As we define SourceCred as a social/governance system, we should think about what interest groups will form, and how we can engineer the meta-system so that these groups will act as checks and balances on each other.
At first glance, there seem to be two natural interest groups in SourceCred:
The Cred contributors. They have created the project.
The Grain holders. They have economically supported the project
Now, if we were in the shareholder capitalism model, we would give all of the power to the Grain holders, and Cred contributors would receive the smallest share that the Grain holders can get away with offering. Fortunately, this isn’t shareholder capitalism! (For some related thoughts, see: Purposeful Communities, Closed Hierarchies, and Decentralization)
However, I worry that explicitly enfranchising just the Cred contributors (based on all-time Cred) and the Grain holders (based on total Grain balance) could still create some bad power dynamics. In a sense, there are actually three groups:
Old-guard Cred: Been around since the beginning of the project, lots of cred and social capital, may not be as involved day-to-day
New-blood Cred: Recently started contributing, may be adding a lot of value, not as much lifetime cred
If we only enfranchise the Old Guard and the Grainholders, we may see unhealthy dynamics where those two groups try to pay the New-blood as little as possible, and turn SourceCred more into a gig-ified system where people just get paid subsistence amounts.
The interesting thing about having three parties is that three is the smallest number where two parties can gang up on the third. So we might want to ask, when would two of these parties want to gang up on the third? Which dynamics are helpful, which are perverse?
Old Guard + Grainholders gang up to give new blood as little as possible?
Old Guard + New Blood gang up to inflate the Grain, ruining Grainholders?
Grainholders + New Blood gang up to increase the Cred inflation, inflating away the Old Guard?
Ideally, we’ll want to reason through these dynamics in depth, and convince ourselves that the game is stable, and the different power centers check each other without perverting the system. We might be able to come up with “rock-paper-scissors” dynamics where each group gets to control a system parameter that sets the rewards for one of the other groups.
if the Grainholders have outsize say in how new Cred is created (via boosting?), then they check the New Guard.
If the New Guard has outsize say in how recent parts of the Cred graph are configured, they can check the Old Guard (by making it so that the newly created Cred sticks to recent contributions).
if the Old Guard has outsize say in how much new Grain is issued, then they can check the Grainholders (punishing them with inflation).
The above is just a super initial stab to prime the imagination, would love to see folks propose alternative systems of checks and balances.
Also, as @mzargham pointed out during our recent community call, the game is more complicated because the parties may overlap. The Old-Guard Cred earners may also be big Grainholders, for example.
I am going to unpack this further because I believe that we cannot simply define a balance of powers between 3 overlapping groups. Not only can groups collude, but actually the density of the community may actually accumulate in the overlaps which creates built in collusion, even if it’s not nefarious. In this post I will further unpack the overlapping groups tendency to accumulate ‘mass’, building on the groups @decentralion has proposed.
a. Current Cred Contributors who are also high lifetime cred. These are ostensibly old guard but if the categories are defined by behavior then someone who is still laboring extensively to improve sourcecred is different from someone has previously done a great deal of work and wishes to steer future work or community decisions without continuing to provide labor. When I was researching the feedback dynamics for time based cred versus the snapshot view of cred it became clear that the timeline cred parameters could have a large effect on persistent past contributions are. As we dive deeper into the question of balance of power it is important to note that being high lifetime cred and high current cred will not be uncorrelated, it is very likely that inner circle community members will be both. We might think of these as maintainers, there is an old brainstorming thread about that here: Maintainers User Stories discussion.
b. Current Cred Contributors whi are also Grain Holders. Assuming one is not also an old guard contrubutor this group represents an entrant to the community who is providing both capital and labor. Such a person is unlikely to be in need of payouts to cover their expenses and thus their contributions could be seen as magnified. Their labor will result and cred which will accumulate over time and eventually they may also become high lifetime cred. This could be increased by boosting work that they intend to do themselves (as @decentralion suggested above). Additionally, lacking a need to cash out to cover expenses, their level of grain may compound significantly faster than other new cred contributors. This to me feels a bit like joining a MMPORPG but buying your first character level ups and gear. In some ways this could be good, we’re all one team here, that contributor can have more impact faster and they put more on the line. However, we still want this power leveling mechanic to be muted enough that this accelerate onboarding is balanced with the other community members.
c. Old guard cred holders who are also grain holders. It is extremely likely that old guard cred constributors will also be grainholders. In fact even with the limited history of the credsperiment this is already visible: Week 7 Grain Distribution … Distribution for Weeks 9 & 10 . @wchargin is the canonical example of this group. One of the original core developers, he is responsible for no small way fort the existence of sourcecred as we know it today. Second in lifetimecred and Grain accumulated only to @decentralion. His week to week contributions are modest but valuable and come largely from code reviews and feedback thanks to his deep experience in the project. I think this role has reasonably balanced power today in ‘level one’ but i think it is important to note that many old guard contributors are likely to be grain holders.
d is for @decentralion: old guard, current top contributor and highest grain holder. This category today contains only one person, but it’s important to note that they are carefully distributing power. If @decentralion were an AGI, this would be a computer aided democracy. The benevolent dictator discussion has been constructive, but here I we should ask: in 5 years when 20 people are the old guard + core contributors + top ranked grain holders, do we have a plutocracy? What can the other 6 categories do if they don’t like the direction this group is bringing the project? Forking is of course the governance of last resort, but that’s also a form of breakdown because it represents unreconcilable differences within the community.
Within the discussion of balance of powers i propose we discuss power balance in terms of 7 categories (sorry to make it so messy), but then observe where the ‘mass’ if the distribution is likely to accumulate, and make sure the balance of power appropriately enfranchises all the subgroups in the partition. I mocked up a sample of the visuals I would find most useful for reasoning about the flow of power in sourcecred over time. Caveat: this has not been matched to our data, it is meant to demonstrate visually the dimensions of sourcecred stakeholdership. Choosing the metric for the partition would have a non-trivial impact on the qualitative interpretation of the results.
Aren’t the consumers of the signals the system generates also stake holders?
It seems as though 2.5 entity types is very low given the utility of the system. Maybe I’m just confused by delineating two groups solely by the time they enter the system, maybe just calling the old guard founders or initiators addresses one of my issues.
If we can monitor the consumption of the signals, this provides important feedback to help stabilize the system.
We certainly are not enumerating all possible interests or interest groups within sourcecred or within a community using sourcecred as a tool in their governance kit. We are currently enumerating the ‘the quantitive measures of stakeholdership’ as they exist in the current sourcecred designs.
I do think it is worth noting that lifetime cred versus current cred is not just about when people enter and more about understanding that cumulative past contribution is quite different socially and politically from being a heavy current contributor. It is primarily a difference of expectations (and real project specific expertise) in the direction of the current work.
i) it is extremely unlikely we can enumerate a reasonable share of all possible forms of utility and associated interest groups that might arise within sourcecred community alone, much less any instance of sourcecred in use by other communities
ii) by focusing on the measures of stakeholdship we can have a constructive discussion around how mechanisms may take stakeholdership into account to (dis)allow certain actions, or weight certain signals, and more generally explore how influence is exercised rather than the ends to which that influence is exercised.
It is my conjecture that allowing (even designing for) multiple categories of stakeholder to emerge and to exercise comparatively balanced power will lend this system to respectful discussion as opposed to polarization (2 stakeholder groups in rhetorical opposition stymying progress) or capture (1 stakeholder group dictating terms and imposing externalities on the minorities).
Mind you, this is all very experimental so, I am more than happy to hear thoughts expanding the purview of the balance of powers discussion.
This is a really insightful discussion, generating much thoughts.
This scenario seems the most likely, even if all players have good intentions. Imagine you’ve put in 5 years of hard work on SourceCred, put in your blood, sweat and tears, weathered growing pains, taken on financial and emotional risk, have scars to prove it, and have come out victorious. The community is thriving! Your financial investment has paid off, and you have the time and resources to continue innovating. You have big ideas. And those ideas are surely the best ideas, considering your position and experience. Well, yes, and no…
Imagine it’s 5 years from now, you’re a new contributor and have put in 5 months. You’re excited. You’ve finally found a system that aligns with your values and rewards your OSS contributions. You can say you’re changing the world for the better with a straight face . But, you soon start to see some issues…That’s fine. Not system is perfect. These issues aren’t insurmountable, and you have ideas for how to address them. You invest considerable (potentially unpaid) time analyzing the issue and proposing a solution in chats. You get…crickets. Or told you just don’t understand how the system works. Or that there simply aren’t resources for that. Those resources are needed by the ‘cred elite’ to implement their better ideas; in fact, you are one of those resources btw).
The truth is, that the ‘cred elite’ will have the same universal human limitations as any other group, namely:
Limited bandwidth: As the project grows, no individual person will have the cognitive capacity to have visibility into and understanding of all parts of the system.
Limited experience: If successful, the SourceCred of 5 years from now will be very different from what it is today, even if if SourceCred manages to scale its values. Cred elite will have not have experienced coming up in this new system. Just as it is impossible for me to truly understand what it’s like to be black, or female, it will be impossible for cred elite to understand the day-to-day reality of newcomers in the trenches. They will have a different set of problems, pain points.
Limited information: In addition to limits on the ability to process information (limited bandwidth), any power imbalance will keep those with less power from fully expressing themselves. Period. We have all experienced this in our personal and professional life. There is no avoiding it. Telling people, “don’t worry, we’ve built this totally enlightened system and there will be no negative consequences if you speak up”, will not work. Even if the cred elite are walking the walk, nearly all contributors will be coming from coercive hierarchical power structures, and will not take that at face value (and should they?). Compounding this, if SourceCred is indeed successful, its leaders will be subject to a constant barrage of illegitimate criticism. From competitors, to coordinated FUD campaigns, to bad faith concern trolling, sensational media narratives, etc. This will make filtering legitimate criticism from illegitimate difficult, both cognitively (see limited bandwidth) and emotionally.
Limited empathy: neuroscience has shown that power literally rewires your brain. Psychology experiments show giving people even temporary power makes them less empathetic. Do we believe that because our are intentions are pure (and I believe they are), that we are not susceptible to this?
I think everyone on this thread is aware of these limitations, but just felt like fleshing them out.
The only way, IMO, to keep this from happening is to give the ‘new guard’ (and other groups) enough power to speak their mind. Which is why I’m heartened to see this thread. If you are laboring in the system, you need hard sovereignty. The ability to meaningfully influence the system, and not worry about paying rent if you raise an inconvenient truth, even if it causes the cred elite cognitive dissonance and forces them to direct resources from their better ideas. If we can balance the powers well, this can hopefully be avoided in the first place, as the system will develop as a collaborative effort, with all parties feeling seen and heard (by actually being seen and heard, even if they lose some). Where power balances cannot be avoided (for instance, due to gaps in knowledge/experience (old guard will be better at deciding some things)), the “deal”, and any paths to more power need to be explicit and clear up-front, so that newcomers can make an informed decision before investing with their time and labor.
I will admit, I hesitated to even reply to this. Mainly because I’m currently living paycheck to paycheck and do need to sell some Grain every month (in addition to selling other crypto from work in another DAO) to make this financially feasible–as you can see from the Week 8 Grain Distribution, I’m the only one currently selling grain. Not because I fear speaking out (I’ve felt free to express myself since the beginning and we’re still in Level 1). But because I can’t afford to participate in governance for free. This work takes time, energy, and emotional labor. I decided to write this anyway because a) I may get cred->Grain for this post if people find it valuable, b) enough of my previous contributions were found valuable to, in aggregate, to generate “royalty payments” via the CredSperiment, de-risking any single contribution and c) if we get this right, this is a system I’d like to be in long term. The typical time for money dynamic is less present. Note that I realize the risks here and am OK financially if Grain payments stop for some reason.
It’s important to note also that I am an ‘interest group’ here, with my own inherent biases and limitations myself (I don’t pretend to understand a founder or VC’s day). Just wanted to add more data to the process.
I’m not sure how best to accomplish this, but impressed by the ideas and discussion here. @decentralion’s idea of giving different groups different ‘levers’ of power (e.g. cred and grain minting) feels promising…One property it has that I would like to see generally is simplicity. Too much complexity inexorably tilts power to those with the time and resources to exploit that complexity (i.e. those already in power). It also increases the educational barrier to entry for new participants.
@s_ben, I appreciate your decision to write the above post. As you’ve noted, there’s a limit to the human capacity to understand others perspectives. I think it’s worthwhile for each of us to do a personal perspective breakdown.
For my own part, I may be a bit of an oddity. I’ve never thought about influencing decisions in terms of explicit political power. I’ve predominantly relied on my ability to convince people who do have power to adopt policies I believe in. SourceCred is unlikely to be different. I’m far too financially constrained to have as much grain as an investor and my available attention is constrained by running my research firm. I’ll likely remain middle of the road at best in grain and cred. This hides the fact that I could use that attention share to retain the ears of those of you with decision making authority. In fact this is essentially what I am doing now and what I would hope for in the future. I justify that on the grounds that: I’ve done enough time as a SourceCred contributor to understand the algorithms and the community values, my non-SourceCred-specific attention is being applied to researching technology enabled social-political-economic structures, and the learnings there are relevant to SourceCred. That said I’m always leery of this trap:
The social network of influence will never exist totally in the observable data and it will always provide a perturbation to the perceived power structure. This is in some ways the role of academics in politics in the traditional world; stay apart, even theoretically (but not actually) disinterested and unbiased. The systemic effect of this dynamic is often academics/researchers lending legitimacy to the current power holders ignoring minority interest groups using “people just don’t understand as the experts do” rhetoric.
In my opinion, that power reinforcing dynamic is a symptom not a cause. In a healthy balance of power, one cannot simply cozy up to power brokers. In a community with many balanced stakeholder groups, the academic class produces signals but those signals are processed by the community into decisions. In fact there are also cases where research is used to bolster minority opinions, evidenced by the cases where Autocratic crackdowns often target the educated dissenters first. This is done precisely because research facilitates free thought and is capable of amplifying and lending legitimacy to a message regardless of its source.
As with everything in this thread, there are upsides and downsides. There will be no panacea but I do feel that it’s important to recognize the natural dynamics of decision making power, both measured and unmeasured in an effort to engender a balance of power.
What other kinds of entities? I think of SourceCred generating two signals right now: cred attribution (who did work) and the grain balances (who paid for it). So I started with just thinking of cred-earners and grain-holders.
But if you reason through the social implications, you realize that the “founders” and “new contributors” will have different perspective and interests (as @s_ben does a great job exploring), so I think we need to enfranchise them separately, to avoid the new contributors getting marginalized.
In the long run, as we flesh out an economy of interconnected projects, the downstream users will have influence over their dependencies, since a lot of the value of a project’s grain will relate to the upstream “homage” flows from its dependents. But I think this is an external influence on a project’s governance, but I don’t see why they should get modeled as a first-party participant.
@mzargham, thanks for fleshing out all the different intersections between these groups! The categories you enumerated (and the example instances) make a lot of sense to me. I’m also very entertained by the idea that we can model computer aided democracy if only I can transform into an AGI .
Glad to hear that you think building a multi-stakeholder model will make the system more robust and responsive.
@s_ben, I super appreciate your post; thanks for writing it! You’ve done a really good job of exploring how there will inevitably be tension between the “cred elite” (founding class) and newer contributors. Your point about power corrupting is well taken too (it’s something I’ve observed directly in my personal lived experience).
I think enfranchising the “new contributor” class (with explicit governance power) can do a lot to ensure that their voice is heard and that they feel empowered down the line.
On a meta note, I think there have been some really valuable contributions in this thread, and likes don’t let me express how valuable some of the posts are. I wish we had boosting implemented so that I could signal boost these posts explicitly.
Appreciate your openness and perspective. The project is certainly lucky to have you as a contributor. Numerous times now have found your contributions useful in exploring this new territory.
Have also been recently thinking about the role of academics and ‘public intellectuals’ in power structures. A podcast I’ve become obsessed with (Philosophize This!, forget the episode) explores this in depth. And nearly all frameworks analyzing social power structures recognize this group’s role in building consensus. So it’s good to have that represented as we explore these issues.
So far what I’ve taken away from this is, we’re not dealing with a level playing field.
Old-guard vs new-blood being a source of inequality.
How many resources you have are a type of inequality.
But to say these categories are a good predictor of the interests of different stakeholders is probably oversimplifying things.
So if categories are too complex to define, my inclination would be to look at broader principles that serve to level the playing field for individuals. Things like transparency, equal access to governance, equal access to speak with people who have more influence, having recourse if you couldn’t attend some meeting where a decision was made. In short, maximizing inclusion in spite of inequalities. I think we can find checks and balances to ensure this.
For example, with that in mind it would be a bad idea to have a minimum requirement of “skin in the game” to participate in decision making. As the financial risk might have prevented them from making a choice to have that skin in the game, making a chicken-or-egg situation for them, possibly leading to structural exclusion of these interests.
Yep, that seems like a reasonable claim at this point.
Considering interconnected projects as a weakly defined monolithic entity in the first-party system creates a better mapping to where the ultimate utility in the system will lie. But it makes sense that I was jumping the gun a bit trying to include that now.
There is also some point where most of the activity occurs in downstream projects, but again, it makes sense to worry about that years from now
I think it’s pretty interesting! A quick takeaway is that @wchargin and myself have just under 50% of the power. The next ~5 most engaged contributors have about 25% of the power, and the long tail of everyone else has the remaining 25% of power.
The “participation power” metric is indeed quite democratic, as just having a few recent forum posts is enough to get >1% of that power. (This is because we take the square root of the raw value, for power computation purposes–this tends to push power towards the long tail of small contributors.) Since it’s intended to be a counterbalance to the more concentrated cred and grain power, I think it’s working as intended.
Nice to see the cred ‘gini coefficient’ trending lower:) These categories and scores feel about right…I like that the participation metric gives voice to the long tail.
Could be useful for voting. I imagine we might see differing balances of power depending on what was being voted on and the turnout. A big, important decision such as Grain issuance would presumably turn out the Cred Power and Grain Power. Smaller decisions impacting the day-to-day of current contributors (e.g. how should we spend small amounts of money in an initiative, or a minor policy change) may turn out the Participation Power, giving them more leverage where it matters.
I might experiment with distributing some of the weekly Grain based on Participation Power. Right now the Grain flows concentrate really strongly with people who are the most committed (and thus get the most cred). It would be cool if you could accumulate meaningful Grain just by being consistently engaged over long periods of time.
Also, if we are going to enfranchise the “participant class” with explicit political power, it just makes sense that they would get explicit Grain flows too.
Starting next week, I will likely change the Grain payout so the 20% we’re already distributed based on “weekly cred” will get distributed based on “participation” instead.
Would have to not feed the trolls (they’re very engaged), but this might be a good catchall for contributions that fall through the cracks too, because they’re not part of an initiative or liked by those with lots of cred.
Like this idea as well. I feel like the 20% fast payout always had the goal of providing more inclusion opportunity. If this is a better formula to achieve that that seems objectively better to me. It might even be a reason to shift the balance to say a 30-40%, in line with the recruitment goals.
On top of that, how about we split this idea out into it’s own proposal topic and do a test-run of voting on it?
It’s definitely interesting. Some limitations also stand out to me.
@Protocol at 0.6%, seems like a measurement problem. Being mostly invisible in the value they’re adding having only a Grain stake.
@mzargham at 2.8% seems low in certain contexts. In part this shows the raw activity bias, but also hints at what this metric might be useful as. I would argue this is a proxy for skin in the game, not as a predictor of who’s likely able to make an informed vote.
For a question like: do we want to spend more resources on X? The skin in game measurement seems great.
For a question like: should we adopt this proposed new algorithm? I think it fails to recognize who could be domain experts.
Yeah, the fact that PL has zero cred is a bug and not a feature. I think the Retroactive Initialization will flow more cred to PL, for things like providing the operational support for the CredSperiment, helping found SourceCred, paying me to work on it.
Yeah, I expect he’ll also be a beneficiary of retroactive initialization.
I like this idea! I think I’ll first build a notebook that explores just the participation power concept, since it has some parameters (e.g. how fast participation power time-decays). Then in addition to voting on whether or not to tie the fast payouts to participation power, we can also vote on different parameter settings.
+1000 on this! Would be great to create a simplicity artifact (or maybe a usability artifact) to help make this happen. Keeping things clear and simple is essential for success
If the only way to get “skin in the game” is to buy it, then yes. If you can earn it, however, then it’s a different story. Reputation that is earned and cannot be bought addresses this.
My experience with this is that it’s mostly correct, except when it’s not. If any one whale has an opinion on something then the community just backs off. They know it’s a fight they can’t win. Over time this results in people generally just discussing things that the whales approve of. Even if the whales don’t vote, they shape the community landscape considerably.
One way to address this I think would be making people’s income “uncensorable”. Most people wanting a voice in a decision (and who have skin in the game) are either a) working for a traditional company and can be fired, or b) working for a crypto project with a centrally controlled dev that can defund their work. People generally aren’t going to risk their livelihood/community speaking out. Over time they just learn from the others around them and start self censoring according to the whale’s opinions. Reputation systems provide the opportunity to transcend the binaries of employment, and allow rewards to flow from the value a contributor provides to the project, as determined by the community as a whole, not just by the boss/whales. Even if a whale has enough power to sway votes their way, I believe that giving people the power to speak up and freely express themselves without fear of losing their income would be huge. That alone could make for a much better working experience and better decision-making over time, as good ideas won’t get killed on the social layer before they have a chance to make it to the governance layer (or at least inform the discussion).
Another important aspect I think is anonymous voting. Not sure where that comes into play here, but there’s a reason American’s voting system is anonymous. If someone can prove you voted a certain way, they can coerce you to vote a certain way. And will.
Another strategy is to “localize” votes. Whales by definition don’t tend to worry about small decisions. Their time and attention is limited. If there were votes, for instance, about small funding decisions within a bigger project, or minor policy change, they may be happy to abstain and let the “plebs” in the trenches with more knowledge about the subject determine the outcome, or be fine with being excluded by some mechanism.
Yet another strategy that’s been discussed are mechanisms like quadratic voting, which SourceCred could be in a good position to experiment with, as contributions over time are generally more Sibyl-resistant than other metrics for determining who the humans are.