A sociological understanding of how AI adopts and reproduces oppressive, traditionally ‘feminine’ gender roles.
The relationship between gender and artificial intelligence grows increasingly complex. Many are familiar with deepfakes (media created by AI tools that depict women and girls without their consent—a form of technology-facilitated gender-based violence). However, not all AI-based subjugation of women is as visually violent as deepfakes often are. AI also reinforces traditionally ‘feminine’ gender roles, thereby perpetuating oppressive notions that harm society overall.
The reinforcement begins with the power of naming. Many AI chatbots have traditionally feminine names. For example, in New York, a school district purchased an AI teaching assistant named Sally. Further, Amazon’s Alexa and Apple’s Siri both use feminine names and default ‘feminine’ voices. In fact, some servile AI assistants are named after real-life women, like McKinsey & Company’s AI assistant Lilli, which is named after the company’s first ever female employee, Lillian Dombrowski.
Linguistic relativity posits that language shapes how we view the world, thereby influencing our behaviour. Giving AI tools feminine names is not only based on outdated views of gender, but it also influences and perpetuates gender norms. AI companies giving their AI-powered assistants female names reinforces the oppressive notion that women must be subordinate in society—forever taking orders from the dominating man.
Pairing these women’s names with assistant roles is particularly heinous. Lilli is named after a pioneering woman; Alexa is inspired by the Library of Alexandria, one of the most significant libraries in history; Siri is named after one of the co-founders’ coworkers. From a critical view, these naming instances are belittling and dehumanizing. Not only is femininity assigned to something non-human, but two trail-blazing, knowledgeable, hard-working women are reduced to mere assistant roles.
The concept of hegemonic masculinity can help illustrate the impact of such a practice. Hegemonic masculinity promotes male dominance and female subordination within society by placing great value on certain social behaviours, such as aggressiveness, self-confidence, and lack of emotion. Among these performances of gender is the idea of women as subordinate. Because society is deeply entrenched in hegemonic masculinity, it bleeds into beliefs which then motivate actions, such as feminizing AI.
Assigning gender to something non-human perpetuates the idea that women are less than men. The depiction of women as virtually non-human allows further dehumanization, as they are placed into inherently subordinate roles wherein they can only take orders and exercise no autonomy or agency. AI companies naming their non-human products after actual women and female-associated learning institutions is not a tribute—it’s dehumanizing, and ensures that the power of those who benefit from hegemonic masculinity is maintained.
Moreover, AI creators choose to give their ‘knowledgeable’ and ‘assertive’ AI models traditionally male names. Some examples include Claude, Watson, and Jasper. These tools are marketed as ‘equals’ to the user. For example, Claude’s website describes the AI as “your thinking partner for big ambitions.” Contrarily, Amazon’s Alexa is “happy to help.” Even within the digital world of AI, there exists a gender hierarchy, where masculinized AI tools are seen as smart and objective, and feminized AI tools are seen as mere servants; this hierarchy is no different than what exists in our society.
Feminized AI models are described and used as assistants. Users can offload their entire workload to these models, saving time for other activities. This reflects another sociological concept, known as “the second shift,” which illustrates an issue working women face: not only must they focus on their careers—their “first shift”—but they must also work a “second shift” at home (completing household tasks). This not only places physical and emotional burdens upon women, but it also reinforces oppressive, traditional gender roles of women being caretakers.
Feminized AI models don’t work separate shifts, but they are continually assigned work by their male users—work male users can do themselves, just like the household tasks that husbands, fathers, and sons can complete themselves. Even the feminized AI is subject to subordination, burdened by the assignments of its male user while he takes credit for its work and benefits by maintaining power and control, further legitimizing hegemonic masculinity’s inescapable influence.
As AI becomes virtually inescapable, it also influences gender-based oppression across the spectrum. For example, AI naming adheres to a strict gender binary, which can influence views on gender performance and conformity, as there is no understanding of non-binary or gender fluidity in how these tools are constructed.
Lastly, there lies another aspect. There was a recent internet trend of users creating slurs for AI tools. However, the slurs evidently drew on historically anti-Black tropes and experiences. AI models themselves are not inherently racialized, yet just as companies assign gender to these tools, it seems some users and non-users have assigned race to AI. Such a depiction only worsens the unique oppression women of colour face; likening them to something non-human further dehumanizes them. Returning to the idea of the “second shift,” women of colour have historically engaged in never-ending shifts of unpaid labour; AI assistants similarly work endlessly, but they lack the actual experiences of misogyny and racism that real women of colour experience. Nevertheless, this continuation could further oppress women of colour, as it reinforces the idea of women of colour as subservient.
Further, Sally, the AI teaching assistant, was purchased by a school on Seneca land with a large Indigenous student population. But Sally is a pilot program—her interactions with students, teachers, parents, and guardians are mere tests. This decision concerns community members because it’s a parallel to how residential schools conducted unethical tests on Indigenous children.
The way individuals interact with feminized AI assistants can bleed into their interactions with real-life women and girls. Depicting AI assistants as female perpetuates gender norms. Users will absorb and practice these ideas of gender roles, like burdening women with work male counterparts can do, believing women are subservient, and treating them as non-human things. As such, we ought to be more cognizant about AI construction and usage, as it will affect users and society overall—including women of colour and individuals across the gender spectrum.
👏🏾👏🏾👏🏾