作者：刘清堂，秦丽娜*，叶蕊秋，左云帆

出版刊物：Global Chinese Conference on Computers in Education(GCCCE2026)

出版时间：2026年

内容摘要：

AI debate agents show promise for cultivating critical thinking through dynamic dialogue, yet the influence of their conversational roles remains underexplored. This randomized controlled trial investigated how three distinct roles—supporter, arguer, and comprehensive debater—affect learners’ critical thinking. Results showed that all three roles similarly enhanced self-perceived critical thinking but did not improve the quality of written arguments. These findings indicate that a single interaction with an AI debate agent can boost learners’ confidence in their thinking skills, regardless of the agent’s role. However, fostering long-term argumentation skills requires structured, iterative feedback mechanisms that address the gap between self-perception and actual ability.

AI辩论智能体通过动态对话展现出培养批判性思维的潜力，但其对话角色的影响仍未得到充分探究。本项随机对照试验考察了三种不同角色——支持者、论证者和综合辩论者——如何影响学习者的批判性思维。结果表明，三种角色均同样提升了自我感知的批判性思维，但并未提高书面论证的质量。这些发现表明，无论智能体扮演何种角色，与AI辩论智能体的一次交互即可增强学习者对自身思维能力的信心。然而，培养长期的论证能力需要结构化的、迭代式的反馈机制，以弥合自我感知与实际能力之间的差距。