The Georgetown University Law Center hosted a roundtable discussion exploring the ethics, law, and policy of data re-use in a machine learning age. Georgetown Law Professor Paul Ohm led the discussion, which included representatives from academia, computer science, government agencies, and private companies, including Georgetown’s Institute for Technology Law and Policy and Center on Privacy and Technology, the KIE, and the Center on Law and Information Policy at Fordham Law School. The KIE was a co-sponsor of the event, which KIE Director Maggie Little and postdoctoral fellow Elizabeth Edenberg attended. The roundtable is supported by grants from AXA and the Sloan Foundation.
The day’s focus was on the reuse of data in machine learning that goes beyond the initial purposes outlined in the consent procedures when collecting that data. The workshop was designed to guide the conversation and help distill points of consensus and contention on the future of data privacy in an era of big data and machine learning. Conversations emphasized two competing moral imperatives:
- – Protecting the individual data subject’s privacy and respecting initial terms of consent.
- – The potential benefits for society of new knowledge that could be gained by using machine learning to reanalyze the data to try to solve difficult social problems.
“The day was spectacularly stimulating,” KIE Director Maggie Little said. “There are amazing troves of data that carry great potential for advancing our knowledge in health, security, and other areas to benefit society, but it also carries challenging questions.”
The workshop sought to address the following questions:
- – To what extent does current law around data re-use serve the mutual goals of protecting privacy and using data to help solve societal problems?
- – Does the advent of machine learning require a rethinking of current law?
- – To what extent are the Fair Information Practices (FIPs) useful for the machine learning age?
- – Are there changes in technology, law, or practice that should be considered to increase the utility of data and/or decrease the privacy concerns? Are there changes that are not zero-sum?
“The machine learning aspect is fascinating,” Little continued. “The ability of specific systems to find patterns in vast quantities of data is revolutionizing the world, and while machine learning systems are finding heretofore unseen patterns in troves of data is wonderful, it is also dangerous. When policy is based on these correlations, it can have unintentional consequences, such as entire systems profiling by race… Magic can happen when you bring together advocacy leaders, academics, and the technical experts… it certainly happened here.”
The discussion was held “off-the-record” and conducted under Chatham House rules. The team will produce a white paper following the event for public dissemination.