Daniel Lemire's blog

, 3 min read

Is research sick?

One of the most important database researchers of all time, Michael Stonebraker, has given a talk recently on the state of database research. I believe that many of his points are of general interest:

  • We have a lost our consumers… Researchers write for other researchers. They are being insular and ultimately irrelevant.
  • We ignore the important problems… Researchers prefer to focus on the easy problems they know how to solve right away.
  • Irrelevant theory comes to dominate… Making things work in the real work is harder than publishing yet-another-theory paper.
  • We do not have research taste… Researchers jump on whatever is fashionable with little regard to whether it makes sense (XML, Semantic Web, MapReduce, Object databases and so forth).

The core message that Stonebraker is trying to pass is that the seminal work he did years ago would no longer be possible today.

I do not share his views, but I think that they are worth discussing.

The fact that we are publishing many papers is not, by itself, a problem. The irrelevant theory can be ignored. The fact that many people follow trends mindlessly whereas only a few make their own ways, is just human nature.

So what is the real problem?

My take is that there is an underlying supply-and-demand issue. We are training many more PhDs than there are desirable tenure-track jobs. This puts early-career researchers in a overcompetitive game. If you can signal your value by getting papers accepted at competitive venues, then that is what you are going to do. If you need to temporarily exagerate your accomplishments, or find shortcuts, maybe you will do so to secure the much needed job.

You cannot magically make a field less competitive without tackling the supply and demand. Either you reduce the number of PhDs, or you increase the number of desirable research jobs. I’d vote for the latter. Stonebraker says nothing about this. In fact, though I like him well enough, he is an elitist who cares for little but the top-20 schools in the US. The problem is that even just these top-20 schools produce many more PhDs than they consume. Is it any wonder if all these PhD students desperately play games?

The reason Stonebraker was able to get a tenure-track job in a good school when he completed his PhD without any publication is… Economics 101… there was no shortage of tenure-track jobs back then.

Is it any wonder if the new professors recruited in an overcompetitive system are status-obsessed folks?

There is no doubt that some PhD students and young faculty members will pursue volume at the expense of quality and relevance. However, there is no reason to believe that the best American universities (the ones Stonebraker cares about) do not already take this into account. Do you really think you can go from a PhD program to a faculty job at Harvard, Stanford or MIT without one or two great papers? Do you really think that these schools will pass on someone who has produced seminal work, and pick the fellow who has a large volume of irrelevant papers instead? If that’s the case, where are the examples?

Meanwhile, there are more great papers being published in my area, in any given year, than I can read. Here is a hint: you can find many of them on arXiv. No need to go to an expensive conference at some remote location. It might be the case that some disciplines are corrupted, but computer science research feels quite worthwhile, despite all its flaws.