Decoding Human and AI Persuasion in National College Debate Through Aristotle’s Rhetorical Principles

• In Event: AERA Roundtable Session Sunday 7:45 am - Gold 4
  In Roundtable Session: Cognitive Processes, Human–Technology Interaction, and Learning Analytics

Abstract

This study explores how GPT can get involved in debate training in English Language Arts by comparing arguments generated by GPT-4o and human debaters. Debate, a key method for developing critical thinking, relies on selecting strong evidence and crafting persuasive claims with skills taught through labor-intensive human coaching. We prompted GPT-4o to generate evidence cards and evaluated them against those created by students using Aristotle’s rhetorical principles consisting of ethos, pathos, and logos. Through qualitative and quantitative analysis, we identified where GPT’s reasoning aligns with or diverges from human argumentation. Findings highlight GPT’s potential to enhance debate preparation while revealing areas where human debaters still outperform, offering insights for designing AI-assisted learning interventions.

Authors

• Mengqian Wu, McGill University

• Jiayi Zhang, University of Pennsylvania

• Raymond Zhang