January 25, 2004

PhDMeatMarket.com, or, Favoritism and Intrigue Mechanized

Warren Ellis is not impressed by Orkut:

Orkut, for those who haven't been online for the last few days, is a new Friend Of A Friend system which just began beta testing. The hype point is that it's from Google.

Right now, it looks pretty much like an iteration of the Tribe.net system, with an eye on Friendster's apparent main function as a dating system. (Which means, oddly, it requests your business profile at the same time as it's asking you where you like to be fingered.) (Okay, maybe not.) And, you know, it's a standard beta experience -- it's kind of ugly, the code chokes every now and then, it kicks you out of the system once or twice. Nothing egregious. It's coping pretty well as it starts taking the weight of several thousand early explorers. Most of whom, if they follow the accelerating process that's left Friendster a relative wasteland and given Tribe a bit of an echo, will be out of there again in a few weeks.

I'm happily married, and no more enterprising than the next sloth, so this kind of software is really of no use to me. But I am putting myself on the academic job-market, or what passes for one (a year before my post-doc ends, just to be safe). This means I've been spending my time writing research statements, teaching statements, cover letters, etc., and recruiting friends, collaborators and various patrons to write letters of recommendation. Really, however, hiring committees have two very simple questions about me: (1) Will he cover us in scientific glory? (2) Will he bring us mad cash from the NSF/NIH/DoE/DARPA? It is notoriously hard to estimate these variables on the basis of written statements, sample papers, letters of recommendation and impressions gleaned in interviews. (Scholars should know that interviews are useless, but we don't.) It seems to me that a well-designed social network database could simplify this process, as well as make it more objective.

The system I have in mind would have two types of nodes, one of authors and one for papers, and two types of links --- an "author-of" relationship, and a "cited-by" relationship. So, by entering a scientist's name, you could see what papers they've written, and so who they've written them with, and which papers cite their papers. So far this is just what many universities have, via their access to the Institute for Scientific Information's "Web of Science" database. PhDMeatMarket.com would, however, offer, and charge for, two additional services. (1) Calculation, using various standard metrics on graphs, of how important somebody is, either to science as a whole or some specified field. (2) Automatic suggestion of people competent to evaluate a candidate's work, by examination of patterns of co-citation and authority. (Update: Further network analysis would identify those too closely allied to the candidate to be objective.) Departments would then pay them to do so (implementing a suggestion of D. McCloskey's). A potential third service would allow people to add "acknowledgment" links to each other, reflecting the kind of discussion and evaluation which is often very helpful but not formal enough to show up in the bibliographic record. Update: For a fee, it would let you list your enemies, and exclude them from the list of those it recommends to evaluate you; for an additional fee, it would not display your enemies list.

This doesn't perfectly answer the questions about fame and fortune that hiring committees have, but it's pretty safe bet that somebody who's close to the center of a scientific network is going to go far, especially if their collaborators are also close to the center. (This follows from the pious observation that important papers will be read, built upon and cited, and should be held altogether free of sociological heresy.) And winning scientific glory does help secure funding, though not so reliably as it might.

Now not only do I have a friend who knows an awful lot about scientific collaboration networks, and another friend who started a dot-com, but PhDMeatMarket.com is currently untaken. (Update: Not any more.) If I had only had this idea three years ago...

Trackback: Johnny Logic

Networks; Learned Folly; Modest Proposals

Posted at January 25, 2004 20:00 | permanent link

Three-Toed Sloth