Pages: pp. 2-3
Last issue, my letter focused on some trends in academic publishing that journals, magazines, and other scientific-publishing endeavors are facing. I argued that we computer scientists should take a leading role in helping create technologies that will break down the walls between different disciplines.
An obvious response is, "haven't we already?" That is, Tim Berners-Lee's invention of the World Wide Web, coupled with Sergey Brin and Lawrence Page's PageRank algorithm that powers Google, has led to a world where scientists publish preprints and papers on the Web and can find each others' results without having to subscribe to the same journals. Doesn't that solve the problem? Just pick your favorite chemical, check it out on Google (or even better, Google Scholar), and there's the paper you need. What more is there to say?
Unfortunately, the reality of sharing papers on the Web doesn't live up to this ideal. With rare exceptions, Web-based journals and open source publications and preprint servers have been modeled on the same field and subfield considerations that print journals and scientific communication have been using for decades. (For example, the Journal of AI Research, one of the best online journals in terms of impact and reputation, isn't organized much differently from the traditional print journal Artificial Intelligence.) So while we've made journals electronic, with positive results on distribution, we still haven't really done much to revolutionize the scientific-communication aspects of scientific publication. If a computer scientist searches for some term in the field—say, "case-based reasoning"—in an online journal, he or she will likely find papers of interest. However, he or she won't likely find, say, a Science article written by a chemist who used a similar approach to do data provenance in a lab setting. Despite the prevalence of online literature and the ability to search with great ease, jargons—the different ways different fields describe similar things—still get in the way.
This is true even of sites that have tried hard to provide organizational structures to help people find each other's work. Scientific publication sites such as PubMed ( www.ncbi.nlm.nih.gov/sites/entrez) or arXiv ( http://arxiv.org), for example, organize papers on the basis of disciplinary terms familiar to the scientists who use them. One can find papers on similar topics via common Web mechanisms (hyperlinks to related papers, menus representing ontological categories, and so on) by browsing, thus compensating for some of the domain-specific aspects of search. However, these sites still tend to be of greatest use to those in a particular field, who know the terms and can navigate the terminologies. The significant effort put into these sites does help build bridges, especially between subdisciplines, but they still primarily use the same jargony terms used in searches.
This is particularly troubling because many, if not most, of the key scientific discoveries of the next generations will require interdisciplinary approaches. This becomes particularly clear when you talk to thought leaders in science or explore the sorts of problems described in the United Nations' Millennium Development Goals ( www.un.org/millenniumgoals), such as eradicating hunger and ensuring environmental sustainability. Funding agencies, although still organized along disciplinary lines, are putting more money up for projects that cross the traditional lines. Biotechnology centers are flourishing at many universities and encouraging biologists, computer scientists, and others to work together. New organizations created to develop technologies that span fields are seeing increasing growth in academia and industry. Unfortunately, despite the growth of these larger science projects, the publishing still tends to be in traditional journals, arranged along disciplinary lines and reviewed by experts in the traditional fields. So, large interdisciplinary projects tend to be published in bits and pieces, with each part published in an article that briefly reviews the overall project but primarily discusses the aspects of interest to that particular audience.
This need for better support for interdisciplinary work is exactly what's motivating many of the new solutions being considered online. Rather than simply replacing print journals with online journals of the same ilk, people are considering new models that can more easily cross disciplines and find papers (or projects and so on) related to their interest areas. One model that's becoming more common is the overlay journal (sometimes called a deconstructed journal, after John Smith's 1997 paper "The Deconstructed Journal"). The idea is to create an online publication by taking some crosscutting theme and providing links to papers published elsewhere. By providing links rather than republishing, the overlay journal provides a service to both the reader, by linking to many publications, and the publishers, by bringing more eyeballs to their sites.
The best of these journals provide organization based more on domain features than on disciplinary jargon. One well-respected overlay journal is the Virtual Journal of Quantum Information ( www.vjquantuminfo.org/quantuminfo). It provides links to papers appearing in many specialized disciplinary journals ( Low Temperature Physics, Optics Express, Science, Physical Review A, and so on), organized by themes relating to quantum processing such as "information theory" and "algorithms and computation." So, it uses the jargon of the one field (quantum theory) to help organize the materials appearing across a wide swath of other fields (physics, optics, mathematics, computer science, and so on).
Overlay journals are becoming more popular, and the Web's odd economy makes them very feasible. Basically, the profits from downloads, if they aren't free, go to the material's original publisher. The overlay journal makes its profit (or provides its service) through being a repackager—getting many eyeballs on the site and selling advertising, or promoting the new field. Authors get tremendous benefit because more people are directed to their papers (increasing citation counts and visibility). Readers get relevant material culled from a wider range of publications than they could monitor on their own. Win-win-win-win—how often do you see that?
A number of different models are being explored for how best to "curate" such overlay publications. One model is to have the overlay journal function as an independent publication. The IEEE Computer Society will be taking this approach with Computing Online, a site that will serve multiple purposes. One purpose is to provide an overlay functionality across the different Computer Society magazines (including IEEE Intelligent Systems, of course). The new publication will thus provide a means for a wider audience to see our magazine's articles, and we'll be providing articles to that site, reducing its need to solicit and produce new material. The curation will likely be through an editorial board, with liaisons to the magazines, thus giving Computing Online dedicated staff and the functions of a separate publication.
Another model has a more open Web feel to it. For example, the University of Southampton's Leslie Carr is creating a submitter-based Web science overlay journal. Submitters can fill in a simple form (or share metadata from other publishers) to have a paper listed on the site. The submitter
The submitter can be the paper's author or someone else who feels the paper will be of value to the community. Mechanisms for determining how to moderate the site (whether by having a process to screen the submitters or to review the submissions) are still being discussed. This mix of submission with overlay seems to combine the best of several worlds.
Oh dear, I've run out of space, and I was just getting warmed up. Beyond the overlay journal, we start to look at more novel ideas that depart further from traditional publishing. Such ideas include blog-based scientific publication, Web 2.0-based community sites, and sites enhanced by the Semantic "Web of data." More on this in part 3.
P.S. Last issue I called for creative ideas in the publishing domain. This issue's Expert Opinion (see p. 16) features a response by Josep Lluis de la Rosa and Boleslaw Szymanski. If you have ideas, please consider sending me a letter or an essay like theirs; we'll certainly publish the best of these.
IEEE Intelligent Systems is looking for the top 10 young AI researchers. To be eligible, candidates must have received their PhD in AI or a related field on 1 Jan. 2006 or later. To nominate someone, send a pointer to their Web page and a short (1–2 paragraphs maximum) description of their work and qualifications to firstname.lastname@example.org by 1 January 2008. For examples of the kinds of candidates we're looking for, see www.mindswap.org/blog/ieee-intelligent-systems-ten-to-watch for a list of the 2006 recipients, and www.computer.org/portal/cms_docs_intelligent/intelligent/content/x3005.pdf for the online version of the 2006 winners' statements.