The mission of WIDS is to provide talks about the latest developments in data science and networking opportunities for women in the field. To that end, I attended 13 talks, a career panel, a lunch discussion, and a final mingling session. The talks were beneficial in three ways.
First, I was excited to learn about cutting edge developments in data science. For example, Finale Doshi-Velez presented on how advanced models were used to optimize treatments for patients with HIV. Imagine you have two patients, one who is very similar to several other patients we’ve seen, and one who is very distinct in terms of their health features. For the first patient, you might want to predict what treatment will work best for them based on what worked best for those who were similar to them, their “neighbors.” However, for the latter patient this won’t work, because they don’t have any nearby neighbors! Here you may need to use a model where what you know about their health features (age, gender, comorbidities) determines treatment. Finale Doshi-Velez proposed a hybrid model, where the influence of each of these approaches is mediated by how near the patient’s neighbors are. This allows to build models that predict what treatment will have the best outcome for the patients.
Second, the talks and panels exposed me to the variety of career paths in which data science is applied. In addition to health care I was able to hear about the application of data science in more traditional tech sector such as google cloud, government such as the National Security agency, and non-profit such as the Human Rights Data Analysis Group (HRDAG). The last was particularly novel to me. I was excited to learn that HRDAG is using many of the technical packages I use in my own work to answer questions relevant to human rights. They used analysis of databases to determine a more accurate count of the number of victims of violence in Syria. It is much easier to count the number of documented, identifiable, victims than it is to estimate all victims. Application of data science tools can help agencies to make accurate of assessments of where there is need for intervention in the world.
Third, the talks were an excellent summary of some of the best practices in data science. Claudia Perlich from dstillery (a targeted advertising company) presented the talk with my favorite name “The secret life of predictive models.” She discussed two examples where a model was actually predicting things TOO accurately. For example, in one case the model was making predictions by identifying the camera a photo was taken with, rather than the content of the image. She emphasized the importance of digging in to models to be really aware of what they are doing.
In addition to the talks, one of my favorite parts of the day was the lunch discussion session. For this, we signed up based on topic, and I attended a lunch on Data Ethics and Governance. This lunch was important to me because one of my values as I pick companies to target after graduate school is integrity. This lunch discussion allowed me to see how companies were (or were not) dealing with ethical issues as they come up in their big data analysis. I also got to meet women at different companies, and get their perspectives on how to select companies of integrity for employment.
Attending WIDS was hugely beneficial to me. I got to meet many women who are successful in Data Science, an often male-dominated field. I learned a lot from the talks, and also from chatting with these women about how they found the roles they love. I think the best practical tip I got while attending WIDS was to attend meet-ups. Meet-ups allow one to learn about the culture at different companies, which will help me when I ultimately decide where to apply.