Friday, February 27, 2009

Almost Orwell: Tracking and Reccomendation Systems

We now live in a world where our preferences are tracked. Online the ads we see reflect what we have been viewing. Internet radio recommends new artists based our previous track selections. Every purchase we make from major online retailers results in more product selections that the retailer believes may interest us. All of these applications of data tracking and recommendation systems have been largely accepted as useful. However for some reason the public has been much less accepting of these technologies out side of cyberspace.

The majority of major super market chains now offer some sort of "Loyalty" program most require customers to sign up for a card that they use at check out to receive sale prices on products that they are buying. The corporations save information about the sale in massive databases. This information is later applied to customize marketing promotions to individual customers. at the time of the promotion manufacturers offer sales on certain products. Coupons for these products are then distributed to the consumers. Traditionally all consumers in the same region would get the same coupons (It wouldn't make sense to offer a sale on Ice Melter in Florida). With the databases of information gathered by the stores these marketing efforts are even more specialized. It is now entirely possible that two people living in the same house would receive entirely different promotions if they have distinctly different purchasing habits.

This big picture view of the system appears to be very practical and does not look like a source of controversy. How ever it has become one. The largest point of outrage comes from activist groups like CASPIAN (Consumers Against Supermarket Privacy Invasion and Numbering) who see this tracking as an invasion of privacy. Many would say that they are willing to give up a little privacy if it saves money. CASPIAN however argues that the coupons and sales are actually only offering the normal price and that the new prices are actually gross mark-ups. (

Now that it is clear where the negative effects of suggestion systems potentially come from I would propose that it is not a necessity that these effects be exploited.

We live in a world where experts are being replaced by Expert Systems. Our ability to number crunch is beginning to out weigh our natural ability to analyze. Sabermetrics is becoming the standard for prospect ranking in baseball. Computer algorithms are now more effective at predicting the best wine grape crops than Sommeliers. Computers may actually be able to tell us what we want more effectively than we can predict it our selves. For a long time the common model in expert systems was to teach the computer all of the rules that industry experts learned and then let the computer decide what people will think is best. As it turns out this does not work. The current model is to tell the computer what people think is best (for example through consumer trends) and then let the computer deduce the rules.

Now that we have enourmous databases to develop recommender systems from it is the responsibility of the coroporations to responsibly apply this information. Netflix is an example of a company that employs a positive model for reccomendation systems. Netflix does not make more money if you choose one movie over another, they simply make money as long as you continue to choose movies. This model means that it is in the best interest of netflix to continue suggesting movies that you will like and will rent. This model can easily be applied in any system. Wal-mart, often considered the worst corporation as far as consumer rights, could apply a simmilar model, if they were to track your purchaces and deduce that you enjoyed fishing, it may be in their best interest to offer you a sale on select fishing equipment. While it is in their best interest to charge you as much as possible they wont lose money on the sale items and it is likely that you wont leave the store with only items that are on sale. It is far more convenient for you to pay a little more for other products that you need in order to save your self from having to go to another store.

This model seems straight forward and seems to be common sense. However corporations will tend to exploit information if they have it. It is the responsibility of the consumer to ensure that these suggestion systems are used only for mutual benefit and not for customer exploitation.
Supercrunchers by Ian Ayres

1 comment:

  1. Interesting post. You may be interested in the work of the Electronic Frontier Foundation, which works to protect internet privacy. I think Raj Patel mentions them in his book, with respect to supermarket cards, too.