3 Female AI Leaders on Dismantling Gender Bias in AI

In the world of AI development, there’s clear evidence of entrenched bias against women, which prevents female pioneers in the field from getting the recognition they deserve.

Back in 2020, a World Economic Forum report found that women make up only 26% of AI positions. A new report from Deloitte helps to explain why: Women often face a continuous battle for credibility in the AI space.

“Women in AI, regardless of position or seniority, are often constantly faced with resistance, questioning, and judgment,” the report found.

Deloitte surveyed 200 experts in AI and 71% of respondents agreed that adding women to AI and ML will bring unique perspectives that are needed in high tech, that AI solutions would benefit from having more diversity in designer and developer positions (66%), and that AI models will produce biased results as long as the field remains male-dominated (63%). 

The message is clear: This needs to change.

We wanted to hear from the women building with AI and get their take on why this bias exists and how to dismantle it. So we interviewed three experts: Mida Pezeshkian, a member of A.Team’s AI Guild and the founder of STEMA_cg; Yael Burla, Principal Product Marketer at Vimeo who leads the company’s GTM strategy for AI, search, collaboration and analytics capabilities; and Sarah Bird, Global Lead for Responsible Engineering at Microsoft.

Here’s their vision for challenging bias and elevating women in the AI space.

Mida Pezeshkian

As a former scientist turned business leader and member of A.Team’s AI Guild, Mida Pezeshkian serves as a translator to connect technical and non-technical teams to drive innovation. She immigrated to the U.S. from Iran when she was in high school and quickly adapted, learning to combine her culture and upbringing with possibilities and opportunities that were available to her in America. “I truly believe that we need to be open-minded to take advantage of AI’s capabilities,” she said.

Here’s why AI has such entrenched gender bias 

There is a history of fewer women landing C-Suite roles and over time fewer women pursuing training and work in physical sciences, which is showing itself in the sparse women representation in AI. There are a variety of reasons why this is the case so it’s paramount that we ensure the next generation of women in the workforce succeed in STEM. In the US and European countries, women’s representation in STEM tends to be less than countries in the Middle East. For example, women in Iran – where I grew up – represent about 70% of university graduates in STEM.

Women such as Ada Lovelace, Katherine Johnson, Marie Curie, Rosiland Franklin, Maryam Mirzakhani and Jennifer Dauda are great role models.

Get creative about drawing girls into STEM programs

In affluent countries, there are more opportunities to reach similar status or income in a variety of fields so there is less concentration in STEM. We need to promote the value of deep STEM education in other ways. I am convinced that we need more girls to become fascinated with science at an early age. That means getting caretakers and the broader society to encourage exploration and experimentation. Women such as Ada Lovelace, Katherine Johnson, Marie Curie, Rosiland Franklin, Maryam Mirzakhani and Jennifer Dauda are great role models. We need their stories to be well known and appreciated by the next generation. 

AI can solve massive problems in healthcare

The appeal of AI lies in its ability to solve complex problems efficiently. I am very proud of leading a team that built the first physician-group developed electronic health record system integrated with various AI techniques and natural language processing to improve operational efficiency and clinical decision making. We were leveraging data to not only inform our decision-making but using it to automate and optimize our operations. This was during the pandemic, when we had fleeting resources so with every improvement we were helping our providers care for their patients. It was very rewarding to know that we were supporting people who were on the front lines delivering vital services to the elderly population in long term care and skilled nursing facilities.

Yael Burla

With a background in studying human behavior and expertise in multimodal AI, Yael Burla has a unique perspective and deep passion for building AI in a responsible and human-centered way. She’s a trained neuroscientist from the University of Pennsylvania, where she worked in a decision-making lab for two years. Now, she works as a Principal Product Marketer at Vimeo, where she leads the GTM Strategy across a number of critical product areas, including AI. 

“I’ve always been fascinated by how humans behave and make decisions,” she said. “In order to build intelligent machines—especially in a safe way—you have to understand how the brain works.”

We can address the inequalities that are exacerbated by AI—if we act now‍

With multimodal AI developing at breakneck speed, there is a heightened sense of urgency for us to come together as humanity and address the inequalities that can be exacerbated by technology and that exist in the workplace. By 2025, 90% of online content is expected to be generated by AI. The rapid democratization of this technology puts it in the hands of humans who are naturally biased, and who are largely unaware of their own biases. The striking power and underlying mechanics of AI, especially in its ability to further amplify existing biases and content, is not pervasively known or fully appreciated. 

Correcting bias in AI isn’t easy 

Unless the technology is built with the guardrails and awareness of biases that exist within us as humans, women and minorities more broadly will continue to be impacted through underrepresentation and discrimination. Remember the classic example in 2018 when Amazon’s resume scraping tool discriminated against women? The algorithm was trained on existing top performers, favoring resumes with words that men tend to use, like “executed” and “captured,” while downgrading terms centered around women as well as graduates from all women’s colleges. Naturally, this only further worsened the problem of promoting the dominance of white males in technical and senior roles.

A recent study found that women only hold 26% of analytics-related and AI positions.

We need new standards for responsible AI implementation 

We need to set a standard of how we define let alone implement AI responsibly, and hold companies accountable to doing so through transparency, fairness, accountability, and privacy. But how can we do that when the other main problem is the gender gap in the global AI workforce? A recent study found that women only hold 26% of analytics-related and AI positions. There needs to be a more concerted effort around upleveling women in the field of AI in order for them to participate equally in these conversations, and be appropriately acknowledged for their contribution to the space.

AI is biased because our data is biased‍

Many believe that biases only get created through the algorithms at play. But humans as well as data itself can be biased. When the data we’re using to train models is inherently biased—for example by not containing an equal representation of individuals across demographics—or if the algorithms are only being built by white males, we experience inherent gender bias. Bad data in, amplification of that bad data out, leading to potentially harmful outcomes. A good example of this is that women weren’t even included in medical studies until the 1990’s.

Sarah Bird

‍With a Ph.D in computer science from Berkeley, Sarah Bird has worked in the field of responsible AI since its inception. She works as one of the founding researchers of Microsoft’s first responsible AI group and contributed to the creation of the Microsoft Responsible AI Standard. Before joining the team at Microsoft, she worked as a founding member of the AI ethics group at Facebook. She was first introduced to engineering when two women—who were engineers—came to speak to her fourth-grade class. Now, she dedicates her own time to mentoring and encouraging more women to get into the field of responsible AI.

AI needs women—full stop 

Having women in the AI field is crucial for ensuring that the technology addresses the needs and concerns of women. Without their involvement, we risk missing opportunities to solve important problems. The economic impact of AI is enormous, and it is essential that women are not left behind in this economic revolution. Women must have the opportunity to participate in and benefit from the new industries and companies created by AI.

If AI technology is not developed and tested with intentionality it can inherit gender bias

AI is biased because data is biased

If AI technology is not developed and tested with intentionality it can inherit gender bias. This is because AI is built from data, which can include historical examples and content from the internet that is biased. This is why responsible AI is necessary; it’s why I do what I do. Having more women and more diversity in the field can help ensure the development of AI that is inclusive and addresses these biases. However, it is not enough to simply have diverse teams. It is also important to work with experts who understand and study different types of biases to help build and test the technology.

AI presents an opportunity to right old wrongs

I was drawn to AI because I believe it has the potential to make the world a better place by creating solutions that are designed for individuals, rather than a one-size-fits-all approach. Historically, many solutions, including medicine, were not designed with women in mind. AI has the potential to change that. AI is the perfect bridge between people and technology, and it requires a deep understanding of both to create effective solutions. Innovating across these two areas is incredibly rewarding because both are interesting and important.