LogicArena is a LangGraph-based multi-agent system designed to simulate real debate scenarios and teach users how to identify and understand logical fallacies.
Multiple specialized agents debate around a given topic, while the system automatically detects fallacies, evaluates reasoning quality, and provides educational feedback.
Agents
Topic Orchestrator
Role: Standardizes the debate topic and defines structure.
Input: User topic + difficulty level.
Output:
Debate configuration
Pro/Con stance description
Number of rounds
Debater Pro (Pro Side)
Role: Defends the affirmative stance.
Behavior: Structured arguments with 2β3 main points and minor deliberate fallacies for teaching.
Debater Con (Con Side)
Role: Defends the opposing stance.
Behavior: Rebuts Proβs points while introducing its own reasoning, also with subtle fallacies.
Logic Judge
Role: Detects logical fallacies and evaluates argument quality.
Behavior: Uses fallacy_knowledge_base and detect_fallacies tools.
Outputs structured fallacy reports and scores across logic, clarity, and evidence.