强制披露下用自利LLM司机智能体评估网约车调度策略
原标题：When Disclosure Is Compulsory: Evaluating Ride-Hailing Rebalancing Policies with Self-Interested LLM Driver Agents
AI 摘要
该研究评估欧盟《平台工作指令》要求网约车平台以通俗语言披露自动决策规则后，司机会将规则交给LLM助手代为决策所带来的影响。作者用Claude Haiku 5.5和Jev两个LLM构建自利司机智能体，在三区域小模型和曼哈顿路网中对比反应式（R）与预测式（A）两类调度策略。结果显示规则披露后司机拒单率从约五分之四降至约四分之一，但R策略在曼哈顿使平台利润下降3.7%，A策略利润不变，两者最终相差6.1%；在完全合规情形下A策略实现司机收入、平台利润和服务率各提升约10–12%的三赢。
正文节选
When Disclosure Is Compulsory: Evaluating Ride-Hailing Rebalancing Policies with Self-Interested LLM Driver Agents Abstract The EU Platform Work Directive, which member states must transpose by December 2026, obliges ride-hailing platforms to state in plain language the main parameters of their automated decisions and how a worker’s behaviour affects them. Plain-language rules are exactly what a large language model (LLM) can read, so drivers will be able to hand the disclosed rules to an AI ass