I wrote this comment before I made it to the end of what you wrote about LinkedIn endorsements. The implementation is still up for debate, but the idea is compelling at the very least. I do think that LinkedIn endorsements are worthless because I haven't yet used them, but that idea distills a person down to their core. We're basically trying to solve that problem through the single question that gets more responses than any other on our site: "describe your friend using only three words."
Being able to find a set of N people who are into ancap/lib ideas if not politics, blinded tokens, HSMs, and are within a 100 mi radius of X would be interesting.
So the natural consequence is assuming the absence of a verified fact for "is under 300 pounds" is "is over 300 pounds" with equally high certainty, if you have a well structured system which covers everything.
My theory is the problem of people lying on their profiles will go away when their friends are using the site with them.
e.g. for a dating site you'd have "less than 300 pounds", "graduated from Stanford", "not married", "hivfree"
for an HR site you'd have "not a CISSP", "graduated from Caltech in 2000 in CS", "US Citizen", "worked for Google 2000-2004", etc.
Individual tags would be searchable, high confidence, and ideally structured (so "College Graduate", "CS", "MIT", "MIT SB", "MIT 2000", "MIT SB 2000", "MIT SB 6 2000" would all be known.
(The stupid way LinkedIn does Endorsements isn't it, I think. What my friends click just to get a stupid box to go away isn't high assurance information. I'm not sure if crowdsourcing has any place, but the ideal is to get actual validation from the data owner.)