Building Gender Equity in the Age of Artificial Intelligence (AI)

By: Patricia Santillán

From software that creates images or targeted advertising, to avatars for affective relationships, it seems truly magical. However, the reality is that many times we don't realize how artificial intelligence perpetuates the deep-seated inequities in our society, reinforcing and exacerbating biases, prejudices, and injustices.

Artificial intelligence not only learns, but also recreates our biases, as it operates through machine learning algorithms and is fed by databases. For example, if you have used Google's servers, you have likely encountered reCAPTCHA, a security step in which different images are displayed and you have to indicate those that show a traffic light or a motorcycle. This is a basic example of an algorithm machine learning, Google is using our responses to train its AI and improve image recognition. This makes databases a fundamental aspect because if the data is biased or incomplete, to give a crude example related to Google, there are only sport motorcycles, then the algorithm will learn and replicate those same biases; every time you ask for a motorcycle, it will represent a sport one.

Following the same line of thought, one of the main problems when considering the implementation of an AI model lies in the availability and representativeness of the data (in addition to the criteria considered in its collection). The data reflect power relations, inequities, and representations that respond to a patriarchal, hegemonic, and heteronormative society, which benefits and represents a few and dictates what is «right» or «wrong» in terms of gender roles and expressions.

For example, we have less data on people working informally (where most are women), less data on clandestine abortions, and less data on trans or non-binary individuals. The lack of data means there are aspects that systems will never be able to «learn,» and this will affect their outcomes. This includes the creation of unrepresentative images, where luxury in Mexico is represented by white people with foreign features, or algorithms designed to facilitate the hiring process that favor male candidates because they are trained on a history that reflects gender biases.

In the age of artificial intelligence, incorporating a gender perspective is fundamental. AI can have a profound impact on society, and it is essential that it be designed in a participatory, open, and democratic manner, not solely for the purpose of improving efficiency and productivity. Ultimately, artificial intelligence is a tool, It has the potential to facilitate and streamline processes, but also to cause harm, depending on how we use it. The question nowadays is no longer whether we use it or not, the question is How can we build and use artificial intelligence that is more inclusive and representative?

Whether using artificial intelligence occasionally for everyday tasks or developing models or systems, it's important to recognize that we contribute collectively, each from different levels and perspectives. 

Here are some aspects to consider:

Taking a critical approach to AI's results and implications. We must not accept the results as absolute truths, but rather question them, especially from the perspective of communities and individuals who have historically been rendered invisible and oppressed.

Foster AI access and education. When used responsibly, it can become a tool that takes us very far. This involves not only ensuring physical access to technology, but also providing training to understand and use it from an inclusive and diverse perspective.

Promote diversity in the data and algorithms used to train AI. In order to ensure that systems and models can capture a wide range of human perspectives and experiences, which translates into more equitable and representative outcomes.

Demand regulations that define the parameters and scope of AI. From a diversity and equity approach to address issues related to decision-making, accountability for outcomes, and the protection of potential impacts of AI on society.