Did an honesty researcher fabricate data?
STEVE INSKEEP, HOST:
This next story begins with a basic truth. Academic research sometimes falls apart. The data don't hold up or results can't be replicated in further studies. People make honest mistakes. But there's also a bigger deal, dishonest mistakes, fabricated data. Nick Fountain from NPR's Planet Money podcast brings us the story of a massive scandal in the field of behavioral science.
NICK FOUNTAIN, BYLINE: Duke Professor Dan Ariely is one of the biggest stars of behavioral science. A lot of Ariely's research is focused on how we can get people to be more honest. One of his solutions, prime people to think about honesty before you need them to tell the truth. Ariely tested this in a very clever way. For a paper, he partnered with this insurance company. Periodically, they sent their car insurance customers a form asking them to report their latest odometer readings and then, at the bottom, sign that it was the truth. Ariely tried moving that signature from the bottom of the form to the top. And, he said, that made people more honest. Here's Ariely.
(SOUNDBITE OF ARCHIVED RECORDING)
DAN ARIELY: Think about it. Just signing something a second before you fill the number - not you, other people...
ARIELY: Dramatically changes the level of cheating.
FOUNTAIN: And this is where the story gets weird. A few years ago, Uri Simonsohn and his colleagues at the blog Data Colada, known for their investigations into research, got a tip. Something was fishy in Ariely's paper, so they took a look.
URI SIMONSOHN: So it was just self-evidently faked.
FOUNTAIN: Self-evidently faked. Here was the tell. Normally, you'd expect most people to drive a medium amount - say, like, 14,000 miles a year - and way fewer people to drive very little or a lot. But in Ariely's data, the same number of people drove around 1,000 miles as 10,000, as 50,000.
SIMONSOHN: It's just so - I've never seen anything so blatant in my life. It's just incredibly, like...
FOUNTAIN: So as far as smoking gun evidence goes, this for you is like, this is it.
SIMONSOHN: Yeah, nothing will ever match this.
FOUNTAIN: Simonsohn's team wrote to the paper's original authors, told them they suspected the data was fake.
SIMONSOHN: So almost immediately, Dan Ariely replied and said, just to be clear, I was the only person responsible for this in this team.
FOUNTAIN: This meaning Ariely was responsible for getting the data from the insurance company. But Ariely wasn't taking responsibility for forging the data. He basically blamed the insurance company. His numbers, he said, came from them. Still, he and his co-authors retracted the paper. Now, for years, the Hartford - the insurance company - hasn't responded directly to Ariely's claim. But a couple of days ago, the company was finally willing to go on the record. They told us in a statement that they'd pulled the data set they'd sent to Ariely, and it looked dramatically different from the one in Ariely's paper.
The company said that they had only given him data for about 3,700 policies. But in Ariely's paper, he cited data for more than 13,000. I got the data file from the insurance company, Ariely said in a statement. Getting the data file was the extent of my involvement with the data. While we can't say with certainty who falsified the data set, the Hartford statement makes it clear the company believes the change occurred after it was sent to Ariely. Quote, "though some of the data in the published study data is originally sourced from our data, it is clear the data was manipulated inappropriately and supplemented by synthesized or fabricated data." Fabricated data in a study about honesty. You can't make this stuff up. Or can you?
Nick Fountain, NPR News.
(SOUNDBITE OF BEAR GARDEN SONG, "HEY YE YE YEAH (SLOW)") Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.