Famed Duke expert on human dishonesty suspected of fraud

A Duke University researcher known globally for his popular writings and stage presentations on the phenomenon of human dishonesty is confronting allegations that his work may have been built on fraud.

Dan Ariely, a professor of psychology and behavioural economics at Duke, is facing the retraction of a 2012 paper in which insurance company data on cheating in car odometer readings appears itself to have been faked.

The case was investigated by a team of three academics who specialise in probing inaccuracies in behavioural sciences and whose evidence has painted Professor Ariely as the person most likely responsible.

If affirmed, the discovery could prompt a thorough reassessment of the Israeli-American academic known for popular TED Talks and works that include a book titled The (Honest) Truth About Dishonesty and a movie called (Dis)Honesty – The Truth About Lies.

“My guess is that if this is fraudulent, other things are fraudulent too,” said one of the investigators, Joseph Simmons, a professor of operations, information and decisions at the University of Pennsylvania. “And if we’re going to retract all of it, those are whole literatures that are massively affected.”

The fraud appears to have been “surprisingly careless and incompetent”, said Daniel Kahneman, the Nobel prizewinning emeritus professor of psychology and public affairs at Princeton University who served as editor for the Ariely team’s article in the Proceedings of the National Academy of Sciences. “The person who fabricated the data did not expect anyone to look at them with care.”

Professor Kahneman is high on a list of academics stunned by the possibility. He largely created the field of behavioural economics and then advised Professor Ariely on his pathway to bringing it into popular culture and advising top corporate leaders on navigating human nature.

The purpose of the Ariely team’s study in PNAS was to test whether people obtaining a car insurance policy would be more truthful about the odometer reading on their vehicle if they signed a statement attesting to its accuracy at the top or the bottom of the form.

Those customers signing at the top were reported in the study to have admitted to 10 per cent more miles than those who signed at the bottom, which Professor Ariely and his PNAS co-authors described as an indication of people being pushed towards honesty by making their promise in advance.

The online investigative forum DataColada began looking at the matter after a Harvard University team tried last year to replicate the central finding, but could not, and asked the PNAS co-authors for a copy of the underlying insurance company data.

Professor Simmons, part of the three-person DataColada team, said it was important to see what insight the insurance company can provide. One theoretical possibility, he said, is that someone at the company may have found it easier to generate random data than conduct the necessary queries of its own customers.

But the DataColada probe identified far more factors that Professor Ariely has yet to address. They include his name appearing as the creator in the Excel file that he was understood to have received from the insurance company, changes in font styles that fit with the key data alterations, public discussions of the findings by Professor Ariely years ahead of the PNAS paper’s publication, and his inability to produce his own copies of emails and data files from the time.

Professor Ariely and his four co-authors on the PNAS paper had all responded to the investigation by issuing statements thanking DataColada for its work, with most calling for their paper’s retraction. Professor Ariely, however, took sole responsibility for obtaining the data from the insurance company.

“I did not suspect any problems with the data,” Professor Ariely said in his statement. “I also did not test the data for irregularities, which after this painful lesson, I will start doing regularly.”

But the DataColada team − which also includes Leif Nelson, a professor of business administration and marketing at the University of California at Berkeley, and Uri Simonsohn, a professor of operations, innovation and data sciences at Ramon Llull University − made clear the data was full of characteristics showing it had been actively manipulated.

“It’s fraud,” Professor Simmons said. “We would never use that word if it wasn’t actually that.”

While pursuing his doctoral degree, Professor Simonsohn worked in the lab of Professor Ariely at the Massachusetts Institute of Technology.

“They are very close,” Professor Simmons said of professors Simonsohn and Ariely. “We all really like him − it definitely didn’t start with: ‘Oh, let’s dig into Dan Ariely’s work.’”

In part, he said, the situation reflects differing expectations a decade ago when anyone committing fraud with large datasets had very low expectations that anyone would subject it to close computerised scrutiny.

In a brief response to questions from Times Higher Education, Professor Ariely cited only his team’s agreement to turn over the data file, which had been retained by another of the PNAS co-authors, as evidence that he had nothing to hide.

The insurance company, The Hartford, declined to comment on the matter other than saying in a written response to questions that it was not involved in Dr Ariely’s analysis of its data. PNAS confirmed that it was “looking into the matter” of a possible retraction. Duke University said it would not comment on the case, including whether it was conducting any investigation.

“I expect that Duke will follow up,” Professor Kahneman said. “They must be very concerned, because Ariely is one of their stars.”

paul.basken@timeshighereducation.com