Encoding and decoding gender: Investigating bias and language in artificial intelligence models
Abstract
As artificial intelligence (AI) models become deeply embedded in social systems, discussions on their ethical creation and application have intensified, particularly in regards to the consequences of biased models. This study examines how large language models (LLMs) such as ChatGPT-4.o encode, and potentially reinforce harmful social biases. Through a paired-question experiment, this research assesses (1) how gender is encoded in AI models such as GPT4.o, (2) how language influences gendered outputs, and (3) the extent to which AI-generated gender bias aligns with or diverges from human understanding.
Primary Faculty Mentor Name
Luis Duffaut
Status
Undergraduate
Student College
College of Engineering and Mathematical Sciences
Program/Major
Computer Science
Primary Research Category
Engineering and Math Science
Encoding and decoding gender: Investigating bias and language in artificial intelligence models
As artificial intelligence (AI) models become deeply embedded in social systems, discussions on their ethical creation and application have intensified, particularly in regards to the consequences of biased models. This study examines how large language models (LLMs) such as ChatGPT-4.o encode, and potentially reinforce harmful social biases. Through a paired-question experiment, this research assesses (1) how gender is encoded in AI models such as GPT4.o, (2) how language influences gendered outputs, and (3) the extent to which AI-generated gender bias aligns with or diverges from human understanding.
