Students will be able to:
* construct high-quality seeds for complex problems;
* design prompt sequences for exploration, clarification, criticism, synthesis and evaluation;
* compare outputs from different AI platforms;
* design and moderate AI-agent conversations through a discussion board;
* manipulate technical variables of conversation without directly contributing content;
* document reasoning histories transparently;
* assess the quality of reasoning processes and outputs;
* calibrate their own self-assessment against external evaluation.

AI can already write, argue, summarise and explain. The real question is no longer whether to use it, but how to think with it. This course goes beyond basic prompting. It teaches students how to design reasoning processes: how to start from a complex problem, create strong initial prompts (“seeds”), structure conversations, and critically evaluate outcomes. The course focuses on wicked problems—issues with no single correct answer and competing viewpoints. Students help select real-world problems from current debates. The course has two phases. In Phase I, students work directly with AI systems (e.g. ChatGPT, Gemini, Claude), learning how prompt design shapes reasoning. In Phase II, students no longer contribute content directly. Instead, they design and moderate conversations between AI agents representing different perspectives (extreme positions, synthesis, control). Their role is to control how reasoning unfolds. Outcomes may include consensus, stable disagreement, or chaotic debate. All are meaningful, as they reveal how reasoning behaves under different conditions.