| 英文摘要 |
This study explored methods for enhancing humanities literacy and self-directed learning (SDL). Seventy junior high school students designed personalized generative artificial intelligence (AI)-based learning companions by using a multimodel AI chatbot platform. The learning companions were established using retrieval-augmented generation (RAG), with their responses grounded in specific knowledge bases, such as textbooks and reference materials. This enabled the learning companions to provide real-time, curriculum-aligned feedback on writing and text analysis. Results indicated that using the learning companions led to significantly higher scores in self-management and goal-oriented learning while increasing learning motivation. Qualitative analysis revealed that the students received specific revision suggestions and actionable analytical insights from the companions. Nevertheless, overreliance on AI tended to diminish independent thinking, and the quality of learning companion responses was contingent on knowledge base integration and prompt accuracy. Overall, RAG-enabled AI learning companions can serve as effective scaffolds for humanities education and the development of SDL skills. |