Principles
Liquid democracy, in plain words
On any question you have a choice: vote yourself, or hand your vote to someone you trust to decide it for you. That's delegation, and it is per topic — keep your say on health, pass climate to an ecologist, pass budgets to an economist.
Delegation is transitive: whoever you trust may pass the bundle on to someone they trust, and so on, until it reaches whoever actually votes. And it is revocable — take your vote back any time, on any topic, and it returns to you at once.
Your voice on a topic is simply yourself plus everyone whose vote flows to you there. Influence is earned trust, not a seat — and it is fluid, hence “liquid.”
Background: Liquid democracy · Delegative democracy · Proxy voting.
How Republicate reads it
A few opinions baked into this implementation:
- Topics are a tree. Delegate a branch and its whole subtree follows — unless you carve a child back out. The keep-sentinel keeps “everything-but-this” a single gesture.
- Most-specific wins. A delegation on a child overrides one on its parent, so you can be broad-brush and precise at once.
- The river rule. Trust flows one way to the sea — no cycles.
- Take-back is sovereign. Your direct vote always overrides a delegation.
- The weight is public, the choice is private. Anyone can see who holds the say on a topic (the team revealer) — never who handed it to them.
The maths
The model, condensed (the full treatment is the spec's delegation-math annex).
- Topics form a tree — every topic but the root has exactly one parent.
- A delegation is a triple (a → b, t): dude a entrusts topic t to dude b. At most one is active per (a, t); the history is append-only.
- Effective delegation — for a on t, the active delegation whose topic is the most specific ancestor of t. Children override parents.
- Voice-bearer — follow effective delegations from a on t until the chain ends at someone who did not delegate: that terminal dude carries a's voice.
- Voice —
voice(a, t)= the number of dudes whose voice-bearer on t is a = 1 (yourself) + everyone who entrusted you, transitively. Conservation: voices over a topic sum to the population — no vote is created or lost. - River rule — the graph stays acyclic, so every chain terminates. A keep is a self-edge (a → a, t) that stops the cascade at t — the primitive behind take-back and interest zones.
Deeper: the Sources.
Sources
The liquid-democracy literature this rests on.
Foundational
Bryan Ford (2002), “Delegative Democracy.” The canonical reference — per-issue transitive delegation that preserves the right to vote directly. Our delegation math is Ford's, made concrete. bford.info/2002/05/15/deleg
Mathematical
Gölz, Kahng, Mackenzie & Procaccia (2018), “The Fluid Mechanics of Liquid Democracy.” The super-voter problem — delegations concentrating excessive power. arXiv:1808.01906
Escoffier, Gilbert & Pass-Lanneau (2018), “The Convergence of Iterative Delegations in Liquid Democracy.” When delegation patterns settle, and when they oscillate. arXiv:1809.04362
Kahng, Mackenzie & Procaccia (2018), “Liquid Democracy: An Algorithmic Perspective.” Local delegation with accuracy guarantees and no central guidance.
Aziz, Gaspers, Gudmundsson, Mackenzie, Mattei & Walsh (2014), “Computational Aspects of Multi-Winner Approval Voting.” Complexity of selecting multiple winners under approval voting.
Case studies
Hardt & Lopes (2015), “Google Votes: A Liquid Democracy Experiment on a Corporate Social Network.” The most concrete real-world deployment on record — UX, vote-visibility, churn.
Jabbusch (~2011), “Liquid Democracy in der Piratenpartei.” The German Pirate Party's LiquidFeedback deployment, analysed.