Thumbprint scanners are inaccurate unless you get into the high-end range. If there is a commodity fingerprint scanner for that many students, they could install a single button and get equivalent data.
BUT if we had consistent poor performance for some vendor in a certain category, we could infer that their offerings for that category will also be poor in the future.
It's not as extreme as that--consider just the case where you want to share things with work friends that might not appeal to college friends. You're not acting different; you'd show the college friends if any of them asked, but you know they're bored by it so you want to keep it out of the timeline/tweets/whatev that they see.
I saw an on-going project recently that was attempting just that. They were using texts from the web to infer relationships from recognizable patterns. Like (the head of X, Y) implies X is an organization, Y is a person, and Y is the leader of X.
It would also try to learn new patterns from facts it already knew, like if it notices Jobs and Apple in a sentence it can hypothesize that sentence pattern is about a leader of a company.
For the life of me I cannot remember the name of the project or even the university, hopefully someone else has heard of this.
I get poor search results daily, and half of them get "incorrected". "Did you mean UIView?" No, Google, I am working on OSX. I actually did mean NSView, that's an actual thing.
There's explicit provision in the law allowing for people without a pilot's license to fly an experimental aircraft. I'm sure you can still get in massive trouble for flying through the upside-down layercake above airports, but it's not flat-out forbidden everywhere.
Yeah, the author seemed to argue "because the situation I was in while doing a job I was passionate about was unlivable, therefore pursuing any job you are passionate about will make you less happy than working some regular 9-to-5." Logical fallacy.
The bloom filter could be used in a similar way, you have X hash functions used and Y total bits, so for a bloom filter with Z bits set you anticipate N unique items were hashed. Probably not as good as the HyperLogLog in terms of memory space / % error, but there you go.