Pay per paper (P3)

(Originally published on Applied Mechanics News on 22 July 2006, where many comments provided remarkable insight)

I’ve just stopped subscribing to Science. The magazine is great, but few papers in it interest me. The signal-to-noise ratio of Science, I guess, is just too low to most individuals. Instead, I’ve now subscribed to the RSS feed of Science. If any paper looks interesting, I can access to the full paper online through Harvard Libraries. Outside my office, a color printer is free to use for everyone. A library of an institution seems to be an ideal home for a journal like Science. Nearly every individual paper in Science is of high enough quality to appeal to someone in the institution.

Few journals can make that claim, however. Most journals are only relevant to several people in an institution. Furthermore, few researchers read any scholarly journal from cover to cover. Rather, we all read individual papers. However, libraries subscribe to journals, or even bundles of journals. As a result, the libraries pay for many papers that nobody reads, and miss other papers that someone would like to read.

This business model is bad for authors and readers, and possibly even bad for publishers. Technology now exists to distribute information far more efficiently, in a unit consistent with how people consume the information. For example, many people now prefer buying individual songs to albums. See a recent book, The Long Tail, by Chris Anderson, the editor-in-chief of Wired, for a remarkably perceptive analysis of media industries.

The same business model may apply to scholarly papers. One may argue that journals, like albums, were invented as a packaging technology to suit the old economics of delivery. As scholarly papers are all online, the name of a journal becomes simply a tag to the papers published in that journal. Maybe a powerful tag, but a tag nonetheless. So far as how papers should be distributed, the name of a journal should serve the same function as all other tag-like entities: keywords, names of authors, etc: the tags help readers to sort papers and set priorities. It makes no sense for anyone to insist that papers with any particular tag be delivered as a bundle.

Many publishers already offer individual papers for sale online; for example, the cost is at $30 per paper for many Elsevier journals. Once a reader buys a paper, it seems reasonable to share this paper with his close colleagues, and it also seems reasonable to store the paper for future use. Perhaps we can formalize this practice.

How about we treat a paper just like a book? With one click, a reader will have the paper, and his library will automatically pay for it. Once bought, the paper is accessible to every user of the library. We can also collect statistics. If the users of a library buy many papers in a journal, the library should subscribe to the journal. Libraries will set up an algorithm to minimize the total cost. Publishers will set up their algorithms to maximize profits. However, libraries and publishers do have a common ground: they both want to help people to find papers.

To support such a business model, a third party may provide a web service. It seems to be too wasteful to make every individual library and every individual publisher maintain a separate web service. Something like Amazon.com or Last.fm for papers might do. The service can also be an extension of services like EZproxy or CiteULike.