The irony perform case has attracted the attention of many people: a research on the psychological mechanisms of dishonesty that ended up being portrayed (roughly, unpublished) because of… apparently fraudulent data.
Part of perform buzz around a situation also has to do with the relative celebrity perform researcher responsible for the presence of misleading information zero study. This is perform Israeli-American behavioral economist Dan Ariely, teacher at Duke University (South 2 USA).
Author on three bestsellers, speaker on success on the web electronic columnist perform newspaper The Walls Street Record, Ariely is also the author of The (Honest) Truth About Dishonesty: How We Lie to Everyone, Especially to Ourselves. zero specialized journal PNAS, from a National Academy of Sciences 2 USA. In the work, Ariely electronic plus four co-authors from different institutions claim to have identified a simple measure to increase the honesty of these people in different contexts.
According to a study, it would be enough to see that people sign a form stating Yes , I will pass on reliable information beforehand about participating in different activities so that they learn to behave in a substantially more honest way. Placing the signature on this statement after the activity would not have the same effect, the study suggests.
The group tested this hypothesis through different experiments, however the controversy arose because 2 suspicious numbers perform experiment 3, zero which was analyzed information supposedly obtained with the help about a car insurance company that Ariely had contacted.
According to operating system survey authors, operating system mileage data about About 20 million cars registered with the insurer indicated that when the person first signed the declaration on electronic honesty then filled it in while other information, the mileage reported was highest on average.
This would be a sign that the presence of a zero signature top perform form on fact led to more reliable filling, t that higher mileage tends to increase like fees charged by the insurer. Operating system results turned out to be another example of how small situational therefore changes can positively affect human behavior.
Zero However, members of the same team tried to reproduce these results in further studies, without success. , one that flashed a yellow light about seeing the conclusions of an electronic survey on a Duke University expert.
Basically, the only one who had access collects 2 period field data Ariely himself. Operating system other authors helped in the electronic analysis discussion 2 results, explains Altay de Souza, researcher perform Department of Psychology at Unifesp (Government University of Therefore Paulo).
It was then that a trio about researchers formed by Uri Simonsohn, Leif Nelson electronic Joe Simmons, responsible for the DataColada blog page, in which they analyze controversial issues about data on behavioral studies. The perform blog site team (along with other researchers that operating system helped anonymously) kept scratching their heads at some oddities in the information from the unique survey.
With bottom on information about other insurance companies , they reached what would be the expected proportion of kilometers traveled by policyholders. As sony ericsson would expect, we see that some people drive a lot, others drive a little electronic most stay in the media, they write. of this intuitive pattern. Exactly equal groups of drivers drove 0, thousand, 10 thousand, 30 thousand, 40 thousand electronic 50 thousand miles (unit of measurement normally used in the US). Electronic absolutely no one exceeded the 50 thousand miles in the period studied.
Also, after operating system 2 car owners signed operating system forms, while mileage (or mileage) reported zero include no rounding. In other words, all operating system participants supposedly looked at the electronic odometer and wrote down with exact operating system numbers 25,673 million kilometers traveled, let’s say. Other studies, however, show that it is very common for people to round this type of number on forms.
Both operating system factors indicated that operating system data could have been generated automatically using random numbers. According to the perform DataColada team, although it is not possible to be sure who manipulated the electronic information for what reason, we do not need to know the answer to these questions to know which operating system data was manufactured, without a doubt.
Some 2 authors of a first m research had learned to declare themselves in favor of retracting an article, that is, making it invalid, when zero were able to replicate operating system results in subsequent studies. When the DataColada perform analysis came out, co-author Lisa Shu tweeted: It was a delight to see this brilliant work on investigative journalism on data. The article was eventually invalidated by PNAS magazine.
In a message to the blog site, Ariely said that a process on electronic data tabulation 2 collection was made entirely by the electronic insurer reaffirmed that he had I had contact with the company.
I zero suspected that there was a problem with the operating system electronic data zero I did tests to verify that they contained irregularities, something that, after this painful lesson, I will start doing it regularly, he said.
The situation can be seen as a broader problem of this kind regarding behavioral studies with human beings: the area has suffered serious zero difficulties with regard to replication over studies previously considered classic. This so-called crisis over replicability affects even Nobel performers on economics, such as Israeli Daniel Kahneman (by the way, initial editor perform unpublished article).
Kahneman has been very honest about this, he even sent a open letter to the scientific journal Character calling attention to the need to reproduce classic studies, says Altay of Souza. He put the baton down on staff in these areas of behavioral economics electronic advertising that do not use his theory as a theory, but as an authoritative argument.
For a researcher at Unifesp, a case involving Ariely plus one warns of the need for transparency in the area, with the immediate posting of 2 raw data on each study along with the publication.