3 ms·
Even when data is "anonymized", it's sometimes possible to recover identifying information anyway. For example, knowing a person's age, zip code, and gender is
by pmoriarty 9y ago
Even when data is "anonymized", it's sometimes possible to recover identifying information anyway. For example, knowing a person's age, zip code, and gender is often enough to identify people, even if their name is not explicitly part of the available data.
That's just a really simple example. I'm sure you could do much better with advanced AI techniques and more "anonymized" data.
- icebraining 9y agoIn fact, this was demonstrated on Netflix's own dataset: https://arxiv.org/abs/cs/0610105 https://arxiv.org/abs/cs/0610105
- arkh 9y agoI think one of the best and chilling example is still: https://en.wikipedia.org/wiki/AOL_search_data_leak https://en.wikipedia.org/wiki/AOL_search_data_leak Just some anonymous search strings can give you people. And that was in 2006 so nowadays with the quantity of informations people make public you'd get a lot more results.
- remarkEon 9y agoIn college (years ago...mid 2010s) I did a panel data regression analysis about certain types of behavior in the dorms. I conducted an anonymized survey, that asked a bunch of demographic and behavioral questions as well as ones about extra curriculars across several years. I was working on a “willingness to take risk” study. When cleaning the data (I allowed too much human entry in the first iteration of the survey) I could read the demo data and see the other activity responses and make quite confident inferences about who some of the respondents were. This Netflix thing is no surprise to me, unfortunately. I sort of feel like there should be some kind of code of ethics for Data Scientists, or maybe anyone that uses econometrics for making inferences about human behavior.