Key Concepts

• The idea of linking together databases, known as data fusion, is the bête noire of privacy advocates. So far, however, it seems to be limited to specific contexts, such as gambling casinos and child-support enforcement.
• Data fusion is challenging because databases are riddled with errors and meaningless coincidences. New algorithms overcome some of these hurdles, but do they shift the overall ratio of cost and benefit?

A  few years ago I bought a latte at Starbucks on the way to the airport, parked my car and got on a flight for the U.K. Eight hours later I got off at Heathrow, bought a prepay chip for my cell phone and went to buy a ticket for the train into London, when my credit card gave up the ghost and refused to work anymore. Not until I got back to the U.S. did I find out what had happened. Apparently, the small purchase at Starbucks, followed by the overseas purchase of the cell phone card, had tripped some kind of antifraud data-mining algorithm in my credit-card company’s computer. It tried to call me, got my voice mail and proceeded to blacklist my credit card.

What I found so exasperating about the entire experience was that the computer should have known that the person using my card in England was me. After all, I had bought my plane ticket with that same card and had flown with a major U.S. carrier. Aren’t all those databases supposed to be tied together?

   
Already a Digital subscriber? Sign-in Now
If your institution has site license access, enter here.