The Algorithmic Bias Initiative — founded in November 2019 at the University of Chicago — has brought together experts in healthcare and technology to address biases in medical artificial intelligence tools, the school said Nov. 3 in a post on its website.
Four details:
- The University of Chicago's Center for Applied Artificial Intelligence launched the initiative after its faculty director, Sendhil Mullainathan, PhD, published a study showing an algorithm used to evaluate millions of U.S. patients for enrollment in care management programs excluded many Black patients who should have qualified.
- The initiative is designed to work with health systems and other providers to help them evaluate their use of algorithms and check for biases. Its researchers found that when medical algorithms are biased, it's often because the algorithm's target variable is not the right fit for the decision it’s being used to make, the post said.
- In June, the initiative released an algorithmic bias playbook, a guide synthesizing its research and insights from working with healthcare providers. The playbook offers a four-step process to address algorithmic biases: Create an inventory of all the organization's algorithms, screen the algorithms for bias, retrain or suspend use of biased algorithms, and establish organizational structures to prevent future bias.
- The initiative will cosponsor a conference in the spring to bring together providers, policymakers, payers, AI software providers and technical experts from outside the healthcare industry. The panels will focus on topics such as data sharing and how to build an effective algorithmic management team.