Inquiring Agents Brings AI Ethics Debates to the Classroom

AgentLab releases Inquiring Agents, an interactive teaching tool that stages multi-agent Delphi debates on AI ethics. Built as a single Cloudflare Worker file with zero dependencies, it deploys in seconds and streams philosopher responses in real time.

Five Philosophical Traditions

Each agent embodies a distinct epistemological tradition from Churchman’s 1971 Design of Inquiring Systems:

• Leibnitz (Rationalist) — argues from first principles and logical axioms
• Locke (Empiricist) — grounds analysis in observable evidence and data
• Kant (Duty) — applies the categorical imperative and deontological ethics
• Hegel (Dialectical) — seeks synthesis through thesis-antithesis resolution
• Singer (Pragmatist) — evaluates consequences and real-world outcomes

How the Debate Works

The system follows a three-round Delphi method. In Round 0, all five philosophers respond in parallel to an AI ethics dilemma—topics like autonomous hiring, deepfake regulation, or algorithmic sentencing. In subsequent rounds, philosophers respond sequentially, referencing prior positions and sharpening their arguments.

Between rounds, students can inject their own perspectives, challenging philosopher assumptions or raising new dimensions. A coordinator agent then synthesizes all positions into a final policy recommendation, mapping areas of convergence and persistent disagreement.

Zero-Dependency Deployment

The entire application—HTML, CSS, JavaScript, and API proxy—is a single Cloudflare Worker file. There’s no build step, no package installation, and no infrastructure to manage. It deploys in seconds and runs at the edge globally.

Research Foundation

Inquiring Agents is adapted from Sridhar Nerur’s original CrewAI notebook, reimagined as a real-time web application with streaming responses and student participation.

Visit the Inquiring Agents project page for the full architecture breakdown.