Team: Akkanut Wantanasombut, Luxsnai Songsiengchai, Chitiphat Suntornsaratool, Matthana Rodyim, and Nonthakorn Jitchiranant | Chulalongkorn University & Srinakharinwirot University, Thailand
A team of researchers based in Thailand at Chulalongkorn University and Srinakharinwirot University has joined the F<A+i>r Feminist AI research cohort with a paper titled “Gender Inequalities in AI Chatbots: The gendered language set and dissolution of gender bias in language use in Thai’s service sectors.”
This collaborative study will investigate the language that AI chatbots use through case studies spanning various service industries (i.e logistics, finance, retail, and platform services) to obtain an understanding of the language patterns and decisions employed by AI chatbots. The study brings together linguistic, cultural, feminist, social, and technical viewpoints on Thai language with the objective of developing a holistic approach to resolving gender imbalances in AI chatbots.
The methodology includes a mixed method approach that includes: 1). A critical discourse analysis to analyze how AI chatbots make language judgments, and how participants’ choices shape their online identities, how those identities reinforce or subvert gender stereotypes and power dynamics, and how these are represented to others. 2). Linguistic analysis methods to detect and correct sexism in the chatbots’ output text. This will involve detecting and adjusting language-generating algorithms to account for gendered language patterns, prejudices, and assumptions.
The nexus between AI chatbots in Thailand’s service industries and feminism is an important one to examine since Thai comprises numerous other languages and has gender norms and is embedded with prejudices. For instance, female personalities are more common in AI chatbots, whereas masculine figures are typically connected with qualities like dependability and authority. This user research will provide valuable insights about the chatbot’s architecture and bring clarity for the implementation of gender-responsive frameworks, as well as the consideration of ethical guidelines for unbiased language generation.