From ad-hoc voting to continuous voting
“There are only two hard things in Computer Science: cache invalidation and naming things.”
Conviction Voting is a coordination module that allows communities governing a shared resource to choose priorities in a decentralized way. This module is a part of The Commons Stack — a blockchain-based software initiative that enables communities to build digital economies around a cause, called Commons, that can interface with the real world.
Ad-hoc Voting vs Conviction Voting
The most common solution to coordinate action in current systems is to give all individual agents the ability to vote on a proposed solution. The most popular scheme currently used is to collect all the proposals together and hold a vote — to call the beast by its name: the majority vote. This scheme mostly has the following properties:
- One vote per individual (agent)
- Every vote counts as one, all votes are considered equal
- Once cast, a vote cannot be changed
- Voting is anonymized
- A set timeframe exists when voting is possible and the votes are counted at the end
- Consensus is sought by defining the rules for a successful vote (participation is measured but usually there is no quorum required, the outcome is triggered by achieving a relative majority for a proposal or candidate)
While these properties are not bad per se, they come with some undesirable side effects. Some examples:
- Anonymizing votes makes detecting systemic fraud on behalf of the organizing body hard to detect and only leaves the solution to hold the vote again.
- It is usually required that voters are physically present at a certain point in time which poses a high level of friction towards organizing the vote, as well as participating as a voter.
- Having a set date to vote enables strategizing on shaping the outcome via selective broadcasting of information and enables a whole meta-game that is being played actively and gave birth to some industries that were not always part of the process (e.g: see the history of Campaign Advertising, the Cambridge Analytica scandal, etc.) but now are viewed as integral.
- Not being able to change your vote means that once all votes are cast, accountability cannot be enforced by the voters. This is generally exploited in political voting, where it can be assumed that what was promised is not necessarily shaped into policy.
The concept of voting continuously, which we apply in Conviction Voting, became possible with the advent of cryptoeconomics and token engineering. Conviction Voting got its name from one distinct property: Your vote accumulates more power over time, as long as you don’t change your stance. The first concept was called ‘social sensor fusion’, as we deal with the fusion of different signals. The context for which we developed this module is a larger system, in our specific case a Commons. These Commons consist of an Augmented Bonding Curve that continuously funds a Conviction Voting DAO that governs a pool of funds designed to fund projects on top of the Giveth DApp.
Now let’s look at some of the properties of Conviction Voting:
- Voting is continuous, its much more like signaling.
- The emitter can change their “vote” anytime
- A fundamental property of this particular system is that votes grow stronger over time (as long as the vote is not changed).
- Because the focus lies on the signage of proposals, the choice is binary (only “support” or “no support” is possible, the default is “no support”). Ternary choice is an optional extension to the mechanics, to add an “active resistance” by enabling token holders to “support against”.
- To reduce cognitive overhead, “support” could mean full support (signal with the full amount of tokens or, in the case of multiple proposals supported, an equally divided amount of tokens from the full holdings of the agent).
- A positive outcome (signing off) is achieved by reaching a trigger value that is defined to be proportional to the sum of all funds (and therefore tokens) and the requested funds of the proposal.
- Proposals have an “upper ceiling” in asking funds that is proportional to the total holdings of the commons.
- A larger holding of tokens always means stronger supporting weight.
- The underlying architecture uses the Ethereum blockchain or EVM sidechain like xDai to achieve a trustless environment that is pseudonymous but fully transparent to all parties internal and external.
- The system itself does not discriminate against artificial intelligence and autonomous software, they are “equal” to human actors. In the end, every signaling agent is an address on the Ethereum blockchain.
These differences between ad-hoc voting and Conviction Voting lead us to assumptions regarding properties that we deem positive in its context. A continuous voting mechanism is more agile and reduces friction as well as cognitive overhead for the agent, due to less planning to cast the vote.
I am not arguing for a replacement of (political) voting, but offering a complementary mechanism that concerns itself initially with decisions in the economic realm, and wants to formulate a tool optimized for collectively signing off individual proposals. It is important to acknowledge here that from an ethical standpoint the idea of replacing anonymous ad-hoc voting with pseudonymous continuous voting might violate some of the most important tenants of a functional democracy.
Our assumption is that Conviction Voting is better suited for allocating community funding by being more agile and secure, as well as less biased than modern and historic forms of voting.
Voting and the Blockchain
The blockchain space is no exception to the dynamics of manipulation. We can observe the latest Aragon token holder vote. The vote has the function to govern the “Aragon Network” through proposals (called AGPs). In the second network vote, a highly contentious proposal was decided at the last moment, by a large token holder coming in and flipping the vote:
Michael Zargham, creator of the social sensor fusion concept, comments with formulating the necessity for new schemes that allow decision-making in decentralized groups that are designed with modern game theory in mind.
It should be noted at this point that we don’t argue for strict superiority of our proposed model. We can merely try to improve the blind spots we find in a system and remove or work around possibilities for manipulation. In that sense the “right” coordinated-decision-protocol needs context: “Who is deciding what under what conditions?” — it is important that this design space is explored in the presence of new and disruptive information technology.
Currently there are a bunch of experimental projects in progress to do fundamental design and programming work. These different implementations might be scrapped or mutate along the way:
- Gitfyah strives for simplicity and enables a DAO to use their governance tokens to decide on funding GitHub issues
- A release version as a module of the Commons Stack. It enables token holders of a DAO to use conviction voting as a pluggable coordination mechanism — ideally among other pluggable mechanisms that help in decision-making
- A decision module built on top of Giveth.io, allowing a decentralized community to decide on funding a Milestone in a charitable crowdfunding campaign
More on implementation possibilities:
- Griff Green’s awesome presentation on the commons stack (ETHCC Paris 2019)
- Giveth exploring the commons
- The Commons Stack initiative and its Augmented Bonding Curve.
Fundamentals of Conviction Voting
Signaling is employed by the entity we love to copy from shamelessly: Nature. Ecosystems balance themselves because of carefully weighted signals which are constantly emitted by all agents to let the larger system adapt to external and internal changes. Some great examples we can observe are streams of nutrients in forests, communication of insects, electrical current in the bodies of animals and humans, etc.
The theory and intention, as well as the mathematical engineering behind our proposed Conviction Voting solution was built upon years of reasearch by Michael Zargham, with some tweaks and additions coming from recent experimentation and adaptation to actual implementation on an EVM. Credit for these recent experiments goes to the hacking teams participating in ETHParis and Odyssey hackathons. Some of the challenges we faced (right down to calling it under a different name) are explained in a video interview shot at Odyssey hackathon.
A physicist might call this section old news, especially one who dealt with electrical capacitors. However, a mathematical explanation is the most concrete one can give to leave the least amount of confusion with the inclined reader:
- The original proposal by Michael Zargham
- Formal mathematical justification by Michael Zargham
- More digestible breakdown of math by ETHParis Giveth team
Into a Future of Expressive, Continuous Voting
Collaborative decision-making is an exciting field, where the first beautifully radical concepts are starting to gain traction, like quadratic voting. To reach qualified decisions collaboratively, we need a lot of input and that input has to be broadcast, reflected and again consumed by us in a cycle of learning. A vote has to be trusted, secured and embedded in a system that makes sense to the voter and on top of that is agile enough to deal with the challenges of an ever-changing, interconnected planet.
Internet, decentralized computing and the blockchain give us the ability to explore new ways of human expression in computerized systems that are so important in a time where people sometimes feel overwhelmed by technology and by the bad reputation big data scandals bring. The individual fears that their voice is just not being heard. But transparency and accountability, two key values of Giveth, can bring back the trust that was lost. The blockchain world is steadily becoming more helpful to the individual or community who wishes to achieve higher degrees of self-sovereignty and representation. The revolution will be open source.
Sampling of research used by M. Zargham
- Collected works of Reza Olfati-Saber (Google scholar)
- Evolutionary Dynamics of Behavior in Social Networks by Reza Olfati-Saber
The Commons Stack
- Infrastructure for Scalable Community Collaboration
- Crowdsourcing the Commons
- Deep Dive in Augmented Bonding Curves