Black-box medicine—the use of big data and sophisticated machine learning techniques for health-care applications—could be the future of personalized medicine. Black-box medicine promises to make it easier to diagnose rare diseases and conditions, identify the most promising treatments, and allocate scarce resources among different patients. But to succeed, it must overcome two separate, but related, problems: patient privacy and algorithmic accountability. Privacy is a problem because researchers need access to huge amounts of patient health information to generate useful medical predictions. And accountability is a problem because black-box algorithms must be verified by outsiders to ensure they are accurate and unbiased, but this means giving outsiders access to this health information.
This article examines the tension between the twin goals of privacy and accountability and develops a framework for balancing that tension. It proposes three pillars for an effective system of privacy-preserving accountability: substantive limitations on the collection, use, and disclosure of patient information; independent gatekeepers regulating information sharing between those developing and verifying black-box algorithms; and information-security requirements to prevent unintentional disclosures of patient information. The article examines and draws on a similar debate in the field of clinical trials, where disclosing information from past trials can lead to new treatments but also threatens patient privacy.
