Who's Afraid of Data Fabrication?

Emerging technology is helping uncover fabricated science at even the most prestigious research institutions

There has recently been a rash of accusations of research falsification at the highest levels of elite research institutions. The latest bombshell is from the Dana-Farber Cancer Institute, where four senior researchers were accused of falsified data and plagiarizing images across several papers.

This follows the resignation of Stanford president Mark Tessier-Lavigne over manipulated data as well as an ongoing controversy involving Harvard Business School professor Francesca Gino and accusations of data fabrication that have led to her suspension.

The story detailing the Dana-Farber fabrication is most interesting because of how lazy the fabrications are. Most of the fabricated data and images aren’t even skillfully faked. Were I to fabricate data, I would build a model of what I wanted from it in line with a result I either expected or felt fairly confident I would get if I performed the experiment. Then I would add some randomness to it and maybe run it a few dozen times until I got a result that was the most visually compelling.

That level of data fabrication is not what we are seeing here. This is much lazier. Graphs are copied and pasted. Results are flipped horizontally and blurred. In many cases, it only becomes clear when someone points it out; but, once it is pointed out, it becomes painfully obvious as we can plainly see that even the anomalies have been duplicated.

My first impulse on this story is to say that this is a result of the publish-or-perish paradigm in academic science. Researchers under enormous pressure to pump our papers with novel results are cutting corners because the hard work of real science is time consuming and tedious.

But then why spend all this time faking so many results? Why not just use fewer results?