Data Falsification

A New Scientist Newsflash today reports on the sacking of MIT professor Luk Van Parijs for fabricating data. A story reminiscent of that of Jan Hendrik Schon's sacking at Bell Labs in 2002. Back then, an investigation of alleged data falsfication resulted in the withdrawal of 8 and 7 papers by Schon, from Science and Nature respectively.

It's puzzling why people, who have managed to reach that academic level, resort to data falsification. In a way, we cannot be surprised by the practice. It seems to be the norm these days for researchers to spend a lot of solitary time working on their own projects, without any real means of verification of results by a third party. How many supervisors, for instance, sit next to their protegees to verify their work? None. The very nature of academic research demands an independent work-ethic.

Although the peer-review process is in place to pick up on faults and shortcomings of scientific research presented for publication, there is no water-tight method to actually verify how -and more importantly- whether the results presented, were actually achieved.

One would argue that reproducability of results in itself is a fairly accurate method of verification. A failure by other research groups to reproduce and back up the published results would be an indication that maybe not all is what it seems. But in Schon's case it didn't quite turn out that way. For years, other research groups failed to reproduce his results, and it was widely assumed that this was merely due to a shortcoming of these other labs.

It all puts research in a spot of embarrasment. I can't imagine what the impact must be on the colleagues of the man in question. We all assume the people we work with are skilled at what they do and we trust them to do their part of the research with integrity. Because that's how the well-oiled machine of research works. We depend on each other and assume that while you do your bit of work as best as you can, your colleagues do theirs with the same ferver. We trust that we can build on each other's results and outcomes. That's how science takes steps forward.

Unfortunately, it may be this very drive to move "forward" which pushes some people to such drastic measures as data falsification. Experiments which yield no "good" results are considered a failure. I've always disagreed with this practice. There is as much to be learned from "bad" results as there are from "good" ones. Which is why I'm pleased that the past two years have seen the birth of The Journal of Null Results as well as a Journal geared at the publication of techniques which failed. Maybe in the long term, this will take the pressure off and we can all go back to what science is really all about: Observation, documentation and learning.