Uncovering Implicit Gender Bias in Narratives through Commonsense Inference

TL;DR

This paper investigates implicit gender biases in AI-generated narratives by analyzing characters' motivations and mental states using commonsense reasoning, revealing biases similar to explicit ones but expressed subtly.

Contribution

It introduces a method to uncover implicit gender biases in stories generated by language models through commonsense inference, expanding understanding beyond explicit bias detection.

Findings

Female characters are portrayed focusing on appearance.

Male characters are depicted emphasizing intellect.

Implicit biases align with previously observed explicit biases.

Abstract

Pre-trained language models learn socially harmful biases from their training corpora, and may repeat these biases when used for generation. We study gender biases associated with the protagonist in model-generated stories. Such biases may be expressed either explicitly ("women can't park") or implicitly (e.g. an unsolicited male character guides her into a parking space). We focus on implicit biases, and use a commonsense reasoning engine to uncover them. Specifically, we infer and analyze the protagonist's motivations, attributes, mental states, and implications on others. Our findings regarding implicit biases are in line with prior work that studied explicit biases, for example showing that female characters' portrayal is centered around appearance, while male figures' focus on intellect.