Salvato in:
Dettagli Bibliografici
Autori principali: Williams, Joshua N., FitzMorris, Molly, Aka, Osman, Laszlo, Sarah
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2405.05382
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910439400538112
author Williams, Joshua N.
FitzMorris, Molly
Aka, Osman
Laszlo, Sarah
author_facet Williams, Joshua N.
FitzMorris, Molly
Aka, Osman
Laszlo, Sarah
contents Text-to-image models are now easy to use and ubiquitous. However, prior work has found that they are prone to recapitulating harmful Western stereotypes. For example, requesting that a model generate an "African person and their house," may produce a person standing next to a straw hut. In this example, the word "African" is an explicit descriptor of the person that the prompt is seeking to depict. Here, we examine whether implicit markers, such as dialect, can also affect the portrayal of people in text-to-image outputs. We pair prompts in Mainstream American English with counterfactuals that express grammatical constructions found in dialects correlated with historically marginalized groups. We find that through minimal, syntax-only changes to prompts, we can systematically shift the skin tone and gender of people in the generated images. We conclude with a discussion of whether dialectic distribution shifts like this are harmful or are expected, possibly even desirable, model behavior.
format Preprint
id arxiv_https___arxiv_org_abs_2405_05382
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle DrawL: Understanding the Effects of Non-Mainstream Dialects in Prompted Image Generation
Williams, Joshua N.
FitzMorris, Molly
Aka, Osman
Laszlo, Sarah
Computers and Society
Text-to-image models are now easy to use and ubiquitous. However, prior work has found that they are prone to recapitulating harmful Western stereotypes. For example, requesting that a model generate an "African person and their house," may produce a person standing next to a straw hut. In this example, the word "African" is an explicit descriptor of the person that the prompt is seeking to depict. Here, we examine whether implicit markers, such as dialect, can also affect the portrayal of people in text-to-image outputs. We pair prompts in Mainstream American English with counterfactuals that express grammatical constructions found in dialects correlated with historically marginalized groups. We find that through minimal, syntax-only changes to prompts, we can systematically shift the skin tone and gender of people in the generated images. We conclude with a discussion of whether dialectic distribution shifts like this are harmful or are expected, possibly even desirable, model behavior.
title DrawL: Understanding the Effects of Non-Mainstream Dialects in Prompted Image Generation
topic Computers and Society
url https://arxiv.org/abs/2405.05382