A Comparative Study on the Meta-functions of Artificial Intelligence and Human Discourse Markers

This study explores the use of discourse markers in AI-generated and human-generated texts, based on Halliday's metafunctional theory.By analyzing 20 Shanghai college entrance exam essays generated by Baidu's AI model Wenxinyiyan and comparing them with 20 human-written essays, it is found that both share similar use of discourse markers.However, AI tends to use transitional and topic-follow-up markers more frequently, while human texts show a higher usage of conditional and causal markers.These findings suggest AI's capacity for maintaining discourse coherence but highlight its limitations in more complex reasoning.