Of the 100 highest-grossing Hollywood films in 2015, those with female leads out-grossed those with male leads by 15 percent. Yet research by Shri Narayanan and his team at the Signal Analysis and Interpretation Lab (SAIL) found that women in Hollywood spend significantly less time on screen or speaking than their male counterparts, about only 30 percent.
Narayanan, professor of electrical engineering and computer science, partnered with the Geena Davis Institute on Gender in Media to develop a cutting-edge tool to analyze unconscious gender bias in Hollywood.
“This tool is useful because it picks up on details that a human observer might miss,” said Narayanan, who is also director of the Ming Hsieh Institute at USC Viterbi. “For example, we were able to show that even when a woman was speaking, the on-screen presence was more often than not that of a man.” Another finding revealed that in male-led movies, men enjoyed three times as much screen time as women. However, in female-led movies, men and women were on screen about the same amount of time.
Known as the Geena Davis Inclusion Quotient (GD-IQ), the tool uses signal processing and machine learning to objectively process the details of movies. Fast and relatively cheap, GD-IQ detects gender by facial recognition, tracks on-screen time, analyzes linguistics and word choice, and processes audio. Narayanan is quick to point out that GD-IQ is still a tool that relies on human ingenuity.
“We want to empower and support human decision making,” he said. “While we’re passing on some of the dirty work to algorithms, it’s up to people to interpret the information and come up with real-world solutions.”