Discussion about this post

User's avatar
The AI Architect's avatar

Excellent deep dive on the small-town reidentificaton risk. The math here is crucial—standard deidentification breaks down when a "35-year-old female with Type 2 diabetes" becomes uniquely identifiable in a Vermont town of 3,600. I've worked on healthcare datasets where similar assumptions failed once you layered in even basic demographics. The 98.9% opt-in rate feels less like consent and more like inertia, especially when most pateints probably dunno the data flows to a research enclave or gets used for commercial analytics downstream.

Expand full comment

No posts

Ready for more?