With the proliferation of female robots like Sophia, chatbots like Amelia , and the popularity of female virtual assistants like Siri (Apple), Alexa (Amazon) and Cortana (Microsoft), artificial intelligence seems to have a gender issue.

This gender imbalance in AI is a strong trend that has drawn strong criticism in the media. Even UNESCO warns of the dangers of this practice, as it could reinforce stereotypes that women are objects.

But why inject femininity into artificial smart objects? If we want to curb this bias, we must begin by better understanding the deep roots of this phenomenon.

Make the inhuman more human

In an article published in the journal Psychology & Marketing , we propose that research into what makes us human (versus machines) may shed light on why AI is almost systematically personified by avatars. female. We propose that if women tend to be more objectified in AI than men, it is not simply because they are more associated with assistants, but because we tend to attribute more humanity to women. women than men.

Why ? Because women are seen as more sensitive, warm and more likely to experience emotions than men, the feminization of objects in AI helps to humanize these objects. Indeed, being warm and feeling emotions are qualities perceived as fundamental to being totally human (unlike simple competence)… and these qualities are lacking in machines.

Drawing on the theories of dehumanization and objectification , we show through five studies and a total sample of over 3000 participants that:

• Women are seen as more human than men, globally and in relation to non-human entities.
• Female robots are endowed with more positive human qualities than male robots, and they are seen as more human than male robots, whether compared to animals or to machines.
• The perceived humanity of female robots increases the perception of how it is thought that they will treat us in a medical context (we believe that female robots will take more into account the uniqueness of our case), which leads to more favorable attitudes towards female AI solutions.

We have used several different scales to measure the “perceived humanity” of robots, both in relation to animals and machines. For example, to measure the perceived humanity of female and male bots versus animals, we used the Ascending Humanization Scale based on the Progress Walk . We explicitly asked respondents online to rate how well they viewed female or male robots on this scale, using a slider ranging from ancient primates to modern humans.

To measure humanity’s perception of female and male robots in relation to machines, we have created a scale that measures mechanistic (de) humanization, adapting the march of progress. This scale represents the evolution from robot to man (instead of ape to man). Of course, we also created a female version of each of these scales.

Other measures captured more subtle and implicit perceptions of humanity, asking respondents what qualities they attribute to male and female robots. There are certain qualities that are supposed to distinguish humans from machines (eg, “sympathetic”, “happy”), and other qualities are supposed to distinguish humans from animals (eg, “organized”, “polite”). Finally, we also used an implicit association test to determine whether female robots are associated with the concept of “human” rather than “machine”, and more so than male robots.

The ghost in the machine

While female robots are perceived to be more human than male robots on most measures of perceived humanity , our results show that male robots are perceived as more human (than female robots) on certain negative human qualities. Overall, these results suggest that, of course, female robots are endowed with more positive human qualities (benevolent sexism), but beyond this benevolent sexism, female bots are perceived as more human and more inclined to take into account our unique needs in a service context.

This would explain why female robots are preferred over their male counterparts: People probably prefer female intelligent machines because they are more strongly associated with humans.

If femininity is used to humanize non-human entities, our research suggests that treating women as objects in AI precisely lies in the recognition that they are not. The most frequent hypothesis, called the “dehumanization hypothesis”, is however based on the idea that it is necessary to consider the “other” as an animal or an instrument, before objectifying it. In other words, dehumanization would be a prerequisite for objectification – the targets of objectification generally being deprived of their humanity. Contrary to this mainstream view, the transformation of women into objects in AI could happen not because women are seen as sub-human, but because they are seen as superhumans in the first place.

This is in line with Martha C. Nussbaum’s ideas on objectification:

“Objectification involves transforming into a thing … something which is not at all a thing” ( Martha C. Nussbaum , 1995)

as well as in the direction of Kate Manne’s ideas on misogyny and dehumanization:

“Often it is not a woman’s sense of humanity that is lacking. His humanity is precisely the problem ”( Kate Manne , 2018).

The widespread use of female identity in AI artifacts is thus likely based on the implicit recognition that women are seen as human, and more so than men.

Objectification of women in the real world?

In summary, this research draws on what makes us human versus machines to better understand the deep roots of the pervasive feminization of AI. Because emotional sensitivity forms the essence of our humanity, and because women are seen to be more likely to experience this sensitivity, we suggest that the feminization of intelligent objects increases the perception that these objects are human and capable of understanding our unique needs.

However, this process of turning women into AI objects could lead to the objectification of women in “real” life by conveying the idea that women are objects and mere tools designed to meet the needs of their owners. This could potentially contribute to the objectification and dehumanization of women in the non-virtual world.

This research thus sheds light on the ethical dilemma faced by AI designers and decision-makers: women would be transformed into smart objects, but by injecting the humanity of women into smart objects, these objects appear more human and are more adopted. .

These results are not particularly encouraging for the future of gender parity in AI nor for ending the objectification of women in AI. Developing neutral voices could be a way to move away from this practice and stop the spread of this benevolent sexism. Another solution, similar to Google’s recent experiment , would be to randomly assign a male or female smart robot with equal probability to users.

Author Bio: Sylvie Borau is Professor of Ethical Marketing at TBS Business School