Exploring how randomized response can help collect sensitive information responsibly
AI Explorables is an ongoing series of interactive essays that walk through important concepts in machine learning. For the full interactive experience, check it out here.
The availability of giant datasets and modern computing power is making it harder to safely collect and study information about sensitive topics. Demographic data from the census, for example, is critical to understanding our society but people won’t respond truthfully unless they’re sure their information will be protected.
Our latest AI Explorable dives into privacy. Ellen and I take a toy example of private data and used it to show how removing names from a dataset isn’t enough to securely anonymize it — and how applying randomization can be.
Within the Explorable, you can tweak the randomization process to see the tradeoff between accuracy and privacy for yourself.
