5 ms·
Anonymizing can be implemented by simply randomizing usernames, but in a way that keeps users uniquely identified in a scope of a single discussion. I think yo
by patio11 17y ago
Anonymizing can be implemented by simply randomizing usernames, but in a way that keeps users uniquely identified in a scope of a single discussion.
I think you underestimate the ability of people to identify others based on what English teachers call their "voice". I mean, here's a test you can do in forty-five seconds in Ruby or your language of choice: grab the RSS feeds from searchyc.com for pg and any other user you want to play the game with, have your script randomly spit out a recent post from one of the two lists, see if you can guess who said it.
This turns out to be really, really easy, particularly for good writers who one spends a lot of time reading. (You can do this with Bayesian classification, too. In the "Pick who wrote this out of two possibilities for which there are many samples" case it's like taking candy from a baby in terms of classification tasks.)
Also, do you think you or anyone else is going to have much trouble telling who baby_rutabega_spoon is when he talks about his bingo business or ruthless_carrot_fork when he gets exasperated about password hashing?
- tptacek 17y agoMy carrot fork is anything but ruthless, Patrick!
- jacquesm 17y agoHere I was thinking he meant Colin :)
- poronski 17y agoAlso, do you think you or anyone else is going to have much trouble telling who baby_rutabega_spoon is when he talks about his bingo business or ruthless_carrot_fork when he gets exasperated about password hashing? These are exceptions. I can bet that a vast majority of people here would not be able to tell even pg apart from other posters if it were not for his username. Besides these exceptions are fine, the anonymization is not meant to target them, but rather generic blabber-mouths whose comments are floating up only because of the group upvoting.