Gender and AI: What Algorithmic Bias Has to Do with You
Who does it concern — or should concern — the challenges surrounding algorithmic bias, gender, and artificial intelligence?
We'll get back to that question, but first, an invitation. Something common at Taqtile: when someone takes an issue to heart, goes deep on it, and brings it in an organized way to the right forums, that issue moves forward.
That's what happened when part of the Taqtile team became sensitized to the ethical and responsibility challenges associated with the uses and implementation of AI-powered solutions. First came the committee, then the study plans, exchanges with specialists, initial experiments, and, of course, iterations. A few months later, we're on the verge of launching a responsibility and ethics assessment for AI projects.
Structured across 10 dimensions — governance, data, quality, security, and human impact — the assessment makes visible what needs to be done to make AI solutions safer and future-ready. Whether in production or still in development.
I mention this because we're currently testing the assessment with partners and have decided to open a few spots for new ones, at no cost. If you have an AI solution in development or in production and are interested, reach out via LinkedIn for a test.
Invitation done, back to the subject.
Who does it concern the challenges of Gender and AI? This article begins with an initiative from the Impostoras Club, a women's affinity and support group at the company, and ends with a provocation that reached the entire organization a few weeks ago.
In this article, I'll tell you about the training Mulheres no Comando: What No One Ever Told You About Reaching the Top, a women's leadership program, led by Jéssica Paraguassu*. What started as an internal event for women ended as a company-wide workshop titled: Who Programs the Future? Gender, Power, and Decision-Making in the Algorithmic Era.
The Importance of Naming the Discomfort
With Naomi Sato's permission [Hi Naomi o/], I learned that discomfort in the corporate environment truly has many faces, but for women, it is far more relentless. It's not a new insight, but listening and reading about it — as you can in the article Naomi published (What Kind of Discomfort Do You Want to Sustain) — is a different experience.
Everyone has felt a certain discomfort at work: a difficulty expressing themselves, or feeling heard and recognized. For women, this is a reality that starts long before adulthood, it traces back to childhood and upbringing.
Present, and often intensified in corporate environments, these discomforts — small and large — accumulate and drive a kind of self-silencing. Part of it is exhaustion; part is a disbelief that speaking up will lead to being heard or acknowledged.
What's at stake is not only that women are being silenced, but an entire perspective, a repertoire, and a way of experiencing the world that would contribute to activities like business. Jéssica made this clear in her workshop, showing that separating this conversation from business is a mistake. In the video we published on LinkedIn — in case you haven't seen it — she illustrates the point well.
That's why, as Naomi explored in her article, it is vital to create spaces where these discomforts can be named and debated. Not by coincidence, this is one of the Impostoras Club's core beliefs: organizing women-only spaces may seem like a small step, but it holds real value for expanding female representation in leadership positions.
Women and Technology: A Broader Picture
This matters just as much in the world of technology and AI. As Jéssica shared in her workshop, over the last 20 years, the share of women graduating with computer science degrees in the US dropped from 37% to 19% (NCES). Beyond that, studies indicate that only 11% of AI leadership positions are held by women, and that globally, women make up 22% of AI professionals (Global Gender Gap Report, World Economic Forum, 2023).
In Brazil, despite a 60% growth in female participation in the tech sector between 2015 and 2022, 83.3% of the market is still made up of men (CNN, CAGED). The figure becomes even more significant when we find that 63% of people working on micro-labor platforms in Brazil are women.
This matters because it is through micro-labor — not only in Brazil — that LLM models are trained. And unlike more formal and well-known positions, pay and working conditions are considerably worse (see the BBC report).
As the film Hidden Figures (2016) shows, the idea that women are not interested in computing and technology is quite inaccurate. Women not only helped found the field — they continue shaping it to this day.
But What Does This Have to Do with Me?
If you care about building relevant products, solving business problems, and innovating — or simply being a more engaged colleague — this topic should matter to you too.
You've probably come across some of these figures:
97% of AI-generated "CEOs" are white men, while "social worker" is almost always depicted as a Black woman (Bloomberg).
Facial recognition error rates reach 34.7% for Black women, compared to 0.3% for white men (Gender Shades).
LLMs are 4x more likely to associate women with "home" than men (Unesco).
But algorithmic bias doesn't end with AI systems. As Jéssica put it while walking us through this landscape, it also shows up in our work. Because this bias doesn't just render invisible the participation and contribution of a significant portion of the population — including in the development of technology itself — it makes clear that women's perspectives are being left out.
Allowing this to persist as AI scales only amplifies the risks and problems, making that part of the population even more invisible. It's no coincidence that as early as 2021, a Recommendation on the Ethics of Artificial Intelligence was formally adopted at UNESCO by its 193 member states — the first global normative framework on the subject of responsible AI.
But beyond familiarizing ourselves with guidelines and frameworks (as we explore below), it's important to recognize in ourselves — especially as men — the reality of this "silent silencing." Because the perspectives lost through the constant discomfort and silencing imposed on women are exactly what innovation processes, and any real business problem-solving, sometimes need. As Jéssica put it:
"We think we are making rational decisions, from the most rational perspective possible, but are the questions we ask, the research we run, the angles we take actually leading us to see the world from a predominantly male perspective?"
What Can I Do Differently? Addressing Algorithmic Bias at Work
This was one of the central proposals of the Gender and AI workshop. Working through the idea that AI is not neutral, and that technical decisions distribute risks and benefits to specific social groups, the training brought a series of provocations.
While we can't cover all the recommendations, nor the depth with which she addressed the topic — explore her training program — we've organized here part of what was shared with us. Among the references was the work of Catherine D'Ignazio & Lauren F. Klein (MIT, 2020) on Data Feminism as a working method.
Framework: The 7 Core Principles
Catherine D'Ignazio & Lauren F. Klein — MIT Press (2020)
Examine Power: Map who benefits from and who is harmed by data;
Challenge Power: Build data that challenges rather than legitimizes inequalities;
Elevate emotion & embodiment: Recognize bodies and emotions as legitimate dimensions of knowledge;
Rethink binaries & hierarchies: Abandon categories that naturalize oppression (binary gender, race, etc.);
Embrace pluralism: Multiple perspectives produce more complete knowledge
Consider context: Data doesn't speak for itself — context makes the data;
Make labor visible: Make visible who collects, annotates, and cleans;
Another model presented was the Algorithmic Impact Assessment (AIA), a standard created by the Canadian government that establishes strict rules for the use of AI and algorithms in public services. The AIA is a tool for evaluating impact through a gender, race, and vulnerability lens before deploying products, revisited continuously.
Map affected stakeholders: Who is affected by the system's decision? Who becomes invisible?
Classify risk by category: High, medium, or low — with a gender, race, and class lens;
Audit training data and outputs: Fairness metrics, underrepresentation, differential error;
Define mitigations and oversight: Human-in-the-loop for sensitive decisions. Contestation channel;
Continuous review: The system changes. The world changes. The assessment must stay alive;
For those who want to go deeper, Jéssica put together an excellent bibliography to dive into these discussions:
Image: Taqtile. Organization: Jéssica Paraguassu.
How Is Taqtile Internalizing These Provocations?
Following the training, some questions began to circulate internally.
First, as Naomi framed it at the start of the workshop — one of the Impostoras Club's main areas of focus — it's not enough to ask how many women exist or are joining the company. We need to look at positions of leadership and influence:
How and with whom are we making decisions?
Which perspectives are being considered?
How have these decisions impacted internal diversity and the diversity of users of our products and services?
Building on this, another question began to shape Taqtile's engagement with the topic: who are we when we are programming the future of others? Danilo Toledo opened the workshop by connecting this question to the company's primary goal for 2026:
"It was not simply a bet on technology — it was a choice about people. The choice behind it was to ensure that everyone here at Taqtile could work with what is recent, current, and emerging. That we stop doing things the way we used to, look toward the future, and that each person can grow, develop Taqtile, develop our clients' businesses. And that requires moments like this one, that make us question who we are while we are accelerating and programming the future of others."
He continued: "Every technical decision we make here, day to day, leaves Taqtile with scale. So leading this transition means understanding the biases in what we are accelerating — that's why moments like this matter." The provocation ultimately became two key questions, reinforcing that this topic demands attention not as spectators, but as decision-makers.
Who are we when we are programming the future of others? Leading this transition means knowing which past we are automating.
What decision in my next sprints already carries choices about data, AI, and bias? What do we need to do so that not only we — but the AI we train — makes better decisions?
Our deepest thanks to Jéssica for this important exchange, to the Impostoras Club, and to Tuanny Martins and Naomi Sato, who led and organized the entire event and training with the help of Bianca Lima, Ysabella Andrade, and Erica Urushibata.
If this topic sparked a question, connect with Taqtile on LinkedIn. We keep publishing on responsible AI, gender, and technology applied to business.
*Jéssica is a Master's student in World Political Economy at UFABC, where she coordinates the Center for Gender and Technology Studies, with research focused on the impacts of artificial intelligence on women's work and lives. With over 15 years of corporate experience and creator of the Mulheres no Comando methodology — which has reached more than 2 million women — she works at the intersection of academic research and practical transformation, dedicated to building a more inclusive future in the face of digital transformation.