Modeling Cross-Disciplinary Communication in Smart Education via Generative Agent

The application of generative agent models in smart education has become a hot topic in artificial intelligence research, with widespread use in fields such as personalized learning recommendations, intelligent question answering, and automated assessment. However, existing researches in smart education face challenges in cross-disciplinary communication, making it difficult to accurately simulate the language styles and thinking patterns of students from different disciplinary backgrounds. This leads to limited semantic transfer ability and weak knowledge integration capabilities, thereby limiting their effectiveness in multidisciplinary contexts. To address this issue, this study introduces the generative agent model driven by large language models (LLMs), based on Qwen-Max, to simulate interactions between humanities and science learners. Six metrics are designed to quantify and evaluate the quality of communication. Experimental results show that cross-disciplinary communication has lower information redundancy and stronger language adaptability compared to intra-disciplinary communication. The memory mechanism is introduced to improve communication coherence and knowledge integration. This research provides empirical support for the application of agents in cross-disciplinary collaboration and offers new directions for the development of smart education systems.