Tackling the gender-blindness in the legal and policy framework for new digital technologies: Artificial Intelligence and Gender Bias
The recent Covid-19 pandemic has sparked unparalleled rates of digitalization in all economic sectors during the last five years. However, women are falling behind at every stage of the life cycle of artificial intelligence, the technology driving the digital revolution. The self-reinforcing nature of the gender gap in AI results in a technological and economic system where women are disproportionately underrepresented, which has an impact on the progress of gender equality.
Although AI can be used to improve our daily lives, research clearly demonstrates that not everyone equally benefits from the progress of digital technologies as gender biases are deeply established in AI data sets. Artificial intelligence has the capacity to replicate human prejudice and sustain bias, thus furthering the marginalization and stigmatization of women worldwide. Consequently, it takes a committed and methodical effort to combat bias in machine-human interfaces and make sure that technology does not reinforce negative gender stereotypes and societal injustices.
Based on UNESCO estimates from 2019, women only make up 13% of AI researchers, while they represent only 6% of software developers. Demand for digital and AI-specific skills is rising as AI changes the labor market. However, when it comes to developing digital and AI skills, women keep being left behind. In the 16–24 age group, more young men than women in the OECD countries have achieved programming capabilities, which are crucial for the advancement of AI. Only 18% of C-suite executives in AI companies and top start-ups globally were women in 2019. Given these realities, a reasonable question to ask is: How does this representation gap show up in the technologies themselves?
"Prejudiced actions or thoughts based on the gender-based perception that women are not equal to men in rights and dignity" is what the European Institute for Gender Equality (2023) defines as gender bias. Thus, this is the fundamental process via which gender prejudices both impact and are influenced by technologies. Although AI may be viewed as an objective, neutral technology in and of itself, its human application in certain situations gives it new meanings and implications. Since gender biases are internalized in our society and culture, they contribute to the contextual factors that affect how AI technologies are used and understood, which in turn causes the technologies to become entrenched with the same biases.
AI-generated decision-making, dataset training, and algorithm development are three processes where gender bias can manifest itself. AI programs are operated by algorithms, which are sets of guidelines for problem solving. This process comprises the computation of converting input data into output data. Thus, the type of data that is entered can have a direct impact on the decisions that algorithms generate later on. As a result, the algorithms may reproduce any biases present in the original data. Due to over- or underrepresentation of certain social categories in the database, algorithms can contribute to discriminatory outputs. AI systems that produce prejudiced predictions and outputs for particular segments of the population are biased. Therefore, a biased dataset does not adequately represent society and reflects the representation of an unjust society.
The UNDP Accelerator Lab is testing two AI image producers' perceptions of women's representation in STEM disciplines as part of an ongoing study project. Upon requesting visual representations of an engineer, scientist and IT specialist, the researchers saw that between 75 and 100 percent of the generated results featured male individuals.
Gender biases are embedded in AI by design as algorithms reflect the social and cultural norms of those who design the systems. An individual's safety is endangered, opportunities, resources or information are unfairly distributed. Civil liberties are violated and certain demographics do not receive the same extent of opportunities as others. Moreover, a person's wellbeing is negatively impacted by biased AI systems that are derogatory and offensive.
Additionally, the growth of AI technology has facilitated new forms of cyber violence and digital marginalization by reinforcing gender stereotypes, sexism and discrimination online.
AI has the capability to generate content that is deceptive enough to produce fake news, deepfakes and other forms of digital manipulation. Deepfake pornographic videos are an example of gender bias driven by AI as a new kind of gender-based violence. These are created by superimposing photos of people's faces or bodies on top of one another using AI, giving the appearance of real recordings. Gendered misinformation and deception efforts are increasingly directed on women, especially those in politics and other leadership roles. For female political leaders from minority groups who are well-known in the media and outspoken advocates for feminism, this tendency is typically even more noticeable.
The Wilson Center report Malign Creativity: How Gender, Sex, and Lies are Weaponized Against Women Online of 2021 assesses the use of online gendered and sexualized disinformation campaigns against women in politics and beyond through an analysis of online conversations about thirteen female politicians across six social media platforms, amounting to over 336,000 pieces of abusive content shared by more than 190,000 users over a two-month period. Women are mostly silenced by these risks or after experiencing such abuses, which lead them to stay away from politics and other male-dominated fields and to disengage online.
UNESCO reported in 2020 that 20% of female journalists have been subjected to physical violence and assault as a result of online attacks. As a consequence, 17% of the women who were the primary targets of online harassment stated that they had begun to feel physically insecure as a result of the attack, and a smaller percentage even stated that they had begun to miss work due to the possibility that the attacks could turn into physical abuse.
Established gender roles serve as the foundation for structural hierarchies and preconceptions in society, which are replicated and reinforced by the trend to “feminize” AI technologies. AI can be gendered in a number of ways, including voice, appearance and the usage of feminine pronouns or names. The default voices of home-based virtual assistants, for example Apple's Siri and Amazon's Alexa, are feminine. The "submissive personalities" and stereotypically feminine traits, such as being helpful, polite and obedient, were intended for these devices. Voice assistants can be even subject to verbal and sexual assault, which could trivialize such actions directed towards women. Technology has the potential to assist achieve gender equality, but machines that imitate patriarchal notions reject this promise. With an increase in the representation of women in technical and managerial roles inside technology businesses, it would seem less likely that digital voice assistants would take a lighthearted approach to sexual harassment or provide an apology for verbal abuse.
In order to handle issues related to gender, race, ethnicity, socioeconomic status and other factors in an ethical manner, it is necessary to adopt an intersectional perspective. Moreover, human rights-based AI governance that is founded on openness, accountability and human dignity must be implemented. Various parties, such as IT companies, academia, UN agencies, media, civil society organizations, businesses and corporate entities and other pertinent actors, must collaborate and look at cooperative solutions to achieve this objective.
In April 2021, the European Commission presented a proposal for an artificial intelligence regulatory framework within the EU, that is the draft Artificial Intelligence Act. The specialized application of AI systems and the risks involved are the main topics of the proposed legal framework, however .an intersectional perspective has not been provided yet.
To create and sustain gender-sensitive and ethical AI tools, diversity in the industry is essential. Enhancing the gender balance in the AI sector can be accomplished in two primary ways: by eliminating the gender gap in AI-related fields of study and focusing on women's entry into and retention in the field.
A positive move in this regard is the UN secretary-general's proposal for a Global Digital Compact, which is to be adopted at the Summit of the Future in September 2024. "Outline shared principles for an open, free and secure digital future for all" is what the Global Digital Compact is supposed to do. The Common Agenda report outlines potential topics for discussion, such as digital connectivity, preventing Internet fragmentation, giving users control over how their data is used, upholding human rights online and fostering a reliable Internet by instituting accountability standards for deceptive and discriminatory content, including potential gender biases.
At Politics4Her, with the purpose of properly addressing gender discrimination in AI, we strongly believe that, first of all, more gender-equal digital skills education and training are essential and, second, the AI Act and future AI-related regulations must include an intersectional and gender-sensitive perspective to avoid keeping reproducing a discriminatory and unequal society in the digital progress. The latter instead is a great source for advancing gender equality if institutions and IT corporations start to work on enforcing equal representation from the inside and providing equal access and education to digital resources to all individuals. Gender-blindness in the development of new digital technologies does not only reflect gender roles and stereotypes already existing in society and does not allow everyone to equally benefit from digital opportunities, but also does enhance obstacles for women to achieve gender equality and economic empowerment, especially in the technological sector, in which gender and power hierarchies keep strengthening gender gaps further and further.
