In Conversation with Dr. Sirine Taleb
By Nahida Shehab
Dr. Sirine Taleb is currently a lecturer at the American University of Beirut, teaching machine learning courses at the Maroun Semaan Faculty of Engineering and Architecture, as well as supervising Master of Science in Business Analytics students in their capstone presentations at the Suliman S. Olayan School of Business. She is also a researcher at the provost office’s newly established AI, Data Science, and Computing Hub, under which she further offers an online asynchronous course on customer analytics for professionals. After connecting with her, she agreed to let me interview her about her experience as a Lebanese woman in machine learning–the following outlines our conversation.
When did you first get interested in a career in STEM, and how do your many degrees relate to one another?
I enjoyed technical classes like math and physics in the ninth grade, so I decided to pursue a Lebanese Baccalaureate in General Sciences. After ranking first in Lebanon in the official exams, I earned full scholarships to many universities and picked the American University of Beirut (AUB). I wanted to study civil engineering, but my family discouraged me claiming that it is a male-dominated field and that I am not capable of doing, as a woman, the fieldwork that men do. I then chose to pursue a bachelor’s degree in Electrical, Electronics, and Communications Engineering, a less field-based major that was booming in Lebanon and the Middle East.
I maintained my interest in application-based courses and completed a senior project about having an energy-efficient network selection between 3G and Wi-Fi in mobile devices. I then continued my education at AUB. In 2012, machine learning was also booming, so I decided to go for a PhD in EECE with a concentration on Machine Learning, spanning everything from fuzzy systems to artificial intelligence. Apart from the technical aspects, I was also interested in numbers and wanted to broaden my knowledge in the financial sector, so I completed a certificate in financial management. Following that, I enrolled in Science to Data Science, a bootcamp in London for PhD holders, with the aim of bridging the gap between academia and industry by applying what I learnt in machine learning to a real-world business problem. I collaborated with a fintech firm on a project to identify fraudulent transactions in the blockchain domain, specifically in bitcoin.
Are you still affected by your family not encouraging you to pursue a career you were interested in as it required a lot of field work?
I don’t regret not going into civil engineering now as I ended up liking EECE, but my desire to help women in STEM grew. I began volunteering with NGOs like the Institute of Electrical and Electronics Engineers (IEEE), which supports all engineers worldwide. Part of it is an affinity group called Women in Engineering, for which I have served as treasurer for five years and secretary for two. The primary objective of IEEE’s Women in Engineering affinity group is to encourage women to study engineering and remain in the profession, as many women migrate to other disciplines when faced with barriers. And these obstacles can arise from within. As a woman, I had, and occasionally still experience, to be honest, impostor syndrome, during which I do not feel competent enough get the work done, yet I always look back and feel as though I was worried for nothing. I further sense that people value my inputs more than I do myself. I believe that my field being male-dominated plays a role. Indeed, when you follow a male-dominated field, you find that there are specific standards you must adhere to in order to be considered a leader. In my case, according to UNESCO, women make up only 29% of those working in science research and development worldwide.
I was volunteering with no prior knowledge of how NGOs operate or what skills I would need in that setting. So, I decided to fill this gap by participating in MEPI’s Leadership Development Fellowship. Throughout the program, I attended courses by Duke University and a research institute in Tunisia about leadership, civic engagement, social entrepreneurship, and social inclusion, which I especially found interesting. I was aware of the necessity of involving women in engineering, but I had not considered, for example, those from rural regions, those with physical disabilities, etc. Hence, I learnt how to help support these minority communities. There were also some lectures on how to establish partnerships with NGOs. I am a member of Arab Women in Computing, which has the same goal but is more focused on the Arab world. These lectures made me think about how different organizations that operate independently could make a greater impact if they joined forces.
From being a student to becoming a lecturer, has the number of women in engineering classrooms today, interested in a career in STEM, and more specifically machine learning, evolved?
Undoubtedly, it has evolved very noticeably. While I was in engineering school, women made up 10% of the class; in contrast, now, I believe women make up 50% of the classrooms I teach. Moreover, I never worked in an all-girls team throughout my university years; teams composed of one woman and four men would typically work on a project, with the woman bearing its writing portion. Today, from what I’ve witnessed in class, all-girls teams are working on an entire project, including the technical part, and even creating publishable material. We have papers now with the co-authors all being women. Things have evolved, and I think the problem is being tackled from its grassroots. More women are graduating with engineering and, more particularly, AI backgrounds, and the job market will ultimately have to adapt.
Countries that invest heavily in AI research are mostly developed ones, including the United States, China, Canada, and several European countries, such as the United Kingdom, France, and Germany. Has it had an impact on you, not only as a woman, but also coming from the Middle East?
As an Arab researcher, there are still barriers, though. For instance, when applying for funding from abroad, you notice the requirements often include that the fund’s Principal Investigator must be of a specific nationality or work at an American or European institution. You thus cannot apply from AUB unless you have some international partnership. Additionally, when submitting a paper to a conference, you sometimes feel obligated to add a co-author from a foreign university. However, this issue is getting addressed at several conferences through the implementation of a double-blind revision. In short, authors submit their paper anonymously, so their nationality, gender, and other attributes are not displayed, and it gets reviewed solely based on its content. Later, if accepted for publication, the names get disclosed. On a positive note, though, PwC recently published data forecasting that AI will be valued at $20 million in the Middle East by 2030, with the UAE accounting for a sizable portion of that figure. That is only 2% of the worldwide contribution, roughly 15 trillion dollars, but I think it is still good. The impact can be even greater if our governments start pushing for more AI-based technology and funding.
Do you believe that there are bias errors concerning gender, ethnicity, and other types of discrimination that arise with AI?
Indeed, there are biases at various levels, such as gender, race, nationality, and so on. At the end of the day, we are building AI based on real data gathered from humans, which will surely be biased in one way or another. Some companies, for example, no longer manually review CVs and instead use AI to screen them. Consider a men-dominated company. Various characteristics in their existing CV descriptions would indicate if the CV belonged to a man or a woman. If one of the extracurricular or volunteer activities listed on the CV is to assist women-based organizations, the AI will conclude that the CV belongs to a woman. Men rarely participate in such activities. Maternity leave is another factor that can identify a CV that comes from a woman. A maternity leave is generally lengthier than a paternity one, if available, as many countries still do not offer one. The gap usually leads AI to assume the CV belongs to a woman. If a CV with the mentioned characteristics is trying to get into a male-dominated company, the AI will conclude that it differs from those of the employees who work there and reject it. That will lead to a gender bias in the hiring process. I believe this problem should and can be solved by eliminating these biases, returning to the original data on which we are developing our machines, and attempting to normalize, or balance, this data. Going back to my example, the number CVs associated with a man and a woman supplied to the screening machine should be equal. There are technical solutions such as undersampling and oversampling the problem when one category outnumbers the other; You can undersample the dominant category or oversample the one that requires more data to be equal. Weapons of Math Destruction by Cathy O’Neil is a highly worthwhile read addressing this.
Can you please highlight available opportunities for current students and professionals interested in your field?
The Women in Data Science conference is taking place at AUB on the 27th of April. All speakers are women, but everyone is welcome to attend regardless of gender. I’ve been to it several times before, and I’m assisting in organizing it this year. When I used to go, there were speakers from diverse fields, including medicine, engineering, art, and so on. It has become increasingly multidisciplinary as AI and data science applications in many domains are growing. It would be an exciting opportunity for students and professionals interested in AI or data science to network, for this profession is heavily reliant on connections. Additionally, many firms are coming, which would be helpful for students seeking job opportunities.