Oh, that silly Amazon! It gives me oddball recommendations for products that don’t fit me because I ordered or browsed to them out of context. Therefore, it doesn’t know how to use the data it collects.

At least, that’s the thesis of this entry from The Buttry Diary, the blog of LSU student media director Steve Buttry. Buttry complains that, because he bought a book by his grandmother for the purpose of having it made into a purse for her, it now keeps recommending other Christian romance novels to him. Because he once screen-capped the front cover of a Pete Rose book for an unrelated article, he now keeps getting ads for it on Facebook. The thing is, I think his complaint is a little overblown.

I have little doubt we all have stories of that kind of incident in our lives. For myself, when I go to Amazon’s front page, the first row of images has various WiFi routers in it, because I searched on the Crystalview range-extender router for my recent Karma Go article. Further down the page, the “Inspired By Your Shopping Trends” row has a whole bunch of D&D books in it simply because Dungeon Master for Dummies is in the saved items in my shopping cart.

But I already own a Crystalview router (in fact, it’s sitting on the corner of my desk just a foot from my keyboard right now) and I haven’t thought about that D&D book in months. (Not that I was ever seriously interested in it, but it wound up in my saved items just as a way of setting it aside for later consideration.) Does that mean Amazon’s wrong to show me those things?

No, not really. I was interested in them enough to search on them. Amazon doesn’t know why, but then it doesn’t have to know why. It’s enough for it to know I was, and might be again. Not everyone who was will be again, but enough might be that Amazon deems it worthwhile to show them to me just in case.

When you get right down to it, this sort of advertising false positive isn’t exactly harmful, and it’s by and large the exception, not the rule. Amazon can’t tell why people buy or browse to something; it can only tell that they do. And most people who do are interested in those things, so Amazon does better by showing related ads to everyone who does it because it has no way of knowing that you’re not. That’s not a case of Amazon not knowing how to use the data it collects; it’s a case of it making the most efficient overall use of it, even if it proves unwarranted in a handful of cases.

The funny thing to me is that Buttry is complaining that Amazon collects this data without knowing what it all means, and effectively its recommendation engine just throws a bunch of stuff at the wall to see if anything sticks. But has he thought about what the alternative would mean? Does he really want Amazon (or anybody else) to know enough about us to be able to make accurate predictions?

Indeed, I ran across an article about that very kind of thing just tonight on Ars Technica. I didn’t think I could make it into a whole article in and of itself, but it provides an excellent example of data-gathering gone crazy. The article discusses advertisers using inaudible ultrasonic pitches to link multiple mobile devices together and build a much better picture of people’s day-to-day activities. Might Amazon eventually start doing that sort of thing?

At the moment, it would probably be counterproductive for Amazon to try to figure out why people do what they do—it would take more effort for them to gather sufficient data for that kind of analysis than it would probably be worth. Would Amazon find it worthwhile to run a detailed psychological work-up on you for the purpose of selling you a $20 book? Given the amount of trouble and expense it would take to do that for every consumer, probably not. The danger is that it’s getting ever cheaper and easier to do that kind of thing, and it’s not beyond the realm of possibility that it will be possible someday.

The bright side is, as long as people can point to inane examples of Amazon false advertising positives, that day has not yet arrived.

5 COMMENTS

The TeleRead community values your civil and thoughtful comments. We use a cache, so expect a delay. Problems? E-mail newteleread@gmail.com.