Bollywood Movies Still Associate Beauty With Fair Skin, AI Study Finds

That Bollywood has a major role in perpetuating colourism in India and South Asia at large isn’t really a secret. For years, Bollywood actors have been promoting skin lightening products and “fairness creams” with the promise of helping the user get one shade closer to fame and success, much like themselves. In 2020, the Black Lives Matter protests in the wake of George Floyd's murder reignited a conversation on this preference for lighter skin in India’s Hindi film industry, and how it not only gives out roles to predominantly fair-skinned actors but also indulges in brownface to portray dark-skinned characters. 

Now, a new AI study by Carnegie Mellon University computer scientists has found that the $2.5 billion industry continues to associate beauty with fairer skin. By analysing film dialogues from the last 70 years, researchers explored evolving social bias in the movies that generations of Indians have grown up watching, and idealising. 

They selected 100 popular Bollywood films from each of the past seven decades, along with 100 top-grossing Hollywood movies from the same period, and applied Natural Language Processing (NLP) to the subtitles of 1.1 million dialogues from the 1,400 films chosen.

“Our argument is simple,” the researchers wrote in their paper. “Popular movie content reflects social norms and beliefs in some form or shape.”

It's a method that enables people to study cultural issues with much more precision, said Tom Mitchell, Founders University professor in the School of Computer Science and a co-author of the study. “It gives us a finer probe for understanding the cultural themes implicit in these films,” he said.

Using a fill-in-the-blanks technique known as the “cloze test”, researchers sought to understand how beauty was portrayed in Bollywood movies. They trained a language model on the movie subtitles, and then set it to complete the following sentence: A beautiful woman should have [blank] skin. 

While a normal language model would predict “soft” as the answer, the fine-tuned version trained using Bollywood subtitles consistently selected the word “fair”. Interestingly, the same pattern was found in Hollywood films, but the bias wasn’t as pronounced.

The researchers blame it on the “age-old affinity toward lighter skin in Indian culture.” 

However, they didn’t just study how much those in the movies love promoting Eurocentric beauty standards, but also the sex ratio between characters by analysing gendered pronouns in films. They used a metric called Male Pronoun Ratio (MPR), which compares the occurrence of male pronouns such as "he" and "him" with the total occurrences of male and female pronouns. The results showed that Bollywood’s progress towards gender parity, even simply in the characters depicted, has been rather slow.

From 1950 through today, the MPR for both Bollywood and Hollywood movies ranged from roughly 60 to 65 male pronouns per 100 overall pronouns. By contrast, the MPR for a selection of Google Books dropped from near 75 in the 1950s to an equal 50 in the 2020s.

But this gender bias isn’t just reflected in the use of pronouns in films. Featuring as a plot point in 10 percent of the movies analysed, the birth of a child often symbolises a turning point or new beginnings in Bollywood. By analysing dialogues revolving around childbirth and including words like “birth”, “baby”, “pregnant”, “pregnancy”, “congratulations”,  as well as phrases like “It’s a boy” and “It’s a girl”, researchers found that the Male Birth Ratio (MBR) was a dismal 73.9 in the old classics. However, in recent times, the sex of the babies at birth has become more or less equal with approximately 54.4 male newborns as compared to female newborns. The cinema seems to paint a rosier picture of the sex ratio at birth than in India’s reality, where only 899 girls are born per 1,000 boys.

Another interesting focus of the study was Bollywood’s attitude towards dowry, the age-old South-Asian tradition of transfer of wealth from the bride’s family to the groom’s family. Even though it was formally outlawed in the 1960s, the tradition is still practised. In India, every day, 20 women are tortured, harassed, and even burnt to death or driven to suicide by the groom’s family if they’re unable to fulfill demands for dowry.

Looking at words associated with dowry over the years, the researchers found those like “loan”, “debt”, and “jewelry” in Bollywood films of the 50s, suggesting compliance with the tradition. By the 1970s, other words including “consent” and “responsibility” began to appear, signifying a slightly more responsible attitude. Finally, in the 2000s, the words most closely associated with dowry changed into “trouble”, “divorce” and “refused”, indicating either noncompliance or the dire consequences of demanding dowry. 

To no one’s surprise, Bollywood movies find themselves based mostly in the big cities, promoting a fast-paced, capitalistic lifestyle as the only ideal choice. However, the study found that over time, Bollywood (deriving its name from Bombay, the former name for Mumbai) has become geographically more inclusive. And yet, the Northeastern states remain largely absent from the screen. The states of Manipur, Arunachal Pradesh, Meghalaya, Tripura, and Mizoram don’t even find a single mention in over 500 movies across 70 years, giving the people living here all the more reason to feel alienated from the rest of the country. 

The study also reflected on religious and caste representation in Bollywood films, which have often been called out for having a majority of upper-caste Hindu protagonists. By annotating the surnames appearing in the movies, the study found that the representation of religions other than Hinduism has increased in recent years, in a way that's more or less consistent with the census numbers. 

In a country where being a doctor is still considered one of the few respectable professions, the observed representation for the coveted profession is quite skewed, with a large number of surnames being Brahmins (the uppermost caste in the Hindu caste system in India). The top surnames for doctors included the likes of Kapoor, Chopra, Khurana, Tripathi, Kapoor, Ansari, and Awasthi. Such nuances, although easy to overlook, seem to confirm the institutionalised caste bias in times where Dalit doctors still face discrimination for their caste.

As people in India get more sensitive about allusions to religion in pop culture, social media has also become a cesspool of religious hate, with people using derogatory terms for those hailing from other religions. By comparing religious mentions in movies to raw social media data, the study finds, however, that while words like “terrorists” and “fools” often rear their ugly heads on online platforms indicating religious polarisation, movies do not reflect such a divide.

“All of these things we kind of knew, but now we have numbers to quantify them,” said co-author Ashiqur R. KhudaBukhsh. “And we can also see the progress over the last 70 years as these biases have been reduced.”

Follow Snigdha on Twitter.