March 27, 2026

The World's Biggest AI Study Forgot Half the World

This month, Anthropic released the world's largest study on global AI attitudes. They surveyed 80,000+ people, 159 countries, 70 languages. I thought this could be a really useful resource for me as I prepared for the panel discussion that I’ll be participating in at Harvard’s AI Policy Forum later today. But instead, this mass compendium of data was useless to me. Why? No sex-disaggregated breakdown.

Why useless? Because when we aren’t careful about selecting participants equally or when we don’t bother to assess women’s needs, risks, preferences and concerns separately, we risk “rounding up” (as if women are a set of decimal points to every male round number) to whatever it is that men think.

The Forum’s topic for my panel is AI in the emerging markets. We have far more information about how women feel about generative AI use in developing country contexts. Here again we bump up against another default: we risk assuming that women’s experiences in the US and Europe mirror those of women in places as different as India, Kenya and Mexico.

What we know:

The displacement risk is gendered. The ILO found that women's jobs face 3x the automation risk. UN Women and LinkedIn found that ~80% of women in Asia-Pacific work in roles flagged as augmented or disrupted by AI. But in Sub-Saharan Africa, 89% of employment is informal -- agriculture, street vending, domestic work. Crucially, AI eliminates the formal-sector entry points -- secretary, receptionist, data entry -- that historically served as women's first rung on the career ladder. As economies develop and formalize, the ladder is being pulled up before women can reach it.

AI access infrastructure is uneven. Women in Sub-Saharan Africa are 29% less likely than men to use mobile internet. In South Asia, it's 32%. Globally, 60% of the 885 million unconnected women live in these two regions. And yet: the GSMA reports that progress in closing the mobile gender gap has stalled.

What we don’t know:

How women feel about using generative AI. There are signals. The Anthropic study showed that users in Sub-Saharan Africa and Latin America felt much more positive about AI’s potential - especially with respect to economic gains - than Western users. Does that extend to women in these countries as well? Possibly. In India, 64% of women in India’s tech sector say that AI has accelerated their path to senior roles. 69% say it has opened new career opportunities. Urban women are using ChatGPT to reduce domestic burdens. In China, Female professionals are adopting AI faster than male counterparts and reporting lower anxiety. Africa graduates the highest proportion of women STEM university graduates in the world at 47%, according to McKinsey. People in tech tend to view gen AI’s impact on economies and societies more favorably than those in other sectors. How about every day generative AI users that are female? I’m not sure there’s enough information to know.

There are many well-founded reasons why women should feel more concerned about generative AI’s impacts than men. HBS’s global meta-analysis of gender gaps in generative AI use cites common concerns around (1) differences in knowledge and familiarity with gen AI, (2) differences in confidence and persistence when using the technology, and (3) differences in beliefs about whether AI use is unethical or "cheating." (HBS Working Paper 25-023). Responsible AI Fellow Rebecca Ryakitimbo names others in her thought provoking piece, “Needs and Wants: A Feminist Approach to AI in Sub-Saharan Africa”. In addition to issues around biased data leading to biased results in AI and lack of female representation in machine learning roles, she notes the concerns around an inculcation of values dictated by Global North tech elites and how marginalized voices are excluded from participation and shaping how AI systems are built and deployed.

She also offers a helpful set of recommendations that we would be well guided to heed:

  1. Centering Lived Realities: Start by grounding AI design, use, and evaluation in the daily experiences, cultural contexts, and aspirations of marginalized African groups. Prioritize bottom-up knowledge production and usage of participatory methods such as community-driven needs assessments, co-design sessions, and feminist tech workshops to define what problems AI should solve.

  2. Challenging Extractive Data Practices: Critique and resist data colonialism, where African data is harvested without community benefit or consent. Embrace data licensing frameworks that center downstream impact, such as the Nwulite Obodo open data license.

  3. Intersectional and Contextual Design: Recognize that gender inequality in Sub-Saharan Africa intersects with race, ethnicity, disability, class, and geography. Intersectionality should be applied as a design principle to ask who is included, excluded, and affected at each stage of the AI lifecycle.

  4. Redistributing Power in AI Development: Decenter dominant tech narratives and elevate African women and queer technologists, researchers, and knowledge holders. Prioritize initiatives that promote open access to infrastructure, such as compute power, datasets, and research tools, for feminist actors.

  5. Building Ecologies of Feminist AI Practice: AI should not be built in isolation but as part of a broader ecosystem of justice movements that include digital rights, reproductive justice, climate, and land rights.

  6. Gender-Centered Policy Design: Design policy frameworks, regulations, laws, and acts from a gendered lens to enable compliance in terms of practice. Accountability, fairness, and intersectionality should be at the core of policy design.

I’m skeptical that in the US we will put these ideas to good use, but in the emerging markets, there may be hope for us yet.

Years ago when I was working at an international development organization, I was sitting next to a woman named Lan from the Philippines who was about my age. Lan was concerned about her grown daughter and I’d just said, “Oh, Lan. Remember when we were in our twenties! We were staying out too late with our friends having fun and we pulled it together.” To which she said, “It was different from me. I spent most of my twenties in prison for protesting against corruption in my country.” And then she was gracious enough to tell (clueless) me about her life and about her county’s unsettling history. These are the types of people I got to meet when I was in international development. Courageous, unbroken people who had endured unspeakable traumas and gone on to work for a world where others wouldn’t have to. What so many people get wrong about international development is the assumption that the benefits of aid accrue to the people in “poor” countries from those in “rich” countries. As an American, who by virtue of the vastness of our country and richness of our economy could have been confined to a pretty insular life, the perspective I gained from Lan and my other colleagues was priceless.

I am writing this post as a love letter to a career I’ve had to leave behind. For years now it’s been clear that the design of international development was anachronistic. Too much money and decision-making has historically been trapped in Washington DC and London. The past few years have seen a much needed move to “localization” wherein strategy is set and executed in closer proximity to those who will reap the benefits. It is right. It also left me wondering - after a 20+ year career in this space - what was the point of me?

I am also writing this post as an open letter to others who are suffering career disruptions. Going through this transition - made even more gutting by the violent way in which the US is rejecting international cooperation more broadly - has felt eviscerating. I understand peoples’ pain and, sadly, I foresee a future where there is only more to come far beyond the international development sector.

When you are in a position like this, you are pressed to think about your transferable skills. Probably not a “skill” in the traditional sense, but something I have a passion for, is spotting where women’s unique experiences are overlooked. I learned this from amazing feminists from Guatemala, Vietnam, El Salvador and India during my time at Oxfam and through reading books such one called Invisible Women by Caroline Criado Perez. I learned that “men” have historically been understood as the same as “human”. Women’s interests, experiences and preferences are viewed as “niche”. Sometimes women aren’t seen at all. I had a conversation with my dear friend former Prime Minister Patrick Achi of Cote d’Ivoire where he mentioned that he’d never before in his career been in a room where he was the only man. I said - now you know how most women feel throughout their professional lives. Now I know from conversations with him that he now can see when we aren’t there.

Seeing is one thing. Speaking about it is another. This is my transferable skill. When I started getting interested in generative AI and how it will transform our economies and societies, I was baffled at how little attention was being paid to how women's lives will be uniquely remade. Generative AI is going to hit our economies like a wrecking ball. Unlike a wrecking ball, however, the destruction will not be evenly felt. Take this example of not being seen: a January 2026 paper by Sam Manning and Tomas Aguirre looked into how adaptable American workers are to AI-induced job displacement. Their research finds mostly good news; many occupations that are highly exposed to AI contain workers with strong means to manage job transition. A relatively small percentage work in occupations that are both highly exposed and where workers have low expected adaptive capacity. That is to say, not everyone in an AI disrupted field will feel the pain equally. People are more able to adapt if they have savings, transferable skills, live in places with vibrant economic conditions, and if they are younger. Those without these advantages make up “pockets of vulnerability”. 

What the research fails to mention is this “pocket of vulnerability” is almost entirely female. 89% of the workers least equipped to successfully navigate these disruptions are women. 89%. Though this fact is not seen as sufficiently relevant to even note in the NBER paper; it took a follow-on Brookings analysis of the data to pick up this fact. I was just listening to a podcast about recent research heralding good news on the AI job disruption front. Don’t worry folks… while there may be fewer white collar jobs, blue collar workers like plumbers and electricians will benefit because they can outsource pesky administrative and customer service roles and focus on their trades. I was on the verge of tears thinking… were you planning on mentioning the fact that nearly 90% of all blue-collar work is done by men? Not to mention the fact that admin and customer service jobs are two of the most common occupations for women in the US labor force? And so here comes the speaking out.

I fear that women in the US are going to grow increasingly economically vulnerable. Between this and the Dobbs decision (as of January 1, 2025, roughly 62.7 million women and girls lived under state abortion bans) and child care access and affordability in crisis, I worry that the number of women living in poverty, requiring state support (which may or may not be available) and/or having to depend on others to survive will spike. I worry that women are becoming smaller and even harder to see as we slowly but surely lose our power.

I understand that men are facing similar, guttingly painful disruptions but my view is that policy measures will lean towards initiatives that support sectors in which men thrive (e.g., trades, tech and data-adjacent roles and manufacturing/logistics) because they are seen as intrinsically valuable. Women’s labor is - and has always been  - seen as “nice to have”. 

In the US, unlike in other places, no one is coming to save us. My view is that women - and those that value women’s labor and talents - can do a lot to save ourselves. The NBER study I mentioned earlier states that “AI exposure reflects potential changes to work tasks, not inevitable displacement; only some of the changes brought on by AI will result in job loss”. We can break down our jobs into individual tasks and see which of those tasks can be made easier and faster using AI. As much as I am terrified - truly - by what’s coming I am also so excited by what I’ve learned using AI. I realize this isn’t a 1:1 exchange given the disorientation and chaos that we will have to manage as a society but it has allowed me to deepen my skills and explore my creativity in new ways. 

I started my initiative First Prompt in order to galvanize other women to do the same. Women are less inclined to be using AI for a variety of valid reasons that I’ve described in this piece in the Stanford Social Innovation Review. But at a time where our power is actively under attack, we can not afford to neglect AI tools that can give us leverage. And, we have to encourage other women to do the same. As I said in SSIR, “The response to dizzying hype should not be rejection; it should be fierce ambivalence. That means passionately holding two seemingly contradictory truths at once: We should use generative AI to empower ourselves and others, and we should demand exacting standards of transparency, fairness, and safety from those building and governing these tools.”

With this, I want to extend my gratitude to the Harvard Berkman Klein Center for Internet and Society for offering me a place in its fellowship program starting March 2026. By bringing in someone whose focus has been on addressing gender disparities globally, the Center has evidenced its commitment to interdisciplinary exchange and collaboration. 


PS - This piece was written by me with no help from AI to prove to myself that I still can. For those who want to know why, please see this piece on cognitive surrender.

A love letter to the past.

And to the future.