Online images amplify gender biases, study finds

Big data analysis shows women to be starkly underrepresented in online images, where gender biases are stronger compared to text.

An international team of researchers have analyzed more than one million images from Google, Wikipedia and Internet Movie Database (IMDb) based on gender associations of 3,495 social categories, such as “nurse” or “banker”.

Led by researchers from UC Berkeley’s Haas School of Business, the team then examined billions of words from these platforms and found that “gender bias is consistently more prevalent in images than text for both female- and male-typed categories.”

“We also show that the documented underrepresentation of women online is substantially worse in images than in text, public opinion and US census data,” authors of the paper published in scientific journal Nature.

People are spending increasingly less time reading and more time viewing images online, which “significantly exacerbates gender bias, both in its statistical prevalence and its psychological impact,” researchers said.

A nationally representative experiment carried out by the research team showed that searching for images of occupations rather than reading descriptions amplified gender bias in participants' belief, demonstrating it may have an enduring effect in offline lives.

Artificial intelligence models, trained on the vast amount of data online, including images, could further perpetuate gender, racial, and other social stereotypes, researchers warned.

“Addressing the societal effect of this large-scale shift towards visual communication will be essential for developing a fair and inclusive future for the internet,” they said.

More from Cybernews:

US disrupts Russian espionage botnet

ChatGPT controlled my life for a day

40% of UK adults unaware that AI-generated abuse material is illegal

Volt Typhoon takes the stage: what we know about “defining threat of our generation”

Data-hungry dating apps ranked from top to bottom

Subscribe to our newsletter

Leave a Reply

Your email address will not be published. Required fields are markedmarked