Africa, Access to Opportunity, and AI
On March 13, 2018, I had the pleasure of speaking at a CSW62 side-event entitled “African Girls can CODE: Empowering African Women and Girls through ICT and Coding.” The event was organized by the African Union, Permanent Mission of Denmark to the UN Ministry of Foreign Affairs of Denmark, and the International Telecommunication Union, and hosted at the AU Permanent Mission to the United Nations in NYC.
It was an honor to speak alongside H.E. Amira El Fadil (AU Commissioner for Social Affairs), Ms. Ulla Toernaes (Danish Minister for Development Cooperation), Ms. Ursula Wynhoven (ITU Representative to the United Nations), and Mme Phumzile Mlambo-Ngocuka (UN Under-Secretary-General and Executive Director of UN Women). Below is a transcript of my talk.
When I was 11, I discovered my love for mathematics. I was so interested in the subject that I would often spend entire weekends trying to prove theorems that had been given to us as black-boxes in school.
Now, there was just one problem; in my home country of Ethiopia, students didn’t get to pick their majors in college. Majors were assigned to students based on a variety of factors including performance, preference, and need. At the time, high-performing students such as myself had nearly a 50% chance of being assigned to medicine.
But I didn’t want to be a doctor; I was barely comfortable dealing with my own scratches and bruises, much less anyone else’s. I preferred to study math, but I had limited access to opportunities and resources. My family could barely afford my required textbooks, so extra books were out of the question. And I didn’t have access to summer math programs or extra-curricular activities. My mother has always been a deep believer in the importance of education for women and girls. She had finished her own Bachelor’s degree by taking night classes while raising 3 kids. She was always my fiercest advocate when it came to getting the best education possible. But, there was only so much she could do on her own. So, at the time, doing well in school and becoming a doctor stood as one of the best possible outcomes.
I wasn’t satisfied with this. I wanted to choose what I was going to study. I wanted access to resources that would allow me deepen my appreciation of math.
It soon became clear to me that, if I was to pursue my interest any further, I needed a plan. In retrospect, the plan I came up with relied on so many improbabilities that that it could only have been the product of the unbounded optimism particular to 11-year-olds. But, as luck would have it, the plan worked. Seven years later I found myself at Harvard College, where I would spend four years studying math. I would go on to earn master’s degrees from Harvard and the University of Cambridge before landing as a computer science Ph.D. student at Cornell University.
My story is one of hard work and tireless support from my family, friends, teachers, and mentors. But it is also a story of luck — lots and lots of luck — and that is a problem.
Luck should have less to do with whether or not African women and girls are able to pursue their interests. Every girl, regardless of where she is born, should be given the opportunity to explore whatever field she is most passionate about.
When we talk about the participation of women — and especially African women — in computing and other STEM fields, we talk about how we can get them more interested in these subjects. But, the interest is already there! African women and girls are just given fewer opportunities to develop these interests. And even when such opportunities are presented, they are tenuous in nature and limited in scope. Africans constitute over a sixth of the world’s population and African women constitute one twelfth. No meaningful progress can be made in computing (or any other field) without concentrated efforts to include us.
The lack of inclusion of African minds and especially African women’s minds is especially alarming in the field of artificial intelligence. AI is growing at breakneck speed and impacting almost every aspect of our lives. In the western world, AI’s impact is felt in seemingly benign things as targeted advertising, as well as more consequential areas such as personalized health-care.
Despite this widespread impact, AI conferences that attract thousands of researchers were, until very recently, attended by only a handful of Black and African researchers. Insufficient representation should be a concern in any area, but it is especially troubling in this case. AI holds immense promise to improve people’s lives. But, if approached without care, it can negatively impact already marginalized communities. We’ve seen examples at both ends of the spectrum in the past several years: Ubenwa, a Nigerian AI startup, has developed a machine learning system to detect child birth asphyxia, which can save thousands of babies’ lives every year; on the other hand, COMPAS, a risk assessment tool for determining recidivism in defendants, has been shown to be biased against Black defendants.
In order for any research area to benefit all of society, it must foster the inclusion of a diverse group of researchers, especially those who have been historically excluded. About a year ago, to address the lack of representation of Black and African researchers in AI, along with Timnit Gebru, Moustapha Cisse, and others, I co-founded the Black in AI group. Black in AI is a place for sharing ideas, fostering collaborations, and discussing initiatives to increase the presence of Black researchers in the field. What started as a small Facebook group has now grown to include over 600 members from Africa, Latin America, Asia, Europe, and North America. Over 200 members are women, and we are deliberate about prioritizing the participation of African women. Members of the group span a wide range of career stages, from undergraduates to senior professors, engineers, and entrepreneurs. We work on applications of AI in health-care, algorithmic fairness, development, art and many many other fields.
This past December, we organized the first Black in AI workshop at the 32nd Conference on Neural Information Processing Systems (NIPS). NIPS is the largest annual machine learning conference, which is attended by several thousand researchers, but until recently, there are only a handful of Black and African attendees. This year was different. Our workshop drew over 200 members of the Black in AI group from 22 countries, including 10 African countries. To make this possible, we worked tirelessly for several months. We reviewed over 100 submissions, partnered with several universities and companies — including Cornell, Harvard, Microsoft Research, Facebook, Google, among many others — and gave out over $150k in travel grants. The workshop was a resounding success. We showcased 80 presentations from members of the group, including invited talks and poster presentations. Students who flew halfway across the world to attend the workshop had the opportunity to meet with many researchers, including leaders in the field of AI.
This workshop is proof that African researchers have much to contribute and that the field is not complete without us. The workshop was one small way to showcase some of our research and to amplify our voices, but this important work continues. Black in AI is growing rapidly and members are forming transcontinental collaborations. We are working on a wide range of problems, and many of us are focused on those that affect us specifically. I am humbled to be part of the effort to make algorithms and AI work for all of society. Increasing access to opportunity and empowering African women and girls is key to the future of AI and computer science more generally.
Before I wrap up, I want to tell you a little bit about my own research and leave you with a few closing words. Broadly, my research falls in algorithms, AI, and applications to social good. I am specifically interested in using techniques from these fields to understand socioeconomic inequality and to inform interventions to improve access to opportunity. My work spans a wide range of topics, from understanding the impact of financial shocks, to using data-scientific techniques to understand health needs of Africans, to the role of social capital in economic welfare.
In addition to Black in AI, I also co-founded of the Mechanism Design for Social Good research group with Kira Goldner. The group uses ideas from algorithms, optimization, and mechanism design toward the goal of improving access to opportunity. We cover domains including housing, healthcare, and economic inequality. In all of my endeavors, I aim to work on real-world problems and for my work to have an impact both within and beyond the academy. Both Mechanism Design for Social Good and Black in AI are initiatives aimed at fostering collaborations between individuals with a diverse set of expertise and interests to ensure that algorithms and AI work for all of society.
I hope you are enjoying this incredible event. I look forward to learning more about your work and seeing the collaborations that grow out of this gathering.