AI 中文总结
研究卖家和买家委托不同代理时的个性化定价，解决相关子博弈，探讨信号与执行证明的影响，指出验证在不同情况下的作用及对交易和关系资本的影响，将AI测量等联系起来，不把信任等简单化处理。
AI 中文摘要
相同的公平规则，不同的后验概率，无交易。我们研究当卖家和买家委托给接收不同价值信号并执行机器强制指令的代理时的个性化定价。平等的名义剩余规则可能不兼容，因为每个规则都应用于其代理的后验概率。在一个常见环境中，我们解决嵌套定价、检查、指令选择和重复关系子博弈。信号证明和执行证明有不同影响：高卖家信号的证据可使高价合法化，而受限规则的证据可防止未经授权的提取。只有当验证涵盖有争议的命题时，才能恢复交易并软化双方的政策。在重复交易中，高于买家参考上限的可归因报价会降低关系资本，产生状态依赖的马尔可夫政策和尊重参考的区域。当卖家不将避免的买家检查和关系溢出内部化时，验证会私下供应不足，但当它促进提取时可能会供应过度。该模型将人工智能测量、算法提取和可验证约束联系起来，而不将信任视为软件情感或将验证执行视为真实价值的证明。
英文摘要
Same fairness rule, different posterior, no trade. We study personalized pricing when seller and buyer principals delegate to agents that receive different value signals and execute machine-enforced mandates. Equal nominal surplus rules can be incompatible because each is applied to its agent's posterior. In one common environment, we solve nested pricing, inspection, mandate-selection, and repeated-relationship subgames. Signal attestation and execution attestation have different effects: evidence of a high seller signal can legitimate a high price, whereas evidence of a restrained rule can prevent unauthorized extraction. Verification can recover trade and soften both principals' policies only when it covers the disputed proposition. In repeated transactions, attributable offers above a buyer's reference cap depreciate relationship capital, producing a state-dependent Markov policy and a reference-respecting region. Verification is privately underprovided when the seller does not internalize avoided buyer inspection and relationship spillovers, but can be overprovided when it facilitates extraction. The model links AI measurement, algorithmic extraction, and verifiable restraint without treating trust as software emotion or verified execution as proof of true value.
Comments67 pages, 4 figures, 1 table. Includes the main paper and online appendix
