打破动态博弈中的惯性：动态干预、实施与结构设计
Breaking Status-Quo Inertia in Living Temporal Games: Dynamic Intervention, Implementation, and Structural Design
AI总结：
本文研究如何通过动态干预克服动态博弈中的惯性问题，提出了三种干预类型：有界转移、结构修改和信息信号，并证明了惯性深度的概念和阈值定理，展示了结构主导结果以及信息机制的效率与预算平衡。
摘要：
我们研究规划者如何设计动态干预以克服动态博弈中的惯性问题，其中战略代理在时间网络上控制其状态（活跃、睡眠、部分死亡）。基于我们同行论文中的连续时间随机博弈框架，我们引入了三种干预类型：有界转移（基于价格）、结构修改（边删除、添加或替换）和信息信号。我们正式化了惯性深度的概念，并证明了一个阈值定理：惯性均衡在所有转移扰动的幅度低于临界界限时都存活。一个核心的结构主导结果表明，对于任何有限的转移预算，存在一组游戏，其中无界价格干预无法消除低效均衡，但单个边替换（连续流到离散运输）却能成功。然后我们研究了具有静态类型的私有信息子类。利用统一化约简，我们证明了一个不可能结果：没有直接机制可以同时满足事后激励相容、事后预算平衡和历史隐私，同时始终实施高效均衡。在同一子类中，我们构建了一个动态pivot机制，实现了次优效率并具有有界赤字。最后，我们证明将连续流边替换为离散运输边会弱化可实现的结果集，突显了时间语义在机制设计中的重要性。我们的结果将[5]的静态分析扩展到连续时间战略网络，并为后续关于学习和均场设计的论文提供了严谨的基础。
英文摘要：
Westudy how a planner can design dynamic interventions to overcome status-quo inertia in living temporal games, where strategic agents control their state (active, sleep, partially dead) on a temporal network. Building on the continuous-time stochastic game framework of our companion paper, we introduce three intervention classes: bounded transfers (price based), structural modifications (edge deletion, addition, or replacement), and information signals. We formalize the notion of inertia depth and prove a threshold theorem: the status quo equilibrium survives all transfer perturbations whose magnitude is below a critical bound that depends on the remaining horizon. A central structural dominance result shows that for any finite transfer budget there exists a family of games where no bounded price intervention can eliminate the inefficient equilibrium, yet a single edge replacement (continuous-flow to discrete-transport) succeeds. We then study private-information subclasses with static types. Using a uniformization reduction, we prove an impossibility result: no direct mechanism can simultaneously satisfy ex post incentive compatibility, ex post budget balance, and history privacy while always implementing an efficient equilibrium. In the same subclass we construct a dynamic pivot mechanism that achieves second-best efficiency with bounded deficit. Finally, we show that replacing continuous-flow edges by discrete-transport edges weakly expands the set of implementable outcomes, highlighting the importance of temporal semantics for mechanism design. Our results extend the static analysis of [5] to continuous time strategic networks and provide a rigorous foundation for subsequent papers on learning and mean-field design.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
