The Census Bureau and many large Internet companies like Google and Apple have
adopted formal models for privacy protection in their publications and research. While
the technical details of these models are fascinating, their implications for how we live
and work in a world where vast amounts of data are processed daily to learn about us
are far more important. Privacy protection and statistical accuracy are competing uses of
the same good: the information in confidential databases. Engineers design the
algorithms for implementing the analysis with privacy protection. But society needs to
decide which statistics need to be more accurate than other statistics. That's not an
engineering decision, that's a public policy decision. I will talk about how to think about
that public policy.