人类导师与智能导师效能差异研究——基于动态交互的可供性视角  

Differences of Efficacy between GAI Tutor and Human Tutor:An Dynamic Interaction Perspective Based on Affordance Theory

作　　者：李姗姗 吴筱萌[1] 张鹏 范逸洲 LI Shanshan;WU Xiaomeng;ZHANG Peng;FAN Yizhou(Graduate School of Education,Peking University,Beijing 100871,China)

机构地区：[1]北京大学教育学院,北京100871

出　　处：《开放教育研究》2025年第2期98-107,共10页Open Education Research

摘　　要：生成式人工智能技术的应用推动了学界对智能导师教育功能边界和局限的研究。本研究基于可供性的动态交互视角,以在线英语写作修改任务为例,采用主题分析和滞后序列分析方法对54名学习者与智能导师和人类导师的在线交互日志进行研究,剖析智能导师和人类导师的效能差异及其成因。研究结果表明,人类导师组存在“主动寒暄—自我报告”“按步指导—评估更新”“基于评分准则的请求”三类典型可供性交互模式,智能导师组则涌现“内容代理循环”“基本纠错循环”和“不明指向的对话”三种模式。两组可供性交互模式的差异主要源于学习者与两类导师之间关系及权力位阶的不同。因此,研究建议智能导师设计需强化社交与教学法可供性以缩小与人类导师差距,同时在技术受限背景下需整合双方优势探索人机协同教学新模态。Generative Artificial Intelligence(GAI)tutors have spurred academic inquiry into its functional boundaries and limitations in educational contexts.Following the dynamic interaction perspective of affordance theory,this study examines online English writing revision tasks,analyzing interaction logs from 54 participants engaged with either GAI tutors or human tutors through thematic analysis and lag sequential analysis.The study reveals three distinct affordance interaction patterns in the human tutor group,which are“Initiated Greeting&Self-reporting,”“Step-based Guidance&Evaluation-based Update,”and“Rubric-based Requests.”It also discovers three typical affordance interaction patterns in the GAI tutor group,which are the“Content Proxy Loop”,“Basic Error Correction Loop,”and“Unclear Dialogue.”The divergence in affordance interaction patterns primarily arises from the differences in relationships and status between learners and the two different group of tutors.The findings suggest that GAI tutor design should prioritize enhancing social and pedagogical affordances to narrow the gap with human tutors while exploring new modalities of human-AI collaborative teaching to synergize their strengths under current technological constraints.

关 键 词：生成式人工智能 可供性 智能导师 滞后序列分析 人机关系 

分 类 号：G45[文化科学—教育学]