题目：人工智能时代的有限理性

时间:：2026年6月18日（周四）上午10:00-12:00

地点：承泽园333

摘要：

卡内基学派将有限理性确立为组织判断的核心前提：决策者依赖对情境的简化表征，而表征不可避免地带来认知盲区。随着AI系统在信贷审批、战略筛选与风险评估中大规模替代专家判断，一个隐含的预期正在学界与实践中弥漫——机器或许能够超越人类的认知局限。本文认为，这一预期遮蔽了一个在组织理论中尚未得到严肃对待的问题：AI系统同样是有限理性的，只是其边界出现在不同的地方。本文以表征形成为切入点，运用形式分析模型，将人类专家与AI系统构建相关性的方式刻画为两种在来源上根本不同的机制：人类表征经由经验形成，由过去案例的后果塑造类别边界；机器表征经由数据学习形成，受训练信号的覆盖与分布所制约。这一差异不仅决定了各类表征与环境的契合程度，更决定了在案例空间的哪些区域判断可以被信任。表征的来源，即它如何被形成，是决定契合与否的根本变量。这一视角将卡内基传统中有限理性的研究前沿从人类认知局限延伸至机器表征局限，重新定位了AI替代人类判断的理论边界，并对组织在何时应当依赖机器、何时应当保留人类判断提供可检验的预测。

简介: Stephen X. Zhang (PhD, National University of Singapore) is Maness Chair of Entrepreneurship and innovation at Baylor University. His research examines how entrepreneurs and executives impact their businesses under uncertainty. His work appears in top entrepreneurship journals (Entrepreneurship Theory and Practice, Journal of Business Venturing) and management journals (Academy of Management Journal, Organization Science, Strategic Management Journal). He has secured over US$1.5 million in basic research grants and US$4 million in applied research funding across multiple countries. His academic career spans leading universities in Australia, South America, and Asia. Before entering academia, he worked in multiple industries and founded several startups.