AI 中文总结
提出一种安全的流动民主机制，利用去中心化定时释放加密隐藏委托选择以防止从众和胁迫，并通过排名多委托和个人后备选票应对委托失败，实验表明密封形成能减少权力集中，排名多委托可降低投票损失。
AI 中文摘要
流动民主承诺通过允许选民直接投票、将投票权委托给信任的参与者，或通过后备机制结合两种方式，来改善集体决策。然而，现有的部署通常依赖于透明的委托，这使选民面临从众驱动的羊群效应，使胁迫可验证，并在高支持度的委托弃权时引入系统性脆弱性。在本文中，我们提出了一种安全的流动民主机制，解决了知情专业路由与系统鲁棒性之间的张力。我们引入了一种使用去中心化定时释放加密的密封委托机制，在形成阶段加密隐藏委托选择以防止从众和胁迫，同时为最终计票恢复完全公开的可审计性。为了解决委托失败问题，我们扩展了协议，加入了排名多委托和个人后备选票。我们正式证明了协议的预揭示保密性和重新提交收据无关性。最后，我们在四个真实数据集上评估了该机制：一个带有校准调查的市政参与式预算选举、另外二十个参与式预算选举，以及带有客观能力度量的6万名美国选民。结果显示，委托是否提高代表性准确性遵循一个可恢复差距定律；只有当弃权规模大且系统性不具代表性时，委托才有帮助，否则是中性的或有害的，其中代表性风格的委托比委托给能力精英更安全。密封形成的好处主要是结构性的，显著减少权力集中而非直接提高准确性；而带有个人后备选票的排名多委托在现实和针对性的委托失败下显著减少投票损失，这一结果在全部二十个选举中重复出现。
英文摘要
Liquid democracy lets voters either vote directly or delegate their voting power to a trusted participant. Existing deployments make delegations publicly visible as they form, which invites popularity-driven herding, makes coercion verifiable, and leaves the election fragile when highly backed delegates abstain. We propose a liquid democracy mechanism that removes these vulnerabilities while keeping the tally fully auditable. Delegation choices are sealed with decentralized timed-release encryption while they are being formed and revealed in full for tallying, and each voter names ranked backup delegates together with a personal fallback ballot so that delegate failures do not silence them. We prove pre-reveal secrecy and resubmission receipt-freeness for the formation phase. Experiments grounded in real voting data show that the mechanism keeps voting power dispersed where transparent formation concentrates it into few hands, keeps outcomes stable under herding where transparent outcomes become a lottery on arrival order, and cuts vote loss under targeted delegate failures from 26% to about 3%. The experiments further characterize when delegation improves representational accuracy, namely when abstention is large, unrepresentative, and widely converted into delegation, giving deployments a concrete condition to assess.
