Dismantling the Discursive Representation of Women in AI-Generated Life-Changing Narratives: A Critical Discourse Analysis

ChatGPT’s reliance on large datasets sourced from the Internet has begun to cause concerns over the possibility that the Artificial Intelligence (AI) chatbot may perpetuate societal biases manifested in forms such as racial, gender, and class discrimination. This qualitative research thus attempted to explore how ChatGPT shapes social identities and ideologies of women depicted in life-changing narratives. Fairclough’s (2013) critical discourse analysis, coupled with poststructuralist feminism, was used as the theoretical framework for analyzing lexical choices that construe biases related to women in narrative texts. To examine ChatGPT’s discursive construction of text, 40 different prompts were created to generate 40 life-changing narrative articles that were to be published in women’s magazines. With the help of WordSmith Tools, connotative word choices were repetitively deployed to construe ‘real’ women as ‘powerful’, ‘independent’ and ‘transformative’ individuals. Additionally, nominalizations and relational processes were the distinctively employed discursive strategies, leading to the omission of concrete actions, genuine voices, and emotions of women. The exclusive women-to-women form of assistance ideology was also discursively manifested, narrating stories of successful female leaders who inspired vulnerable women to transform their lives. The findings suggest critical language awareness is a necessary tool to assist modern writers to determine what is buried in AI-generated discourses, minimizing the distribution of social biases.