Persuasion, Cooperation, and Deception in LLM-Based Agents

What it is about

You learn how AI agents based on large language models act, communicate and behave in social and strategic multi-agent environments. In the end you will be able to plan and conduct empirical studies, implement and analyze agents, and draw conclusions about factors such as prompting, interaction protocols and feedback.

What you will be able to do

• Formulate a research question about agent behavior
• Design and implement an experimental setup for LLM-based agents
• Apply computational social science methods to agent interactions
• Analyze and interpret empirical results
• Critically reflect on ethical, societal and governance aspects
• Present results in scientific written and oral form

What the module consists of

• SeminarTheoretical foundations, case studies and critical debates; guest lectures
• Practical exercisesHands-on training with code, implementation and data analysis
• Project workIndependent empirical study of LLM agents in a game setting

Teaching method

• Interactive seminarsConveying theory, discussion and contextual knowledge
• Practical exercisesImplementation of technical skills, coding and data work
• Group and individual workDevelopment of projects and strengthening technical/analytical abilities