Gender stereotypes in artificial intelligence within the accounting profession using large language models

Abstract

This study investigates how artificial intelligence (AI) perpetuates gender stereotypes in the accounting profession. Through experiments employing large language models (LLMs), we scrutinize how these models assign gender labels to accounting job titles. Our findings reveal differing tendencies among LLMs, with one favouring male labels, another female labels, and a third showing a balanced approach. Statistical analyses indicate significant disparities in labelling patterns, and job titles classified as male are associated with higher salary ranges, suggesting gender-related bias in economic outcomes. This study reaffirms existing literature on gender stereotypes in LLMs and uncovers specific biases in the accounting context. It underscores the transfer of biases from the physical to the digital realm through LLMs and highlights broader implications across various sectors. We propose raising public awareness as a means to mitigate these biases, advocating for proactive measures over relying solely on human intervention.