Diversity AI watched all Oscar nominations for Best Picture – the result was not very diverse

The Oscar Academy has for many years been accused of lack of diversity when it comes to Oscar nominations.

At MediaCatch, a next generation audio and video media intelligence company based in Copenhagen/Denmark, we wanted to check the diversity in the Best Picture nominated movies using AI. We subjected each movie to the same analysis, going through the cast of each movie. But in the diversity context: one thing is cast, another thing is actual screen time.

We made our system watch all ten movies in the Best Picture category and used our AI technology to track diversity: age, gender and caucassian/non-caucassian. Having “watched” all the movies – we were ready to open the envelope:

And the Oscar goes to: White males.

Key findings

In general, the nominated movies are still male dominated.

• Only 1 out of 10 movies offered more screen time to female characters than male characters. Diversity high-five to CODA. Dune was the movie with the least amount of female characters on screen – only 21%. 
• In only 3 out of 10 movies, caucassians weren’t the dominant ethnic group. Those movies were Drive My Car, King Richard and West Side Story.
• In 8 out of 10 movies, female characters were ostensibly younger than male characters.
• In 7 out of 10 movies, the dominating emotion was “sad”. For 3/10 it was “neutral”. Dune was the markedly saddest movie, only registering as happy 4% of the time.

So on this International Women’s Day, it would seem pretty obvious:

Diversity action is needed equally in boardrooms AND on silver screens around the world.

How we tracked diversity

We used AI to identify all faces on screen with an accuracy of 99%. For each face we used AI to predict gender, age, racial group and more. The accuracy for each prediction is shown below. The reporting is based on total screen time.

AI accuracy:

• Identify gender: 98%
• Identify if person is cuacassian/non-caucassian: 90%
• Identify perceived age: +/- 4 years
• Identify emotions: 80%

Deep dive in to our data

The following section consist of data and plots. Feel free to use it as long as you remember to credit us. If you are interested we can also provide raw data.

About MediaCatch

MediaCatch is all about using modern tech tools to create actionable insights that enable not only better business decisions – but more importantly aid purposedriven companies worldwide in creating change.

MediaCatch is Scandinavia’s next gen audio and video media intelligence and research company. Our own software helps find the video and audio you want to find. In broadcast, radio, podcast, YouTube – and soon social media.

Our AI software is based on new revolutionary ways of understanding media intelligence gathering in real time – and we are using it to innovate several industries:

• Media intelligence: Real time gathering of media exposure data – content and context – across all modern media. 
• Sponsorship analytics: Our AI quantifies the actual value and exposure of your sponsorship in real time – for both rights owners and sponsors.
• Diversity: Our proprietary technology tracks your brand/companies diversity in media across all outlets – allowing you to track the progress and footprint of your DEI-strategy.

In short: We work with broadcasters, brands and companies to help them make better business decisions based on live media insights.