Understanding Gender Bias in AI-Generated Product Descriptions

TL;DR

This paper explores gender bias in AI-generated e-commerce product descriptions, developing a taxonomy of biases, analyzing their occurrence in models like GPT-3.5, and highlighting unique biases such as size assumptions and stereotypical language.

Contribution

It introduces a specialized taxonomy for gender bias in product descriptions and provides empirical analysis of bias types in e-commerce-specific language models.

Findings

Gender biases frequently occur in AI-generated descriptions.

Biases include size assumptions and stereotypical feature portrayals.

Distinct biases are identified in persuasive language use.

Abstract

While gender bias in large language models (LLMs) has been extensively studied in many domains, uses of LLMs in e-commerce remain largely unexamined and may reveal novel forms of algorithmic bias and harm. Our work investigates this space, developing data-driven taxonomic categories of gender bias in the context of product description generation, which we situate with respect to existing general purpose harms taxonomies. We illustrate how AI-generated product descriptions can uniquely surface gender biases in ways that require specialized detection and mitigation approaches. Further, we quantitatively analyze issues corresponding to our taxonomic categories in two models used for this task -- GPT-3.5 and an e-commerce-specific LLM -- demonstrating that these forms of bias commonly occur in practice. Our results illuminate unique, under-explored dimensions of gender bias, such as assumptions about clothing size, stereotypical bias in which features of a product are advertised, and differences in the use of persuasive language. These insights contribute to our understanding of three types of AI harms identified by current frameworks: exclusionary norms, stereotyping, and performance disparities, particularly for the context of e-commerce.