讲座主题：Agents and Social Intelligence
报告人：Andreas Herzig
讲座时间：2026年3月31日周二下午14:30-16:30
讲座地点：东校区工业中心802
报告摘要：Social intelligence is the ability to put oneself into somebody else's shoes: an agent possessing a so-called theory of mind can understand other agents' beliefs, goals, and emotions and take them into account appropriately in her own reasoning. It is well-known in psychology that theory of mind is of fundamental importance in human-human interaction. This indicates that artificial agents and other intelligent devices should also be equipped with such a theory in order to interact with humans and other artificial agents in a flexible and meaningful way. The question whether LLMs have a theory of mind has been much debated in the last years, fuelled by the Hanabi Challenge that was proposed by Google Deep Mind in 2020. It seems to be fair to say that the progress that was made did not meet the expectations and that some kind of formal methods are also required. Such a formal framework for theory of mind is offered by logics of knowledge and belief. The talk gives an introduction to these logics and their role in an account of social intelligence.
报告人简介：Andreas Herzig 博士，法国国家科学研究中心（CNRS）特级研究员（研究主任），图卢兹计算机科学研究所（IRIT）人工智能系原主任，法国国家科研署（ANR）人工智能与数据科学委员会主席，欧洲人工智能协会会士（EurAI Fellow）、亚太人工智能学会会士（AAIA Fellow）。其主要研究方向包括知识表示与推理、人工智能可解释性等。曾担任《Artificial Intelligence》等多个国际权威期刊主编及编委，在高水平期刊及中国计算机学会（CCF）A 类会议发表学术论文 100 余篇，多次应邀在 IJCAI 等人工智能领域顶级国际会议上作特邀报告，并主持欧盟重点项目等多项国家级、国际级科研项目。
