When LLMs Enter Everyday Feminism on Chinese Social Media: Opportunities and Risks for Women’s Empowerment

Paper Title

When LLMs Enter Everyday Feminism on Chinese Social Media: Opportunities and Risks for Women's Empowerment

Publication Info

• Topic area: Interaction of large language models (LLMs) with digital feminist practices in China.
• Keywords: LLMs, everyday feminism, women’s empowerment, RedNote, DeepSeek, gender bias, feminist HCI, social media, feminist discourse, Chinese context.

Background and Problem

• Problem / challenge: The integration of LLMs like DeepSeek into feminist discussions on Chinese social media raises concerns about how these models reinforce or challenge existing gender norms and structural inequalities.
• Significance: Everyday digital feminism is a key avenue for women’s empowerment in China, where overt feminist activism is constrained. Understanding the role of LLMs in this space is crucial for designing technologies that support rather than undermine feminist practices.
• Motivation and related work: Prior research has shown that LLMs often reproduce gender biases and normative assumptions, but little is known about their role in shaping feminist discourse and advice. This study addresses this gap by analyzing how DeepSeek-generated content is used and received on RedNote, a prominent platform for everyday digital feminism in China.

Solution

• Proposed approach: A mixed-methods analysis of DeepSeek’s role in RedNote discussions on “women’s growth,” combining content analysis, feminist critical discourse analysis (FCDA), and comment analysis.
• Novelty:
  1. Examination of how LLM-generated advice circulates in feminist digital spaces and its alignment with existing gender norms.
  2. Identification of opportunities and risks posed by LLMs for everyday feminist practices.
  3. Development of design implications for integrating LLMs into feminist digital spaces.
• Procedure and key techniques:
  • Data collection: Scraped 430 RedNote posts, 139 DeepSeek responses, and 3,211 comments.
  • Content analysis: Categorized topics and attitudes in posts engaging with DeepSeek.
  • FCDA: Analyzed DeepSeek’s framing of women’s growth using Kabeer’s empowerment framework (resources, agency, achievements).
  • Comment analysis: Identified community stances toward DeepSeek’s advice (legitimization, negotiation, resistance, neutral).

Results

• Concrete findings:
  • DeepSeek’s advice predominantly focused on self-optimization within existing structures (79.1%), with limited attention to structural constraints or transformative goals.
  • Achievements envisioned by DeepSeek included career mobility, financial stability, and emotional regulation, often framed as individual responsibilities.
  • Pathways suggested by DeepSeek emphasized self-regulation and strategic planning, with minimal emphasis on collective action or structural change.
  • Community reactions were mixed: 49.2% legitimized DeepSeek’s advice, 16.0% resisted it, 7.6% negotiated its applicability, and 27.2% remained neutral.
• Advantage over baselines:
  • The study provides a nuanced understanding of how LLMs interact with feminist discourse, highlighting both their potential to support sense-making and their risks of reinforcing neoliberal self-optimization narratives.
• Experiments / evaluation:
  • Data analyzed included posts and comments from RedNote, focusing on the hashtag “women’s growth.”
  • Metrics included the prevalence of topics, attitudes, and stances, as well as qualitative insights from FCDA.
• Limitations and future work:
  • Limited by the incompleteness of the dataset due to RedNote’s search mechanisms.
  • Analysis relied on user-shared excerpts of DeepSeek responses, which may not represent the model’s full behavior.
  • Future work could include controlled studies of LLM outputs, multi-layer comment analysis, and user interviews.

Summary

This study investigates how DeepSeek, a Chinese LLM, contributes to discussions of “women’s growth” on RedNote, a platform central to everyday digital feminism in China. While DeepSeek’s advice was widely welcomed, it largely promoted self-optimization within existing structures, neglecting systemic barriers and transformative goals. Community reactions varied, with some users legitimizing DeepSeek’s advice and others resisting or negotiating its applicability. The findings highlight both opportunities (e.g., supporting sense-making and emotional resonance) and risks (e.g., reinforcing neoliberal narratives) of integrating LLMs into feminist digital spaces. The study offers design implications for creating LLMs that better support feminist practices by emphasizing context sensitivity, transparency, and collective empowerment.