Saturday, July 31, 2004

In my last post, I would have liked to have given you all a link to the full text of the article by Just et al. from the August edition of Brain[...]But instead I just linked to the abstract, because unless you have a subscription to Brain, the full-text link would have taken you to an Oxford University Press screen informing you that "You may access this article (from the computer you are currently using) for 24 hours for US$27.50", and I suspect that only a few of you care enough about the topic to pay that much for that little. [Language Log] Mark discusses tentative steps toward open access, and increased government pressure for open access to the results of government-funded research, and mentions "significant economic issues" on the way to the open access ideal. When I read discussions of open access for the biomedical literature, I'm always struck by how both pro and con sides accept as a given a very high editorial overhead. In my recent experience as an author of accepted papers in biomedical journals, both closed and open access, I didn't notice that the editorial services were substantially better than for the open access, no-author-fees of the Journal of Machine Learning Research (JMLR), which in a few years has become of the highest impact journals in artificial intelligence. JMLR is hosted at MIT, and for the most part managed by volunteers. Maybe Brain with its much higher volume would need more professional support, but the discussion always misses the fact that modern Web-based journal workflow systems make the editorial job way easier than it used to be. Details aside, the main point is that information technology can make the cost of running a journal much lower than it ever was, and tilt the discussion away from the current reader-pays vs. author-pays food fight. Peer-reviewed publication is an important element in the fabric of scientific discovery and communication (but not the only possible model, as arXiv demonstrated), and support for publication infrastructure can and should be folded into other research overheads that we already accept (labs, libraries, offices, networks, computers, travel, group Web sites,...). JMLR demonstrates this is possible. If all research universities got seriously into this using best organizational and technological practices, the need for explicit payment models (reader or author) might decrease significantly. 3:55:52 PM