A great piece by Michael Eisen at his genomics blog on a science textbook whose price on Amazon marketplace ran up to $23,698,655.93 (plus $3.99 shipping). How did this happen?
The story is actually straightforward and has to do with algorithmic pricing employed by used booksellers. One company, profnath, had set their price of the textbook (on flies incidentally) to be 0.9983 times the price of another company, in this case bordeebook. For their part, bordeebook had set their price at 1.270589 the price of profnath. Both had forgotten to input a maximum price in their algorithm. So every time bordeebook raised the price slightly, a few hours later profnath would too, coming in just a bit under their price, leading bordeebook to raise their price slightly, coming in just above profnath’s price. And so on until was born the $23 million book.
It’s an important story not just because it shows what can go wrong with algorithmic pricing. Rather, it shows the more startling fact that algorithmic pricing itself has entered the online book market. Computation is not limited to making recommendations about what to read anymore. It extends to how much those recommendations will cost. You can see the conflict of interest there.
But it’s actually part of a bigger story of the automation of reading more generally. Prices and recommendations are part of a bigger puzzle of how machines increasingly do our reading for us today — or rather, the way we need to distinguish between machine reading and human reading (also known as display view). Even when a text is not on “display” in Google parlance, it is still being read and that reading is creating value for someone.
Another example that a student of mine brought to my attention is the case of booksellers who specialize in reprints of copies from Google Books. This too is almost entirely automated: they massively download files, use the titles generated by Google Books (often erroneous), list them on Amazon, and then print on demand a copy when you buy it. I’ve done this for a few out of print works that I’ve been interested in and I warn people against using such services (presses like Nabu Press, General Books, Kessinger Reprints, etc.). Not only are their prices heftier than an Espresso Book Machine would be (costs are usually around $35 per book). But because everything’s automated there is no quality control (much like Google’s scans). You will receive copies with plenty of missing pages or mis-scanned pages or hands in pages or incomplete editions.
Just the tip of the iceberg of how we need new ways of making sense of our cohabitation with machine readers.