[Data-modeling] Privacy, public figures, etc

[Kirrily Robert]

A while ago I signed up for online banking and, as online banking systems tend to do, it asked me not only for a password but for some additional question/answer pairs to help me sort things out if I lost or forgot that password.  I looked at the list of suggested questions, and saw things like:

- Mother's maiden name
- Your high school's mascot
- Town where you were born

It occurred to me that almost every item on that list of questions was something that you could find in Freebase, if a person's properties were filled in completely.  As you can imagine, this could be a bit scary if someone used the information in FB to login to your online banking.

When you sign up for Freebase you get a "user profile" and then there's a field to link it to a "Person topic about me".  That person topic then lets you fill in your mother's maiden name, your high school, the town where you were born, and so forth.  I suspect that some people do this a bit naively, hardly thinking about how it could be used for identity theft.  (None of *us*, of course!  Me, I filled in my mother's maiden name in the full and complete knowledge that the world can use it to access to my bank statements ;))

But now my mother and father have nodes in Freebase, and what's to stop someone coming along and filling in *their* mother's maiden names, high schools, etc?  Or for that matter, points that might be sensitive for various reasons, like "weight" or "religion"?

So there's been a bit of discussion around the place about changing the "Person" topic to be less privacy-invading, and moving some of the properties to a new type called "Public person", which we can use for well-known public figures and famous people whose privacy is, let's face it, already pretty well invaded.  That way we can record the weight of professional athletes or celebrities with eating disorders, or the religion of Presidents of the USA, or the genealogy of historical figures, without doing the same to ordinary people like you or me.

If we did this, we'd probably then go through and type anyone who has a Wikipedia article as a "public person" for starters.  If they meet Wikipedia's notability criteria, then their birthdates and so forth are probably public knowledge anyway.  Also, we'd need to write a FAQ or guideline somewhere about what makes someone a public person, and when you should (or shouldn't) apply that type.

Anyone got any thoughts on this?

K.


-- 
Kirrily Robert
Freebase Community Director
kirrily at metaweb.com