AI and Human Interpretation of Multilingual Feminist Discourse: A Study of Aurat March Comments

Abstract

Artificial intelligence is widely used to examine social media discourse; however, its ability to understand complex feminist discourse across various languages has been largely overlooked in the existing scholarship. Utilising AI tools, this study analyses social media discourse on Aurat March which is the most prominent feminist activist moment in Pakistan, to show how an AI-assisted discussion of different cultural frameworks, particularly in the Global South environment, can be useful. This study examines how artificial intelligence (AI) systems interpret multilingual feminist online content. It compares the AI results with human analysis. Human analysis used a framework to evaluate AI outputs against human-coded standards. This approach tracked the incorrect classification of user comments on the Aurat March in Pakistan. This research uses technofeminism and Relational Content Analysis (RCA) to study 450 comments on the Facebook, Instagram, and YouTube pages of the film. A multilingual approach was necessary because users responded to the Aurat March posts in English, Urdu, and Romanised Urdu. The research results clearly distinguish between human and AI interpretations. The AI tool (ChatGPT-5) successfully categorised comments into three different languages; however, it frequently oversimplified sentiment categories and failed to identify rhetorical devices such as sarcasm and humour, while its accuracy decreased when evaluating Roman Urdu and code-mixed content. AI tools can assess sentiment across multiple languages; however, their performance declines in multilingual contexts, especially when they need to analyse code-switching and specific cultural languages such as Roman Urdu. The research suggests that, despite AI's potential, it requires further refinement to effectively handle politically and culturally sensitive discourse in multilingual contexts.