Of SWFWAPs and data privacy

The Register reports that a patent filed by Amazon.com ("Providing gift clustering functionality to assist a user in ordering multiple items for a recipient") was published last week. Its title sounds innocuous and even laudable (I frequently need assistance in ordering multiple clustered gift items for a recipient), but there's a line in it to the effect that a database covered under the patent might contain information on "education levels, genders, income levels, interests, races, ethnicities, religions, occupations, sexual orientations." Most of this information, needless to say, could not be inferred reliably from product purchases and would have to come from external sources. Amazon has said that it has no "immediate intention" of creating such a database. Let's pretend for a moment that it has an...eventual intention.

So what?

Data mining (or "analytics," as it's often called, since the term "data mining" lost some luster in the '90s) can certainly throw off interesting and exploitable patterns--just ask the fraud detection teams at credit card companies--so it's likely that Amazon could increase sales by knowing (quite a lot) more about its customer base. It's not clear that this would be such a good idea, however. Amazon would run the risk of spooking customers by offering product suggestions that are just a bit too well-targeted ("What!? How the #$%& do they know I'm a Single White Female Wiccan Anthropology Professor (SWFWAP) with Bunions!?"). It would have to be very careful when deciding how it uses the data it possesses. There's also a more general backlash risk: Data of this kind has been collected for years, but largely out of sight of the general public. If Amazon's activities made it obvious that such data exists and was being used, the reaction could reach all the way to the halls of Congress--which might severely restrict data collection and use. Amazon should probably be very careful indeed.