Study About Canvassers Changing Minds Likely Used Fabricated Data

A study published in Science that This American Life reported on a few episodes back has been asked to be retracted by the author, as one of the co-authors (a grad student) apparently fabricated all the data for the online survey.

Ira Glass has put up a few posts and tweeted about it earlier.

Original reaction and

Obviously this is far removed from being This American Life’s fault given the reputation of the author and the publication, but it sucks to even be associated with a whiff of a falsified story.

Thanks for sharing. I find this fascinating, since I have a special interest in fraud.

I do think that overall it’s a nice example of how the scientific process is supposed to work, with researchers attempting to replicate the study being unable to do so, investigating more closely, uncovering the fraud, and then widely publicizing it.

I like that the original co-author of the study, who wasn’t involved in the fraud, says that the appropriate resolution is to conduct a new version of the study to try to verify the hypothesis.

It’s unfortunate that for scientific studies like this one that are widely covered in the popular press, the original results can sometimes take on a life of their own and resist attempts to invalidate them.

Wow! I just listened to this episode the other day. I thought it was one of the most interesting scientific stories I’d heard. Too bad it’s based on a fake study. But Ira wasn’t kidding when he said the story of the fraud and how it was uncovered is almost as fascinating as the original story!

I can kinda understand why someone would fake data. But c’mon, he had to have known this would backfire. If you’re going to fake data fake about something that no one gives a shit about, not a highly political issue that’s currently in the public’s eye.

According to one of the stories I read, this guy had an appointment at Princeton that is now being reconsidered. That could be a huge incentive to fake the data on a prestigious study, especially if he had been selected to do the work and then couldn’t complete it. He claimed in his submission that he had grant funding, and that he had paid people to participate in the non-existent survey. If the grant never materialized or fell through, and he had no other resources, he may have felt his only choice was to fake it.

Of course he could have done a better job with the statistics. Everyone knows you can’t be too consistent and need to throw in some truly random outliers (he should have read Nate Silver). Irregularities in LaCour(.pdf) is the study reporting on the fraud. It’s fascinating reading even if some of it is far too technical for me.

As It Happens (CBC Radio’s equivalent to NPR’s All Things Considered, I guess) had an interview with the senior co-author for that paper (the guy who asked Science to pull it). He said that LaCour proposed this study to him using preliminary statistical data which looked, at the end, to be very similar – and every bit as fake – to the final published data.

So it appears that LaCour started out with faked evidence and never saw any need to fix it.

ETA and slightly offtopic: I love As it Happens, but every time I hear their opening theme I think to myself “Hey, 1978 called, they want their disco back.” :smiley: