Researchers at the University of Washington (UW) have used machine-learning technology to analyze about 800 movie scripts to measure how much power and agency they appoint to individual characters, and they found widespread gender bias in the portrayal of male and female characters.
The study revealed the consistent depiction of women in ways that reinforce gender stereotypes, such as in more submissive roles and with less agency than men.
The UW researchers assessed the power and agency inherent in 2,000 commonly used verbs, and applied machine learning to the movie scripts to automatically recognize genders of 21,000 characters based on names and descriptions.
Using natural-language processing tools, they studied which characters appeared as a verb's subject and object, then computed how much agency and power were ascribed to these characters with crowdsourced connotation frames.
Male characters' tendency to be imbued with more power and agency than females also was found to be consistent across all genres.
From UW News
View Full Article
Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA