讲座题目 | Strategic Reasoning Under Uncertainty: Insights from Game Theory, Human Experiments, and Large Language Models | ||
主讲人 （单位） | 庄峻 （布法罗大学） | 主持人 （单位） | 王文平 （东南大学） |
讲座时间 | 2026.07.28 10:00am开始 | 讲座地点 | 文科楼914 |
主讲人简介 |
庄峻博士现任纽约州立大学（SUNY）布法罗大学（University at Buffalo, UB）工程与应用科学学院（SEAS）科研副院长（Associate Dean for Research）、工业与系统工程系（ISE）Morton C. Frank 讲席教授，并担任国际运筹与管理科学学会（INFORMS）期刊《Decision Analysis》主编（Editor-in-Chief）。他于 2008 年获得美国威斯康星大学麦迪逊分校工业工程博士学位。 庄博士是美国科学促进会（AAAS）会士、工业与系统工程师学会（IISE）会士以及风险分析学会（SRA）会士。他的主要研究方向是融合运筹学、大数据分析、博弈论和决策分析，以提升自然灾害和人为灾害的减灾、防备、响应与恢复能力。其研究还广泛应用于医疗健康、体育分析、交通运输、供应链管理、可持续发展和建筑设计等领域。 | ||
讲座内容摘要 | Strategic decisions often involve uncertainty, incomplete information, and potential deception. This talk examines how intelligent agents reason in such environments, progressing from game-theoretic models to human behavioral experiments and, more recently, large language models (LLMs). Using attacker–defender signaling games, we show how information disclosure shapes adversarial beliefs and target-selection decisions, and how human behavior systematically departs from normative predictions. We then compare Bayesian benchmarks, human decisions, and GPT-4o under identical conditions. The results reveal both alignment and important divergences, including evidence of belief–action decoupling in LLMs. Together, these studies provide insights into strategic reasoning, adversarial decision making, and the capabilities and limitations of AI in uncertain and deceptive environments. | ||
