Collaborative Learning Environments by Educational Agent Feedback Systems
Xin Qi, Hao Meng, Xinguo Yu · 2024
The creation of large language models (LLMs) has opened up new possibilities for diagnosing learning progress through artificial intelligence (AI). However, the practical application of LLMs in actual educational diagnostic situations is limited ability to gather learner data in a proactive manner. This research introduces a diagnostic system based on LLMs that improves collaborative learning by simulating tutors. Our system involves two external agents to handle collaborative tasks. The first agent employs a reinforcement learning approach to formulate learning progress assessing and conduct initial diagnoses. The second agent uses LLMs to parse tutoring guidelines by utilizing student online learning record data, to generate simulated dialogues between students and tutors and evaluated the diagnostic abilities of our system. Our system showed impressive performance in both tasks of evaluating learning progress and making differential diagnoses. This research represents a step towards more seamlessly integrating AI into tutor settings, potentially enhancing the accuracy and accessibility of learning diagnostics.