Statistics Department Hosts Women in Tech and AI Career Panel
In the United States, 32% of individuals in data and AI positions in the workforce are women and the rest are men, according to the World Economic Forum’s Global Gender Gap Report 2020. The global data is even more stark: only 26% are women (40-41, Global Gender Gap Report 2020). How can we address these gender disparities in technical fields? One approach is to spotlight influential women in tech and to connect them with young women to help them envision their future career path.
Our Department’s Women in Statistics Group organized a few career panels with a reception and dinner, including the most recent Women in Tech and AI Career Panel (sponsored by the Harvard Department of Statistics at FAS and Applied Math and CS at SEAS). The panel featured Ashley Mae Conard, Senior Researcher on the Biomedical Computing Team at Microsoft Research, Health Futures, and Stella Biderman, Executive Director of EleutherAI, and was moderated by Associate Professor Tracy Ke and PhD student Kyla Chasalow. While Conard specializes in building expert-centered, interpretable (meaning that the model and its outputs can be understood) AI and probabilistic models in biomedical settings, Biderman is a world-leading expert in training large language models and one of the foundational members of the open source large language model space. From the panel, certain themes emerged, including navigating different career pathways and challenges, the importance of ethics in their work, and recommendations for self-care.
Ethics in AI:
When questions were raised about the potential risks of open source AI, Biderman responded, “Overall, I think it’s important to develop these tools, e.g. LLMs, and that the benefits outweigh the risks. Whether a model should be deployed also depends on the governance of the model.” Biderman added that EleutherAI rigorously evaluates their technology and has a policy of not releasing any technology that is deemed dangerous. In addition, the audience sought guidance for how to incorporate concepts of interpretability into their work. Biderman and Conard emphasized that there isn’t one way to incorporate interpretability; instead, they recommended focusing on the purpose and context of a model and identifying techniques for increasing transparency within that context. For example, Conard spoke about in her work the importance of incorporating existing biological knowledge into models for biomedical applications (read more about this topic in Conard, DenAdel and Crawford’s article “A spectrum of explainable and interpretable machine learning approaches for genomic studies”).
Forging a Career Path:
The panelists' journeys reveal that there are multiple paths to careers in AI, data science, and statistics. While pursuing her PhD in Computer Science and Computational Biology at Brown University, Conard was thrilled to collaborate with biologists on models that generated insights about genomics data. Based on her grad school experience, Conard embarked on a career in a biomedical setting in which she could leverage her statistical knowledge. While Biderman initially planned to pursue a math or CS PhD, a concussion interfered with her studies and led her towards a different path. Leaving academia allowed Biderman to gain a wealth of experiences, including working in an applied operations position for the federal government and joining a grassroots Discord group focused on building open source models, which was a precursor to the founding of EleutherAI.
Conard and Biderman both expressed their professional interests in developing interpretable machine learning methods and working with large language models. However, while Conard builds open source AI tools for scientists in the biomedical domain, currently for rare disease diagnosis (read more on the Broad Institute’s website), Biderman builds open source tools, such as large language models, and evaluates models for the benefit of researchers across different industries. Biderman and Conard’s career narratives show a variety of valid ways to make career choices; students might choose to pursue a PhD or Master’s, to develop their skills in jobs in different industries, or to throw their talents into a side-project that they are passionate about.
Navigating Career Challenges:
One of the challenges that women in the workforce face is underestimating their qualifications or capabilities, particularly in more male dominated fields. Biderman provided an example of when she, as an undergrad, participated in a prestigious internship but was convinced that she didn’t do well and, thus, didn’t apply for other internships. Years later when Biderman ran into her previous supervisor, who spoke positively about her internship, she realized that her initial self-assessment was incorrect. Biderman added that women underestimate themselves by often only applying for positions in which they meet all or most of the criteria (learn about recent research on this topic at the Harvard Business School: “Breaking Through the Self-Doubt That Keeps Talented Women from Leading"). “Most job ads are aspirational and list many skills; you should apply for a job if you are interested and meet the essential skills – you do not need to be an ‘ideal’ candidate and check off every box,” advised Biderman.
The panelists also addressed the challenge of responding to microaggressions, sexism, and sexual harassment in the tech field. For Conard and other women in grad school or early in their career, it is often difficult to directly confront sexist behavior or to seek support. Conard explained, “These experiences have a harmful impact on women, stifling their abilities to enter into the workforce and to innovate freely.” To counter microaggressions, sexism, and sexual harassment, Conard and Biderman advised that women speak out and use strategies, such as writing and following a script to help them manage their experience. In addition, Biderman and Conard emphasized how other women established in their career can mentor students through difficult times in their careers. When a grant proposal of Biderman’s wasn’t accepted, she turned to other established women, who had faced a similar obstacle, for guidance. By sharing their personal anecdotes, Biderman and Conard encouraged our students to speak out and seek support when they encounter microaggressions, sexism, or sexual harassment.*
Recommendations for Self-Care:
At the end of the panel, panelists focused on providing suggestions for better self-care. Conard described how during graduate school, she often worked nights and on weekends. Reflecting on what she learned, Conard said, “Remember that your time is valuable, and you need to prioritize work that must get done first, so that you can also prioritize personal rejuvenation time.” Both Conard and Biderman shared that, while their professional identities are very important, they also need to prioritize their relationships, hobbies, and sleep. The key is to keep your goals in mind – whether at work or in life – and to assess periodically whether to shift your focus. Conard elaborated on this juggling act: “I love being a scientist, but I also know that if an idea strikes when I’m with my friends, I can quickly write down a note and then get back to my friends.”
By sharing their expertise in statistics, data science, and AI related careers, Stella Biderman and Ashley Conard helped our students and postdocs navigate the next steps in their careers. Through their conversations with the panelists and other attendees at the reception and dinner, our scholars continued to build and strengthen their professional networks. We appreciate the thoughtful contributions of our panelists and look forward to connecting with them again in the future!
*Please note that the Department of Statistics encourages all students who experience sexual harassment, discrimination, or violence to report it to the Office of Gender Equity; all faculty, staff, and other employees are mandated reporters.
