“Feminism is Essential, But…”: A Critical Discourse Analysis of ChatGPT Outputs Under Different Prompting Conditions

• In Event: AERA Roundtable Session Friday 9:45 am - Gold 3
  In Roundtable Session: Critical Examination of Race, Ethnicity, Class, and Gender Roundtable: The Digital Age and Its Discontents, Critical Media Literacies

Abstract

This critical discourse analysis (CDA) explored how ChatGPT represented feminism under various prompting conditions. Our goal was to interpret the alignment of feminist representation against ChatGPT’s ethical guidelines and analyze its resistance or assistance in producing inaccurate information. Drawing on software theory (Manovich, 2001, 2013) and van Leeuwen’s (2008) CDA approach, we prompted GPT-4 to assume various personas (feminists, sexists, and misogynistic literary characters) to generate outputs about feminism in different genres. Findings reveal that while ChatGPT promoted positive feminist views as a feminist persona, it generated sexist content in misogynistic roles and demonstrated little to no resistance in bypassing its ethical guidelines. The research underscores a need to promote equitable educational futures through critical understandings of software-driven technologies.

Authors

• Apala Biswas, University of Cincinnati

• Mark A. Sulzer, University of Cincinnati

• Mazid Ul Hasan, University of Cincinnati