Author: Michaela Jarvis | Laboratory for Information and Decision Systems<\/p>\n
Captivated as a child by video games\u00a0and puzzles, Marzyeh Ghassemi was also fascinated at an early age in health. Luckily, she found a path where she could combine the two interests.\u00a0<\/p>\n
<\/a>\u201cAlthough I had considered a career in health care, the pull of computer science and engineering was stronger,\u201d says Ghassemi, an associate professor in MIT\u2019s Department of Electrical Engineering and Computer Science and the Institute for Medical Engineering and Science (IMES) and principal investigator at the Laboratory for Information and Decision Systems (LIDS). \u201cWhen I found that computer science broadly, and AI\/ML\u00a0specifically, could be applied to health care, it was a convergence\u00a0of interests.\u201d<\/p>\n Today, Ghassemi and her Healthy ML research group at LIDS work on the deep study of how machine learning (ML) can be made more robust, and be subsequently applied to improve safety and equity in health.<\/p>\n Growing up in Texas and New Mexico in an engineering-oriented Iranian-American family, Ghassemi had role models to follow into a STEM career. While she loved puzzle-based video games \u2014 \u201cSolving puzzles to unlock other levels or progress further was a very attractive challenge\u201d \u2014 her mother also engaged her in\u00a0more advanced math early on, enticing her toward\u00a0seeing math as more than arithmetic.<\/p>\n \u201cAdding or\u00a0multiplying are basic skills emphasized for good reason, but the focus can obscure the idea that much of higher-level math and science are more about logic and puzzles,\u201d Ghassemi says. \u201cBecause of my mom\u2019s encouragement, I knew there were fun things ahead.\u201d<\/p>\n Ghassemi says that in addition to her mother, many others supported her intellectual development. As she earned her undergraduate degree at New Mexico State University, the director of the Honors College and a former Marshall Scholar \u2014 Jason Ackelson, now a senior advisor to the U.S. Department of Homeland Security \u2014 helped her to apply for a Marshall Scholarship that took her to Oxford University, where she earned a master\u2019s degree in 2011 and first became interested in the new and rapidly evolving field of machine learning. During her PhD work at MIT, Ghassemi says she received support \u201cfrom professors and peers alike,\u201d adding, \u201cThat environment of openness and acceptance is something I try to replicate for my students.\u201d<\/p>\n While working on her PhD, Ghassemi also encountered her first clue that biases in health data can\u00a0hide in machine learning models.<\/p>\n She had trained models to predict outcomes using health data, \u201cand the mindset at the time was to use all available data.\u00a0In neural networks for images, we had seen that the right features would be\u00a0learned for good performance, eliminating the need to hand-engineer specific features.\u201d<\/p>\n During a meeting with Leo Celi, principal research scientist at the MIT Laboratory for Computational Physiology and IMES and a member of Ghassemi\u2019s thesis committee, Celi asked if Ghassemi had checked how well the models performed on patients of different genders, insurance types, and self-reported races.<\/p>\n Ghassemi did check, and\u00a0there were gaps. \u201cWe now have almost a decade of work showing\u00a0that these\u00a0model gaps are hard to address \u2014 they stem from existing biases in health data and default technical practices. Unless you think carefully about them, models will naively reproduce and extend biases,\u201d she says.<\/p>\n Ghassemi has been exploring such issues ever since.<\/p>\n Her favorite breakthrough in the work she has done came about in several parts. First, she and her research group showed that learning models could recognize a patient\u2019s race from medical images like chest X-rays, which radiologists are unable to do. The group then found that models optimized to perform well \u201con average\u201d did not perform as well for women and minorities.\u00a0This past summer, her group combined\u00a0these findings\u00a0to\u00a0show that\u00a0the more a model learned to predict a patient\u2019s race\u00a0or gender from a medical image,\u00a0the worse its performance gap would be for subgroups in those demographics. Ghassemi and her team found that the problem could be mitigated if a model was trained to account for demographic differences, instead of being focused on overall average performance \u2014 but this process has to be performed at every site where a model is deployed.<\/p>\n \u201cWe are emphasizing that\u00a0models trained to optimize performance (balancing overall performance with lowest fairness gap) in one hospital setting are not optimal in other settings. This\u00a0has an important impact on how models are developed for human use,\u201d Ghassemi says. \u201cOne hospital might have the resources to train a model, and then be able to demonstrate that it performs well, possibly even with specific fairness constraints. However, our research shows that these performance guarantees do not hold in new settings. A model that is well-balanced in one site may not function effectively in a different environment. This impacts the utility of models in practice, and it\u2019s essential that we work to address this issue\u00a0for those who develop and deploy models.\u201d<\/p>\n Ghassemi\u2019s work\u00a0is informed by her\u00a0identity.<\/p>\n \u201cI am a visibly Muslim woman and a mother \u2014 both have helped to shape how I see the world, which informs my research interests,\u201d she says. \u201cI work on the robustness of machine learning models, and how a lack of robustness can combine with existing biases. That interest is not a coincidence.\u201d<\/p>\n Regarding her thought process, Ghassemi says inspiration often strikes when she is\u00a0outdoors \u2014 bike-riding in New Mexico as an undergraduate, rowing at Oxford, running as a PhD student at MIT, and these days walking by the\u00a0Cambridge Esplanade. She also says she has found it helpful when approaching a complicated problem to think about the parts of the larger problem and try to understand how her assumptions about each part might be incorrect.<\/p>\n \u201cIn my experience, the most limiting factor for new solutions\u00a0is what you think you know,\u201d she says. \u201cSometimes it\u2019s hard to get past your own (partial) knowledge about something until you dig really deeply into a model, system, etc., and realize that you\u00a0didn\u2019t understand a subpart correctly or fully.\u201d<\/p>\n As passionate as Ghassemi is about her work, she intentionally keeps track of life\u2019s bigger picture.<\/p>\n \u201cWhen you love your research, it can be hard to stop that from becoming your identity \u2014 it\u2019s something that I think a lot of academics have to be aware of,\u201d she says. \u201cI try to make sure that I have interests (and knowledge) beyond\u00a0my own technical expertise.<\/p>\n \u201cOne of the best ways to help prioritize a balance is with good people. If you have family, friends, or colleagues who encourage you to be a full person, hold on to them!\u201d<\/p>\n Having won many awards and much recognition for the work that encompasses two early passions \u2014 computer science and health \u2014 Ghassemi professes a faith in seeing life as a journey.<\/p>\n \u201cThere\u2019s a quote by the Persian poet Rumi that is translated as, \u2018You are what you are looking for,\u2019\u201d she says. \u201cAt every stage of your life, you have to reinvest in finding who you are, and nudging that towards who you want to be.\u201d<\/p>\n<\/div>\n Go to Source<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"