The Other Facebook Facial Recognition Problem
As I’m sure you are aware, your friends at Facebook rolled out a new “feature” in the last few days that has privacy experts more than a little concerned. Supposedly the new facial recognition software they have deployed will recognize your face when friends post pictures of you (and vice versa) so that all your other friends can see who you’ve been hanging out with.
Privacy experts were surprised (I don’t know why…Facebook has a history of being oblivious to privacy concerns) when the new software was automatically implemented as a “default on” option. In other words, if you didn’t want Facebook to do this for you (or to you), you have to go into your profile settings to turn it off. Apparently, the new feature has European regulators so annoyed that they are investigating (seriously, do these people ever get around to taking action other than “investigating?).
Today I noticed an article published by PCWorld, where the authors actually investigated the facial recognition feature to see how good it is. And, guess what: the facial recognition software doesn’t work so well. The author’s conclusion: stop worrying about your privacy…the feature doesn’t work well enough to warrant concern.
Hmm. The feature doesn’t work. Facebook really doesn’t do a good job of identifying my face on a friend's picture. Excuse me if I don’t breathe a sigh of relief. Let me explain why:
Apparently it works sometimes…and that’s what concerns me. What if it identifies me as a face in a photo and that face doesn’t belong to me? I’m far less concerned that the feature identifies me in things I’m involved with; I’m much more concerned with the potential for false positives!
If you and one of your Facebook friends look somewhat alike, the feature can identify you as being involved in lots of things you didn’t do. No, really, that isn’t me dressed in a bedsheet helping those people load a cow in the back of a van (though it sounds like an interesting evening). No, that wasn’t me drinking beer at the racetrack on the day I called in sick to work. No, that wasn’t me at that party making out with that woman who isn’t my wife. And so on.
Even if the facial recognition software worked perfectly so that there were never any false positives, what’s to keep a vindictive acquaintance from pasting my face onto an incriminating photo and posting it? (This is why, just to be on the safe side, I always buy treats for the Creative Services team at work: they know how to use Photoshop.)
Facebook seems to have some interesting ideas about how “privacy” should work. It remains to be seen if this will be the wake-up call they seem to so desperately need to be more privacy conscious, or if it turns out to be just one more kerfuffle over an ill-timed and badly implemented feature.
And seriously people: if you work with creative professionals that use Photoshop for a living…buy them cookies or candy once in a while.
Dave Higgins has been a student of systems development and improvement methods since 1975. Together with Ken Orr and the late Jean-Dominique Warnier, Dave was one of the principal architects of the Data Structured Software Development (DSSD) methodology--more widely known as the Warnier/Orr approach--that was widely used in the late 1970's and early 1980's. In his capacity as a software ...
Social Media Today