How do you picture AI scientists? Odds are it’s male.

By Scout Barker

Title: Who Makes AI? Gender and portrayals of AI scientists in popular film

Author(s) and Year: Stephen Cave, Kanta Dihal, Eleanor Drage, and Kerry McInerney

Journal: Public Understanding of Science (open access)

TL;DR: What do The Emoji Movie (2017) and Transcendence (2014) have in common? They are two of the nine mainstream movies that featured female AI scientists in the last 100 years. In the same time frame, we have had 107 male AI scientists on the big screen. There also isn’t a single film featuring AI that was solely directed by a woman. This study thus further evidences the link between on-screen representation and real-life diversity imbalance within male-dominated STEM professions.

Why I chose this paper: This paper choice was a snowball of my current thoughts and fears. First, I have had more conversations about AI this week than I can count on one hand. Second, watching Lise Meitner in Oppenheimer struck me because I think she is the first female physicist I have ever seen on the big screen. Third, I am reading Invisible Women: Exposing Data Bias in a World Designed by Men by Caroline Criado-Perez

When I tried to picture an AI scientist in a film, I found myself reconstructing archetypes of the lone male genius, think Tony Stark or the guy from Ex-Machina. But this mindset sets the stage for a troubling reality.

You’re probably aware that women are underrepresented in Artificial Intelligence (AI). Just 22% of real-world AI professionals are female, highlighting an imbalance that is seen in academia, at AI conferences, and within AI industry giants like Google and Facebook.

Hence, with this in mind, it probably won’t come as a surprise that female AI scientists are also underrepresented in film.

It has previously been shown that on-screen role models affect real-life STEM diversity. Thus Cave, Dihal, Drage, and McInerney set out to quantify this imbalance, to determine how AI science is communicated in film and to what extent women were given a voice in this conversation.

The Method: 100 years of AI influencers

Our authors hypothesized that the underrepresentation of women in the AI workforce would be paralleled both on and behind the screen. To test this, they analysed 1,413 of the most “influential” films released over the last century.

These films were collated from lists curated by The Guardian, Science, and Wired UK, among many others, and rated by critical acclaim and revenue (as a proxy for viewership) to determine their influence; these methods mirror previous approaches for investigating Gender in Media.

Cave, Dihal, Drage, and McInerney were strict on what constituted AI and an AI scientist. They defined AI as “autonomous technology,” which excluded any artificial humanoids or non-autonomous cyborg technology from consideration. This meant Metropolis (1927) and The Terminator (1985) were included, but movies featuring Frankenstein and his monster were not.

Cave, Dihal, Drage, and McInerney were interested both in the genders of the on-screen scientists and the gender of the film’s director(s). Character genders were determined by the gender pronouns they used in the film, and the best effort was made to determine the self-identified gender of the director at the time of filming/release to reflect how their work was initially received, supported, and funded.

The Findings: 7 women, an Emoji and an Alien

After watching each film, the authors found 142 films featuring AI, 86 of which clearly showed or referred to one or more AI engineers or scientists. This meant that from 1920-2020, there were 116 influential AI scientists on the big screen.

Out of this total, seven characters were human women, one was a female alien*, and one was a female emoji**. This list is so small we can provide it in full.

The nine female scientists represented were:

  • Frau Farbissina (in Austin Powers: International Man of Mystery, 1997);
  • Dr Brenda Bradford (in Inspector Gadget, 1999);
  • Dr Susan Calvin (in I, Robot, 2004);
  • Ava (in The Machine, 2013);
  • Evelyn Caster (in Transcendence, 2014);
  • Quintessa* (in Transformers: The Last Knight, 2017);
  • Smiler** (in The Emoji Movie, 2017);
  • Dr Dahlin (in Ghost in the Shell, 2017); and
  • Shuri (in Avengers: Infinity War, 2018).

But there isn’t enough space in this bite to list the 107 men.

Furthermore, of the 142 films found to include AI, Cave, Dihal, Drage, and McInerney concluded that only two were co-directed by women (1.4%). These were Jupiter Ascending (2015) and Captain Marvel (2019).

However, the films directed by trans sisters Lana and Lilly Wachowski are impacted by the methods, as one or both of the sisters had not publicly transitioned when their films (the Matrix Trilogy (1993- 2003), and Jupiter Ascending (2015)) were released. Jupiter Ascending is therefore only “co-directed by a woman,” and no Matrix film is included.

Therefore, it was found that zero influential films featuring AI were solely directed by a woman.

The Impact: looking at the big picture (70mm)

AI has moved out of the land of science fiction into our everyday experience and is changing rapidly how we communicate. Whether it is in the classroom or Hollywood’s writers’ room, AI use is becoming ingrained into science communication practice. But the impact of AI is still uncertain, so we need AI researchers more than ever.

Cave, Dihal, Drage, and McInerney have again proven that life imitates art – meaning we need to start picturing diversity in AI to achieve it. The consequences of female exclusion from AI are already evident: women are more likely to be adversely affected by AI, and AI also increases the gender pay gap.

Films have the power to shape perception, and perception can shape actions. Telling women they simply need to get into AI is never going to work. Cultural perceptions need to rapidly shift to create an inclusive atmosphere, and we need to recognise the influence of film on culture.

To pave the way for a more inclusive and equitable AI landscape, we need to recognise this influence, advocate for accurate portrayals, and promote diversity.

Only when we normalize AI diversity on the big screen, can we start to undo the cultural construction of the AI scientist as default male and truly promote diversity in the field of AI.

Edited by: Andrea Isabel López and Kay McCallum

Cover image credit: Andy Kelly via Unsplash