面向复杂任务解决的多智能体协同学习环境构建OA

Against the backdrop of increasing intelligence and scalability in education,learning environments are increasingly required to support more complex learning scenarios and competency development needs.However,several challenges persist while implementing large-scale individualized instruction in complex problem solving tasks,including difficulties in authentically replicating real-world contexts and insufficient provision of personalized scaffolding.To address these challenges,this study adopted a design-based research method and,through three iterative cycles,developed a multi-agent collaborative learning environment for cultivating clinical reasoning in medical students.The system comprises multiple roles-an AI patient,an AI assistant,and an AI preceptor-that collaboratively reconstruct authentic clinical scenarios and provide learners with process-oriented cognitive scaffolding.The study was conducted in three phases and involved more than 70 medical students participating in multiple rounds of diagnostic tasks.By inte-grating system log data,survey responses,and process mining analysis,the results indicate that the multi-agent collaborative learning environment significantly improves diagnostic accuracy in high-complexity clinical reasoning tasks.Moreover,learners who achieved successful diagnoses demonstrated a dialectical cognitive cycle involving evidence weighing,cognitive gap identification,and scaffold seeking.This study provides both theoretical and practical support for the design of learning environments that leverage multi-agent collaboration to facilitate complex problem solving.