630 Billion Word Internet Analysis Shows ‘People’ Are Interpreted As ‘Men’

What do you visualize when you read words such as “person”, “persons” or “individual”? Chances are the image in your head is of a man, not a woman. If so, you are not alone. A massive linguistic analysis of over half a trillion words concludes that we assign gender to words that, by their very definition, should be neutral.

New York University psychologists analyzed the text of nearly three billion web pages and compared how often words for a person (“individual”, “people”, etc.) were associated with terms referring to a man (“man”, “he”) or a woman (“female”, “she”). They found that masculine words more frequently overlapped with “person” than feminine words. The cultural concept of a person, from this perspective, is more often male than female, according to the study published April 1 in Science Advances.

To conduct the study, the researchers turned to a massive open-source dataset of web pages called Common Crawl, which fetches text from everything from corporate white papers to Internet discussion forums. For their analysis of the text – a total of more than 630 billion words – the researchers used word embeddings, a computational linguistic technique that assesses the similarity of two words by looking at how often they appear together.

“You can take a word like ‘person’ and understand what we mean by ‘person’, how we represent the word ‘person’, by looking at the other words we often use around the word ‘person’,” explains April Bailey, a postdoctoral researcher at NYU, who conducted the study. “We found that there was more overlap between the words for people and the words for men than the words for people and the words for women…, suggesting that there is this masculine bias in the concept of person.

Scientists have previously studied gender biases in language, such as the idea that women are more closely associated with family and family life and men are more closely associated with work. “But it’s the first to study this very general gender stereotype – the idea that men are somehow the humans by default – in this quantitative computational social science way,” says Molly Lewis, a researcher in the department of psychology. from Carnegie Mellon University, which was not involved in the study.

The researchers also looked at verbs and adjectives commonly used to describe people – for example, “extrovert” – and found that they were more closely related to words for men than to words for women. When the team tested gender-stereotypical words, such as “brave” and “kill” for men or “compassionate” and “laughing” for women, men were associated equally with all terms, while that women were more closely associated with them. seen as stereotypically feminine.

This finding suggests that people “tend to think of women more in terms of gender stereotypes, and they tend to think of men only in generic terms,” ​​Bailey says. “They think of men as people who can do all sorts of different things and specifically think of women as women who can only do stereotypical things.”

One possible explanation for this bias is the gendered nature of many supposedly neutral English words, such as “president”, “fireman”, and “human”. One way to potentially counteract our biased thinking is to replace these words with truly gender-neutral alternatives, such as “president” or “firefighter.” Notably, the study was conducted using mostly English words, so it’s unclear if the results translate to other languages ​​and cultures. However, various gender biases have been found in other languages.

Although the bias of thinking that “person” equals “man” is somewhat conceptual, the ramifications are very real as this trend shapes the design of the technologies around us. Women are more likely to be seriously injured or die in a car crash because when car manufacturers design safety devices, the default user they envision (and the crash dummy they test ) is a male with a heavier body and longer legs than the average female.

Another important implication concerns machine learning. Word embeddings, the same linguistic tools used in the new study, are used to train artificial intelligence programs. This means that any bias existing in a source text will be detected by such an AI algorithm. Amazon faced this problem when it emerged that an algorithm the company hoped to use to screen job applicants was automatically excluding women from technical roles – an important reminder that AI is as smart or biased as the humans who form it.

Leave a Comment