关键词: 人工智能写作, 人工智能叙事, 大语言模型写作, 人机协作

Abstract: Generative artificial intelligence (AI) is intervening in literary writing practice, but its narrative capabilities remain controversial at the theoretical level. The problem lies in the gap between its generative mechanism based on statistical prediction of correaltions and the intentional causal constructions upon which human literary narrative depends. Through an empirical analysis of LLM-generated texts using the BERT model, this paper reveals the prevalent tendency towards correlation-based writing and its attendant problems in narrative generation. It further examines the evolution of prompt engineering from direct prompts to “chain of thoughts” and “thought maps,” and analyzs how external structures can introduce causal and temporal constraints into large-scale models. In this sense, AI is no longer understood as a substitute for human authors, but as a narrative collaborator. This invites a re-examination of authorship in AI writing, text reading, and the issues of causality and temporality, thereby laying a methodological foundation for the theoretical construction of “AI narratology.”

Key words: AI writing, large language model writing, AI narratology, human-computer collaboration