In academic publishing, the “least publishable unit,” or LPU, refers to the smallest contribution to scientific knowledge that can be successfully sneaked past the reviewers at a conference or journal. Most often the term is used derisively, though I have sometimes heard it used as a modest complement, indicating that a researcher has astutely determined what can or cannot be published, and is extracting the maximum benefit from his or her hard work.
So far in 2010 I’ve participated in the program committee meetings for EuroSys and the USENIX Annual Technical Conference. Both meetings featured a lot of smart people evaluating a lot of very good research in computer systems. However, at both meetings I noticed something interesting: for a significant fraction of the strong submissions (those that were real candidates for acceptance), one or more reviewers had concerns that the paper was too incremental compared to previous papers by the same authors. The thing that struck me was that the authors in question were, more often than not, up-and-coming young researchers at respectable institutions and with strong pedigrees. Another thing that struck me is that I almost never noticed these problems myself. Rather, I tend to take papers at face value, trusting that the material which looks new actually is new. Therefore, at least three times at each of these two meetings, I found myself in the awkward position of supporting a paper that I had liked while others were attacking it as less-than-LPU. This isn’t fun and it looks like I’ll have to change the way I do reviewing by looking much more critically at people’s previous work, as opposed to just trusting them. I believe that the reason this problem showed up so strongly at USENIX ATC and EuroSys is that neither conference used double-blind submission, as is becoming very common. I’m pretty sure that double-blind makes it a lot easier to get an LPU though a program committee.
I am annoyed. The program committee shouldn’t have to be policing people’s publication strategies, it’s hard enough to simply evaluate the quality of the work. Obviously people are under tremendous pressure to publish, but surely this doesn’t legitimize LPU as a tenure strategy.