Marzyeh Ghassemi, an associate professor at MIT’s Department of Electrical Engineering and Computer Science, is working to ensure that machine learning models are robust and fair in healthcare applications. Her research group, Healthy ML, focuses on improving safety and equity in health through advanced machine learning techniques. Ghassemi’s work was inspired by her early fascination with video games and puzzles, as well as her interest in healthcare.
She credits her mother for encouraging her to pursue higher-level math and science, and later, Jason Ackelson, a former Marshall Scholar, for helping her secure a scholarship to study at Oxford University. During her PhD work at MIT, Ghassemi discovered that biases in health data can hide in machine learning models, leading to performance gaps for certain demographics.
Her research has since focused on addressing these issues, including showing that models can recognize patient race from medical images and that optimizing models for overall average performance can exacerbate fairness gaps. Key collaborators include Leo Celi, a principal research scientist at MIT’s Laboratory for Computational Physiology.
Improving Health with Robust Machine Learning Systems
Marzyeh Ghassemi, an associate professor at MIT’s Department of Electrical Engineering and Computer Science and the Institute for Medical Engineering and Science (IMES), has dedicated her research to developing robust machine learning models that can improve healthcare outcomes. Her work focuses on addressing biases in health data and ensuring that machine learning models are fair and effective across different patient demographics.
Ghassemi’s journey into this field began during her PhD, when she encountered the issue of biases in health data hiding in machine learning models. She had trained models to predict outcomes using health data, but upon checking their performance on patients of different genders, insurance types, and self-reported races, she found significant gaps. This discovery sparked a decade-long exploration into addressing these issues.
One of Ghassemi’s most notable breakthroughs came when her research group showed that learning models could recognize a patient’s race from medical images like chest X-rays, which radiologists are unable to do. They then found that models optimized to perform well “on average” did not perform as well for women and minorities. By combining these findings, they demonstrated that the more a model learned to predict a patient’s race or gender from a medical image, the worse its performance gap would be for subgroups in those demographics.
Ghassemi’s work emphasizes the importance of training models to account for demographic differences instead of focusing on overall average performance. This approach can mitigate the problem, but it requires careful consideration and implementation at every site where a model is deployed. Her research highlights the need for developers and deployers of machine learning models to prioritize fairness and robustness in their designs.
The Impact of Identity on Research Interests
Ghassemi’s identity as a visibly Muslim woman and mother has significantly shaped her research interests. She notes that her experiences have informed her perspective on the world, which in turn influences her approach to addressing biases in machine learning models. Her work is a testament to the importance of diversity in STEM fields, where diverse perspectives can lead to more comprehensive and effective solutions.
Balancing Research with Life’s Bigger Picture
Ghassemi is passionate about her work, but she also recognizes the importance of maintaining a balance between research and personal life. She intentionally cultivates interests beyond her technical expertise and prioritizes relationships with family, friends, and colleagues who encourage her to be a full person.
In her words, “When you love your research, it can be hard to stop that from becoming your identity — it’s something that I think a lot of academics have to be aware of.” By keeping track of life’s bigger picture, Ghassemi is able to approach her work with a sense of purpose and perspective.
A Journey of Self-Discovery
Ghassemi’s philosophy on life is reflected in her quote from the Persian poet Rumi: “You are what you are looking for.” She believes that at every stage of life, one must reinvest in finding who they are and nudging that towards who they want to be. This mindset has guided her journey as a researcher, mother, and individual, and serves as a reminder that personal growth and self-awareness are essential components of a fulfilling life.
External Link: Click Here For More
