From Antigone and The Injustice to Dou E to AI Writing: ross-Cultural Structure of Female Silence and Its Algorithmic Reproduction—A Comparative Prompt-Generation Study with DeepSeek

Abstract

The link between biases in generative AI and the narrative structures of female silence in literary traditions remains underexamined. This exploratory study addresses this gap by conducting a qualitative, prompt-based close reading of a single large language model, DeepSeek (version V3), across three controlled groups of gender-swapped prompts. Taking Antigone and The Injustice to Dou E as the anchors, this study extracts a three-dimensional cross-cultural mechanism of female silence: structural closure of appeal channels, recourse to supernatural forces as a final remedy, and exclusionary endings with no institutional rehabilitation. These three groups of gender-controlled experiments on DeepSeek have been converted into prompts. This article uses textual analysis to systematically track and compare the differentiated response path of generative AI in the narrative of male and female protagonists. Close reading shows that under the same narrative restrictions, female-led works fully reproduce the chain of tragic silencing, and male-led works systematically retain the secular institutional correction path. Cross-cultural research has shown that the model is more likely to frame the suffering of women through Chinese “wrongful woman” archetypes, lose the spirit of restorative justice, and add Greek heroic resistance ideas and modern legal relief to the story of men. This study uncovers a covert algorithmic bias: AI does not passively replicate corpora but switches between two narrative rule sets according to protagonist gender. The new method of prompt-based close reading does not require programming skills and can help humanities scholars carry out algorithmic criticism.