

2 Likes

You can vote on memes in discord with your reactions. This should in theory be how the memes channel works. But we’re getting a lot of “I scratch your back, you scratch mine” type behavior when the memes are total trash. But, some sort of asymptotic decrease in cred for reactions per identity rather than linear cred per reaction will probably mostly fix this problem.

4 Likes

hello ~

we dont need to be good to each other , but lets try to be fair ~ i think giving a room for people to have place where they can easily get cred. they make the time for it and log in to the discord ~ is really valuable for the community . the problem is abusing it ~ which i suggested before to have cap on meme channels ~~

1 Like

I like this idea a lot.
In that case, would the reactions per identity be distributed equally among every1 that recieved them? Or would the latest reactions be worth less as the cred minted reaches the asymptote?