Media have an impact on the way people see and deal with different ethnicities and genders. News stories can influence people’s beliefs and stereotypes as well as their behavior in other areas like education and family. Understanding how media and news outlets depict various groups is essential.
Researchers face challenges in this area because stereotypes can be encoded in media in many ways and are not always easily visible or quantifiable. Natural language processing methods have been used to study the association between words in journalist texts.
The importance of images in media is often overlooked. My co-authors, Elliott Ash and Ruben Durante, studied visual stereotypes in newspaper photos using artificial intelligence techniques, and more specifically recent advancements in computer vision.
In particular, we use a custom-trained deep-learning mode that can recognize the identity features (such as gender/race/ethnicity) of people that are shown in the images. This allows us to automate the analysis and not have to manually code thousands upon thousands of images. This approach has another advantage: consistency in the classification is impossible to guarantee for human coders.
Our analysis focuses on Fox and the New York Times, two of America’s most important news sources. We analyzed more than two million articles that were published online by the news outlets from 2000 to 2020. 690k of these articles are accompanying an image.
Our paper also examines occupational stereotypes found in newspaper photographs. We analyze whether newspapers reinforce stereotypes regarding the occupation choices of certain groups. For example, managers are white males who work as managers.
Our computer vision method, combined with text analysis techniques allows us to show how news articles images reflect gender and racial stereotypes. This means that stereotypically female or black jobs are more likely than those that depict the identity groups.
If a greater percentage of the identified group is employed in a particular occupation, it’s considered stereotypical. Because of their large proportion in the U.S., Blacks might be stereotyped to work in occupations such as “mail processor”, even though they are not considered to be a majority. Another example is the “secretary” job, which is stereotypically a female occupation.
Our analysis, which controls for true occupation share in every profession, such as the Black “mail processor”, is striking. We can thus separate stereotypes and pure differences between the representations of different identity groups within occupations in America.
Our results show the many potential uses of A.I. Computer vision tools are used to study social science research questions. Modern computer vision tools allow for a much wider range of analysis due to the widespread use of images in politics and business.
Do media perpetuate stereotypes? (2023, 16 March)
Retrieved 16 March 2023
Copyright is reserved for this document. Except for fair dealings in private research or study, there is no obligation to disclose any information.
Part may not be reproduced without written permission. This content is only for informational purposes.