A Synthesis Toward Equitable Group Decision Making
Special thanks to ItamarGo and SaulThorin for feedback and comments.
The first part of this article describes a system for automating the calculation of reputation, via contribution scores, for each contributor in a DAO.
A common trope circulating through the memesphere is that DAOs are inefficient and only applicable to particular situations. When is decentralization the correct solution, and how much decentralization is optimal?
The so-called “DAO trilemma” consists of scalability, quality (security), and access (decentralization) of which we can only have 2/3 and must compromise the third. We can also think of this in terms of decentralization, organization, and automation.
What’s the real trilemma? Maybe I’ll explore that in another article.
We simply haven’t gotten good at automation and organization yet. We only just recently grew our collective consciousness to a critical mass of awareness about Moloch in the past couple of years.
Here’s my personal favorite and highly eloquent generalization about trade-offs:
Coordination challenges abound, yet, our rate of progress is ever accelerating at an ever-increasing rate. It’s the typical compounding “large number” problem that we as humans have faced so many times in our exploration of science and the universe. We haven’t tried enough to say one way or the other, however, existing progress seems pretty inspiring.
Back to the trilemma question. How do we choose whether it’s optimal, let alone appropriate, to pursue decentralization as the correct solution to a problem? When considering this in the framework of game theory in the context of rapidly changing technology, we have to incorporate how the decision landscape might change over time. Can the existing solution set for a problem evolve such that the landscape shifts from a convex worldview to a concave one?
As technology improves then applying moderate amounts of decentralization may become preferred over centralization. In the generalized scenario shown above, the maximum point on the curve will gradually increase in value along the Y-axis over time in any worldview. However, in a convex worldview, the option with a poorer outcome could be researched and developed more rapidly to eventually yield a better outcome.
Given the trajectory of technological advances, it seems reasonable to assume that many of the decisions we face in implementing decentralization today could be a product of our current stage of infancy. One can imagine the shape of a decision curve can potentially flip from a convex worldview to concave given sufficient innovation.
I try to remind myself of our ignorance on a regular basis. Ignorance is bliss, after all. It helps to keep me happy and a shameless optimist when considering our ability to solve problems collectively as a species and approach problems with an abundance mindset. We can only have our cake and eat it if we first believe we can. Sorry, Ron.
There are numerous ways that quantifying contributions can be integrated into the voting process of an organization. Below is a list of common voting types and how they can be modified to integrate contribution scores.
Contribution scores can be integrated when evaluating the merit of proposal authors. They could even be used to gate permission for submitting proposals in either manually approved or in an optimistic/consent based approval system.
Weighted (linear and quadratic)
Tokens become one of many factors in the model, instead of the only one, creating a contribution score weighted voting system.
Can be combined with the concept of weighting to integrate contribution scores while considering overall order of preferences.
Contribution scores can be use instead of allocating tokens. Voters delegate or “stake” their allocated voting power (determined by their relative contribution scores) on their preferred choice(s). This is distinct from applying the concept of conviction as a time filter when calculating contribution scores as described in part 1 (which also includes a link to the reference on conviction voting).
The limitations of coin voting have been discussed extensively and conceptual approaches to overcoming those issues include: 1) limited governance, 2) non-coin driven governance, and 3) skin in the game. Contribution scores utilize all three of these approaches to some extent. They limit governance with a highly customizable weighting algorithm that includes time delays and forkable components. As described above, contribution scores provide a variety of ways to proportion the effect of token holdings appropriately relative to other contributions and provide sybil resistance through a maximally pluralist approach to quantifying reputation. Lastly, there is a lot of skin in the game when it comes to reputation. The property of sybil resistance confers value which keeps those who use the system accountable because they have reputation on the line.
A fractal-like governance system can support a flexible structure that encourages cooperativity among groups that wish to diverge. Members accumulate opportunities as they build up their contribution score through a variety of actions. This is a positive sum system that builds upon community interactions. We can achieve community value alignment through combination and normalization of scores between categories (e.g. category weights/ranks).
The modularity of this model allows subjective catering of contribution inputs that communities seem relevant. It also allows for fractal structures, with diverse topologies to emerge within the DAO where smaller groups have full customization of their internal governance - which pulls social graph data from the same sources as any other subgroup, and the umbrella DAO itself.
Contribution scores can be applied to any quantitative voting type.
Contribution scores are Sybil resistant due to the broad variety of inputs. The combined modularity of social subgroups in a contribution score model makes it possible to model the network topology in a way that flags automated fraud detection of potential Sybil attacks.
Contribution scores have net positive effect on the accuracy of decision making by increasing the sensitivity and specificity of voting power.
Increased Sensitivity: By enabling groups of people to operate independently under shared conditions we incentivize groups to form of optimal size. (increased sensitivity).
Increased Specificity: Optimally sized groups can optimally choose which contributions to track, quantify, categorize and weight.
Lastly, inter-community functions can be used to bridge voting weights between any number of members in each community, for any amount of time/votes.
Zooming out, it's a simple overall concept. But, the underlying modularity and potential for complexity allow for maximizing self-sovereignty while preserving the social contract members consent to when first joining the umbrella/main DAO. A diverse mix of incentive structures including bounties, tips, peer-reward circles, algorithmic analysis (Contribution Scores), grant programs, and others can improve accessibility to funding and support limited governance through a balanced combination of permissioned and permissionless systems.
It is becoming increasingly easy to create new communities with governance and incentive structures designed to align with an existing community. Seeking alignment ensures that both the original group and the new group are able to maximize their efforts by sharing resources and collaborating in a variety of ways. A few notable examples are AstroDAO, ADAM Vault, Colony, Govrn, and Metropolis (Orca Protocol).
Much of this has been said before.
Due the the flexibility of the integrated approach to governance described above, it could be possible to extend this theory of decision making to any organization. However, there still exist subjective implementation decisions, such as classifying contribution types.
The calculation of contribution scores can most obviously be skewed through the choices made to either include or exclude various input metrics. The ability to quantify an ideal set of metrics in a transparent way may also requires various levels of technical skill ranging from accessing a pre-configured dashboard to building out entire businesses and new platforms.
Similar to the choice of metrics, the incentives that exist outside of the direct translation to contribution score (e.g. token rewards for bounties, etc.) will have a direct impact on which metrics effect the final contribution score the most. Ownership is a large incentive and an underlying principle driving the formation of new networks in web3.
Exploring the design space of DAOs using objective statistical modeling to identify key factors which truly represent what a DAO is will allow us to build systems focused on value creation instead of value extraction. Value flow, where does it lie?
The proposed system of calculating contribution scores requires a flexible iterative approach with numerous points where design decisions need to be made. Assessment of which metrics are working, how incentives impact the system, and implementing changes are all points which could be up for debate.
Typical governance forums on Discourse and other platforms can offer metrics to gain insight into their performance and community participation. The system itself should constantly be scrutinized using statistical modeling and testing. Implementing this should be done in a protocol-like approach with a core functional platform being delivered from the start capable of sustaining itself over time and being designed to resist failure.
Clear boundaries make it easier for contributors to fully occupy the space in the community which is available to them.
Are we all doomed? Not in the mere sense of failure. Rather, the complete existential demise of the human species. Why? How? One word: “robots”.
Does the entire history of biological life merely represent an arc along the evolution of technology? Do we unwittingly serve the machines while deluding ourselves of what we as a species collectively spend countless lifetimes pursuing?
Umm… That got dark quickly…
I take the phrase “spending time” quite literally. I subscribe to the theory that economics are subject to the constraints of thermodynamics. The world of things (contributions) has value because of scarcity, they aren’t easily obtained. Someone ultimately has to transmute energy into matter to make a thing, yet, the sum total amount of energy in the system cannot be created nor destroyed. At least not in the physical world and an economy based on the production of real assets.
The answer to whether thermodynamic constraints apply to digital scarcity seems to emphatically be “no”. This is the core principle of regenerative finance, that we can use economics to not just produce value, but to reduce the sum of entropy in the physical world. An economy based on the physical world is a human construct that extracts energy and converts it into financial value. An economy based in the digital world acts as a sink where endless value can be created and then transferred back to the physical world to create a net positive-sum outcome.
Money is the energy in the system. Money is time. Time is money. Bitcoin is money. Bitcoin is time.
Making decisions is expensive. Even more so with procrastination. An investment of time and money (across any time-frame) is pretty much a basic requirement to achieve efficiency for any sufficiently complex system. My hypothesis is that a system such as contribution scores can be general enough to find broad application and be communicated in a simple way, yet complex enough under the hood to facilitate highly detailed customization.
Managing an algorithmic governance system offloads the majority of the labor burden to a small number of people. This can be accomplished by delegating votes to respected members of a community to act on behalf of others and are incentivized to do so. With an ability to recall delegates at any time, the total governance burden in a population can be minimized.
Hardly anything anymore is truly novel. Most of what I’m saying here is repetition. Big data is everywhere. The only truly novel ideas remaining are bold paradigm shifts that require strong willed early champions. I probably could have included 10x as many references in this article if I was less ignorant. People don’t need to add a disclaimer to everything they say, however, society seems to be going through some contortions over how to define ownership, particularly when it comes to intellectual property.
Our future will be governed by AI, which makes the power of small groups substantially more threatening. It raises the possibility that we’ll be reduced to a set of numbers even further than we are already. The ideas I present here could easily be utilized toward dystopian ends that berate humanist ideals.
What’s stopping this? Plurality. Natural evolution. Variation. Diversity. Equity. Inclusion.
Oh, and, zero-knowledge proofs.
Searching the design space of contribution types and data transformation methods to identify local maxima in the landscape of effect sizes for each independent variable.
There are many methods available in the realm of multivariate statistics that can provide information regarding the effect size of independent variables as well as how to reduce the dimensionality of a large dataset (linearly correlated variables, factor analysis, dimension reduction, multidimensional scaling).
Data is a double-edged sword. It can illuminate our bias or it can be used to gain advantage over those we disagree with.
However, science is inherently a social activity. This essentially makes the scientific method a recursive moral arbiter for how we choose to use the data we generate from applying it. Scientific data requires review and acceptance by other scientists (or at least other experiments) before it can be considered trustworthy.
There are examples of existing organizational design paradigms (e.g. structural adaptation theory) that can be leveraged to generate hypotheses, and testable predictions about the systems we’re building in web3.
The ideas presented here, just as with any other governance model, are just that.
A model is an incomplete descriptions of a system representing some aspect of the real world.
Modeling interactions between contributors is an iterative process where at each step there is some redefining of nodes and edges in an evolving social graph.
Understanding DAOs is a big data problem in the form of an observational research study. No matter how much we plan and build. It will ultimately come down to seeing what works and course correct from there. Many structures are expected, but, for me the fun part of complex networks are the emergent properties. By definition, we can’t predict those or be prepared for them. We just have to figure it out. Mess and all.
Similarly to how an individuals contributions to an organization can be enumerated and quantified, the reputation of an organization itself and it’s contribution to a larger ecosystem can also be evaluated. Internal metrics can also be used to evaluate the health of a community and the groups capacity to contribute overall.
The coordination of groups amongst each other is a highly reductionist, but acceptable, definition of meta-governance. Whether it be parliaments of representatives or tokens; there needs to be a third party, a lingua franca, that enables organizations to coordinate their decision making.
Contribution scores can be abstracted one level further to complement mechanisms used for token-based governance to more holistically represent organizations in their efforts to communicate and align their efforts.
Thank you for making it this far. I’ve explored a litany of pros and cons associated with algorithmic reputation scoring. There is still much more that can be said. I hope to continue this line of reasoning in discussion with others working on similar ideas.
Without going into detail, I’ll simply list a few other applications of algorithmic reputation scoring below:
1. Impact factors in decentralized science (DeSci).
2. Using a data marketplace to create a base layer of revenue for the organization and it’s members. A two-sided “UBI” for participation, per se.
This idea needs token engineers, DID/reputation experts, data scientists and governance experts to assess mechanics of the model (e.g. what parameters have the biggest impact?).
Since this is my first article on my personal Mirror blog, I want to provide some context. I’ve been reasoning through the general topic of how to quantify contributions in an organization for the past ~6 years. The journey to writing this article began with starting my first real business venture nd attempting to implement the “Slicing Pie” method to determine our co-founder equity split.
Around that time I also began exploring applications of blockchain in supply chain management, but never really pursued anything in crypto beyond occasional small value trading. I had also mused about why we weren’t doing digital voting yet when the 2020 U.S. Census (used to determine voting allocations) was being conducted online, but again, didn’t pursue any of those ideas at the time.
At the start of 2021 I began to actively invest the meager income I had in crypto and spent every night learning about the market, who was releasing news, and eventually who were the actual people building amazing things. Within a few months I began exploring how to DAOify my existing business which led to founding a spin-off that year and going full-time web3 in 2022. This article summarizes the observations and conclusions I’ve drawn along that arc and started as an excerpt from the manuscript I’ve been writing for Cannabis Genome DAO.
And, one last little tidbit about digital voting…
If you find value in my work, please consider collecting this article and subscribing.