Science cannot be divorced from other kinds of social issues. If you care about science — not just because you’re curious, but also because you want to make an impact — you must think about the larger social context.
At Morgridge, we’re working to create an environment where people think and talk about responsible science in their day-to-day scientific lives, not just when problems arise.
I’m directing more research toward data-related issues, especially the use of big data and machine learning in healthcare. There is well-justified enthusiasm, but also reason for caution. We need to implement data-driven health studies in ways that do no exacerbate current social inequalities.
Consider, for example, creating an algorithm that finds risk factors for heart disease patients dying within five years of treatment. We would train this algorithm on a variety of datasets, including electronic health records, genomic information, survey results about diet and lifestyle, even exercise data from Fitbits.
But problems may lurk in the information itself. These datasets are not precisely curated for big studies. The algorithms learn on data from the real world, and the real world incorporates all kinds of socioeconomic biases. They’re essentially baked in.
Heart disease is a good example. We know from recent studies that women are diagnosed for heart disease much later than men. And they often don’t get the same treatment recommendations, even when health conditions are similar. Unfortunately, the data reflects these biases. The algorithm learns something we don’t want it to learn.
What can we do? We need to understand how generalizable big data studies are, and for whom they will be medically useful. We also need to build enough tools to help people from all demographics, so we create precision medicine that is relevant across race, gender and socioeconomic status.
This is just one example from this fresh and demanding field. I encounter and address all sorts of interesting issues, from the security implications of gene editing to privacy of genetic information. I have one of the greatest jobs in the world.