Socratic Questioning Scaffolding:A Study on Strategies to Encourage Students to Ask AI Large Models Questions
Posing high-quality questions to AI large models has become an essential skill and fundamental literacy for learners in the era of intelligent and digital technology.However,students currently exhibit a low ability to pose questions to AI models,with issues such as superficiality and lack of cognitive depth in their inquiries,resulting in AI responses that often fail to meet expectations.Socratic questioning provides a valuable framework for improving the quality of students'questions.Based on this approach,the design of a Socratic questioning scaffolding offers prompts,guidance and explanatory support from both"cognitive"and"metacognitive"dimensions,encouraging students to actively and persistently follow up on their inquiries with AI large models,thereby fostering the cognitive development of students.A quasi-experimental study involving 44 undergraduates majoring in Educational Technology revealed significant changes in the depth,the sequences and the structures of students'questions before and after using the scaffolding.The Socratic questioning scaffolding significantly improved students'ability to pose in-depth questions,effectively facilitated the transition within higher-order question sequences,and enabled students to form three typical transformation loops:analytical creation,in-depth understanding,and evaluative creation.Furthermore,it guided students to construct star-shaped,chain-like,and closed-loop question structures based on different perspectives.In the future,educators should leverage diverse scaffolding tools to guide students in posing higher-order questions to AI large models and in establishing logic of questions through follow-up inquiries.By promoting high-quality human-AI dialogue,this approach can enhance deep learning and the development of students'higher-order thinking skills.
Socratic QuestioningAI Large ModelQuestion ScaffoldingHigher-Order ThinkingHigh-Quality Question
NETL
NSTL
万方数据
