Mark Shea/ BBC:

There is a popular idea that artificial intelligence (AI) is out to get us.It was this public image problem that the United Nations was recently trying to address at its AI for Good conference in Geneva. The event in July was intended promote AI to help solve global problems, and it was described as the largest-ever gathering of humanoid robots.
There was Ai-Da (the “world’s first ultra-realistic humanoid robot artist”) and Grace (the “world’s foremost nursing assistant robot”) as well as Sophia, Nadine, and Mika. There was even a rock star robot, Desdemona. All of these androids have one thing in common – they are all female by design. So why is it that creators typically choose to give their robots feminine characteristics?
It is often argued that the choice to make AI voice systems female is rooted in gender bias. But sometimes there is a more innocent reason for the sex a designer gives their robot: they have modelled it on themselves. This is the case with Nadine, whose creator Nadia Magnenat Thalmann describes her as a “robot selfie”. Meanwhile, Geminoid, the only robot at the conference that was explicitly male, is the spitting image of its maker, Japanese roboticist Hiroshi Ishiguro.
One of the keynote speakers at the conference was Ai-Da, an AI machine which can draw, paint and sculpt, and is also a performance artist. Lisa Zevi, head of operations for the Ai-Da project, tells the BBC that in this particular case, there was a good reason for giving Ai-Da a broadly female look.
“Female voices are typically very underrepresented in both the art and technology spaces,” she says. “We want to give a voice to those underrepresented groups effectively.”Specifically, Ai-Da’s persona is inspired by Victorian mathematician Ada Lovelace – considered by many to be the first computer programmer – as is her appearance.
Other than those robots modelled on an individual, one particular reason is often suggested for choosing a female robot: we have an innate preference for women’s voices.
Karl MacDorman, an expert in robotics and human-computer interaction from the University of Indiana in the US, believes this argument may have a basis. He has conducted research which has found that women prefer women’s voices, and men do not really have a preference. By testing their reactions to voices, the research found a discrepancy between what people report on questionnaires and what they really feel – women like a female voice far more than they admit, and men say they greatly prefer a female voice on questionnaires (even though they don’t really care).
“Thus, a female voice may work better for both groups. Women on average are happier interacting with female voices, and men believe they should be happier, even if they’re indifferent,” MacDorman concludes.
This might not tell the whole story though.
Early versions of AI like Siri and Alexa were given female voices, and MacDorman’s work has been quoted as justification for this choice. Yet he believes some major corporate decisions had already been made for entirely different reasons, and his findings were simply convenient. “I suspect they had made their decision before I had published any work on this topic,” he says. “They probably made the decision for reasons that are unconscious, or reasons that they might not like to admit to, and then they need the justification for it later when they are challenged.”
MacDorman believes our own expectations may play a bigger part in the decision-making than many designers are prepared to admit. “In terms of quality of service or customer service roles I think that they may be more associated with women than with men”The initial stereotype then becomes reinforced just because it becomes a popular choice to give artificial intelligence a female voice.”
MacDorman believes that this could be considered sexist, because the roles that AI typically performs – delivering information or customer service – are in a sense servile.
And he hints that this may also play into male fantasies. Kathleen Richardson, professor of ethics and culture of robots and AI at the UK’s De Montfort University, remembers when humanoid robots did not typically take an adult female form.
“In the lab that I was in [15 years ago], they always made them child-like,” she tells the BBC. “The idea was, if they were childlike, they wouldn’t be threatening to people, and people would be more comfortable with inviting them into their home.”
Richardson says this drive to make androids less threatening has morphed into the female forms that we see today, and it is driven by the preoccupation that we have about the increasing role that technology is playing.
“You’ve got to dislodge this very deep fear of depersonalisation and dehumanisation that comes with introducing more technologies into our lives, particularly in our personal arenas, she says.
“People write reams of how terrifying it is, how the terminators are around the corner. That would be terrifying to have in the home, right?” MacDorman – who has also worked with robots for decades – agrees these fears had a role in robot design, especially early on. “A female android is generally considered more approachable, especially for children, so it was considered better suited to human-robot interaction experiments,” he says.This tallies with his experience of working on robots in Japan between 2003 and 2005 – many of the experiments were with children and the team he worked with believed a female android seemed less threatening.
But Richardson suspects that there may be an altogether more basic motive at play in modern designs of humanoid robots.
She likens robots to art – what you see is just an image on a surface – and believes that robot design suffers from the same issues that modern art critics often lament when they appraise historical paintings. “There was a famous theorist called Laura Mulvey who talked about the male gaze in art, and how male artists were representing female figures. They were normally representing them as submissive, as naked, as objects of male desire. And I think in a way, we’re seeing the male gaze just replicated in robotics, because these are just images on surfaces – there’s nothing that that sits behind these images. There’s no sentient being. There’s no life. “We can’t just transfer what’s going on inside people and inside people’s relationships to these new artefacts that are created.”