Dan Ariely: Full-time Faker at Duke University
His dissertations from UNC Chapel Hill are a sham. A summary of his frauds, and analysis of his dissertations and academic conduct.
What does this bench vise have to do with Dan Ariely? So very much, as it turns out.
This exact type of bench vise is what Ariely used to inflict pain on 40 test subjects for his 1996 PhD dissertation work. He used this research as his academic origin story that he peddled in decades to come. It gets much worse. But let’s start from the beginning.
Summary of Frauds
Dan Ariely is a superstar researcher and university professor at the private Duke University in North Carolina. He has written numerous bestselling books, is a popular speaker, and sought after marketing professional. NBC is currently producing a TV series based on his life and his surprising and insightful research.
In 2012 he published, along with others, a novel study that proved people could be nudged into being more honest. This was groundbreaking work. The study was cited numerous times, it yielded speaking engagements, and real-life efforts were made to implement the findings into everyday situations to improve outcomes. Such as Barack Obama using this newfound knowledge to implement certain campaign strategies, and later change tax policies. Tax programs in the UK, Guatemala and Canada also spent money implementing the findings to try to collect more tax revenue.
Doubts eventually arose about the validity of the nudging research published by Ariely and in August 2021 fellow researchers examined the data underlying the study and found it to be undoubtedly fraudulent. The numbers in the Excel file appeared to have a randomly generated uniform distribution, rather than the expected bell curve.
Just to double-check, I re-examined the same data using an additional analysis of the 2nd leading digit with Benford’s law, which predicts the distribution of digits in natural number sets. Examining the first digit likely would not work on this data set. This analysis also indicates that Ariely’s numbers were randomly generated, with the distribution and variability much more similar to a randomly generated set of numbers than to a naturally collected set of numbers.
But Ariely denied having anything to do with having faked the data and apologized only for being too trusting of those who collected the data. Of course there is a major logic flaw in his explanation. Anyone who has ever worked with data in academia knows: That data is your baby. You look through that spreadsheet with the keen eyes of a new parent. By the time you are done with it, you probably have a few favorite data points, and have run statistical analyses up and down your columns and rows. Ariely’s explanations were never even remotely credible. He would have been the one to raise the alarm over oddities and irregularities in the data set. Plausible deniability for the fake data never existed. The fault is on Duke University leadership for failing to take any action against Ariely, seemingly all too keen to discredit their entire institution to save a single researcher.
More frequently than not, those who cover for perpetrators of academic fraud are themselves involved in activities they wish not to have uncovered and thus may feel kinship with the perpetrator. It might be worthwhile to look upstream for the source of the contamination. In this case, the person protecting Ariely from consequences would be Bill Boulding, dean of the Fuqua School of Business at Duke University.
In August 2021 a Buzzfeed News reporter found the insurance company the falsified data set came from but was not able to get a detailed statement from them. This changed just a few days ago when Hartford Insurance made a clear statement to NPR that leaves no doubt about who is the sole person responsible for knowingly and deliberately inventing the data that would earn him a high reputation and millions in salaries, contracts, speaking fees, and book sales: Israeli-American Dan Ariely.
The controversy has caught the attention not only of US media but the story also gained traction in Israel. The Israeli media has generally done a thorough job reporting on Ariely’s other irregularities, in a TV report, a Times of Israel article and a Ha Makom article, the most comprehensive of them. Below is a compilation of the credible accusations of fraud and misconduct.
Sources:
1 Link 2 Link 3 Link 4 Link 5 Link 6 Link 7 Link 8 Link 9 This report 10 Link
As for the reason that there aren’t any recent studies on that list? Perhaps no one has cared to examine. And Ariely himself admitted last year that he has stopped publishing in peer-reviewed journals with a most bizarre excuse:
“In the last seven years, my research is very applied. It cannot be published in academic journals, it is not interesting enough…”
The history of Ariely’s fabrications raises the question of how long this behavior may have been going on. Therefore a review of his dissertations is warranted. He has written a Master’s dissertation and two PhD dissertations.
Master’s Dissertation
Published in 1994 at the University of North Carolina at Chapel Hill, one of the most prestigious public universities in the US.
Experiment-based research thesis on consumer choice building on Ariely’s undergrad work in Israel.
25 pages of written text, 28 references.
Raw experimental data not presented in the appendix.
Supervisor: Thomas Wallsten. Committee members: Edward S Johnson (†) and Christina A Burbeck
Ariely sets out to test if people’s choices between two consumer items are influenced by the presence of an item that is slightly inferior to one of the two. He does not explain what statistical population he is sampling. His sample size for the main part of the experiment is 60. However, this is not a random sample but undergraduate psychology students, presumably from one of the classes he was assisting with. The age or gender make-up of the sample is not mentioned, but undergraduate psychlogy students are predominantly female. They received an hour credit for their time, something they would have had to earn in other ways had they not participated. So participation was not entirely voluntary. And it cannot be ruled out that the participants were familiar with Ariely’s premise and therefore primed to give answers to please the investigator. No assurances are made to the contrary and it seems that a whole discussion on this part is simply missing. What population this sample is representative of remains entirely unexplained.
Because of the poor statistical control on the experiment, the study results are really only valid for the sample he was testing. The population is equal to the sample. Therefore whatever the results are, there cannot be wider implications for psychology or science from this study.
Questionable use of statistical analysis continues in the results of the main experiment. He finds that “none of the results were in the opposite direction of predictions” indicating that he was testing the hypothesis in both directions. His use of a one-tailed t-test is therefore clearly not appropriate and makes the p-values seem twice as significant as they are.
Another concern with the thesis is that the raw data is not in the appendix. Dissertations do not have a limit on the number of pages submitted. It is therefore most common and - even in psychology - it is standard practice to put dozens of pages of raw experiment data in the appendix for later preservation and evaluation. How could he otherwise defend his thesis if the academic committee members were unable to evaluate it fully?
Overall, the thesis setup and experimentation appears to be 3-4 months worth of work, reflected in the extremely meager 25 pages of writing. The results are not significant beyond the test subject population, there are irregularities with the reported significance levels, and the raw data is not available for evaluation. There is no error discussion anywhere. Yet his supervisor and academic committee members approved the thesis as published.
Not only that, but there is a companion published article, in which all of the problems are evident. Over 300 fellow psychology academics looked at this clearly deficient, logically and statistically irrelevant research, thought to themselves that this is relevant and noteworthy, and cited it in their own research articles.
PhD Dissertation 1 - University of North Carolina
Published in 1996 at the University of North Carolina at Chapel Hill, one of the most prestigious public universities in the US.
Experiment-based research thesis on pain perception based on his own hospital experience.
42 pages of written text, 28 references.
Raw experimental data not presented in the appendix. No appendix.
Advisor: Thomas S Wallsten. Committee members: Peter A. Ornstein, Peter C Gordon, Constantine Sedikides and one other (illegible signature).
Ariely uses this thesis as his academic origin story that he repeats everywhere. He suffered horrific burns, had to face the immensely painful treatments and “upon leaving the hospital” (or 11 years later, to be exact), began conducting research on this topic.
The premise of the thesis is based entirely on his personal experience as a burn victim. The research idea is to prove his doctors wrong in that they were trying to minimize the duration of pain, rather than the intensity. He proves that this is a pervasive general issue worthy of a closer examination by surveying a mere 18 medical staff in a single hospital burn unit. No other sources, from the academic record or from elsewhere, are consulted to confirm this idea, and neither the survey nor the data points from it are presented in the dissertation.
It is a doctoral dissertation, so Ariely manages to scrape together a few more pages of writing. He cites a mere 28 references. But we do get a little bit more information on his test subjects than in the Master’s thesis. There is an age range and gender makeup of the sample, all local graduate students. 20 subjects for one experiment and 40 for the other. Again, no attempt to define what population this sample is supposed to represent, no error analysis. Just an implication that the results represent how pain is perceived in general. So again, the scientific relevance is limited to the actual study participants.
Ariely conducted two experiments. The first one did not answer his hypothesis in full because of an issue with the equipment dispensing heat for less of a duration than he desired. This is where the bench vise comes into play. In a much simpler setup than in the first experiment, Ariely decides to inflict pain by tightening a bench vise on people’s fingers. In the US, this kind of research requires review and approval from an institutional review board, involving paperwork which can certainly be found in the appendix. Except that the dissertation has no appendices at all. Contacted by email, the director of Office of Human Research Ethics is also unable to locate any paperwork.
After a very dark history of involuntary scientific and not-so-scientific experimentation on people, in 1975 the US federal government enacted tight regulation for the review and approval of experiments - however benign they may be - involving human subjects. This includes the requirement to get prior approval for any experiments on people from an institutional review board (IRB). Dan Ariely, however, doesn’t have time for lengthy approvals and apparently skipped the mandatory process for his PhD research of tightening a vise on fingers. But not to worry, he assures us in his dissertation that this is perfectly normal and cool:
“I should note that I personally used this procedure on myself […] with no lasting damage or pain. Such self-testing of equipment is the most common and safe method of developing procedures and techniques in this type of research.”
Apparently at UNC Chapel Hill researchers can self-certify their experiments. I should repeat here that the dissertation does not have any appendices at all, so of course no consent form samples or equipment testing data is available for his dubious experiment.
He also wrongly assumes that pressure on the fingers is proportional to the distance that the sliding jaws on the vise move. That is not how pressure works and an outside reviewer of the experimental setup might have brought attention to that fact. Fingers of course have different properties like thicknesses, elasticity, mass, water content, etc. No pressure measurements were performed, no medical professionals consulted. Ariely simply normalizes pain values from each subject by calculating a t-value. Ethically and scientifically highly questionable.
He starts with the thesis with a foregone conclusion that he then affirms at the end without any meaningful discussion on how or why medical staff act the way they do. Once he has concluded that were treating him in a suboptimal way, the discussion ends abruptly after three short sentences. A highly me-focused approach to the research that is intellectually lacking, especially so for a doctoral dissertation. A nuanced examination of the issue that examines the actions in a wider context does not take place and the author clearly has no ambitions to make this research useful for anyone but himself. This could have been a useful examination of the issue of treatment of burn victims but the author’s ambitions did not extend that far. It is doubtful that even a political sciences student could have gotten away with such a minimal effort for a PhD thesis.
PhD Dissertation 2 - Duke University
Published in 1998 at the Duke University
Experiment-based research thesis about the role of interactivity on consumers’ choices.
74 pages of written text, 87 references.
Raw experimental data not presented in the appendix. No appendix.
Co-advisors: John Lynch and Jim Bettman, Committee members: Ziv Carmon and Constantine Sedikides
This one is the academically most rigorous of his three dissertations. I have given this one less examination than the others but the same obvious flaws appear in this dissertation: The experiments are performed with a small number of participants about which almost nothing is known, and no attempt is made to even define the population that is supposed to be sampled by this number. There is no evidence the experiments were really performed as no raw data is presented in an appendix. No appendix exists, which is highly suspicious for any data-producing PhD dissertation. The academic committee members likely did not examine the thesis thoroughly as they approved it presumably without even looking at the raw data.
The thesis does appear to conform to academic standards in the field of research he is studying as we get - unlike in his UNC dissertation - more than three sentences of discussion of the implications of the results.
Harm Done
One may question if any real harm was done by Ariely’s academic antics of faux research and embellished storytelling of supposedly scientific findings. I argue that there has been such harm. His intentionally and deceptively fabricated data on the 2012 car insurance study and the fame he earned form that seems to have netted him a job at the insurance company Lemonade. For 5 long years they paid him real cash to sell them his fabricated theories on human behavior.
How many executives did he motivate to slash bonuses in favor of pizza parties? We’ll never know how many millions employees lost because of his tall tales of human behavior. Money that those people could have spent on goods and services to purchase in the economy. The effect may have been small, but in the economy of the US, small still amounts to tens or hundreds of millions of dollars.
And in the picture below is Ariely shilling hard for one of the least popular ideas in history, trying to convince people to act against their own interests and in favor of a few select commercial real estate owners and antsy managers.
Behavioral Science vs Actual Science
There has been an issue with replicability of Ariely’s experiments. This only appears puzzling to those in his field of study.
In the study on how the recitation of the biblical ten commandments improves honesty in people, the test subjects are described as: “Two hundred twenty-nine students”. That’s it. That is the entire description. No attempt define what these 229 are a statistical sample of. And whose students were they? Likely from one of the classes they were teaching, possibly even primed and motivated to prove their professors correct. The authors pretend that these 229 represent humanity at large. Maybe all of them were male, or gay, or church goers, or billionaires. The authors never attempted to find out because then they would have had to do real science.
It is like a geologist collecting the 229 rocks closest to his office, then performing various tests and measurements on them, and telling the world about the distribution of minerals in the earth’s crust and the age of the formation of the earth. That approach is preposterous, their research would be meaningless. And so are any studies conducted in such a way with no statistical control of the population sample. So of course these studies do not replicate. This is no surprise to any actual scientist reading those studies. The ten commandments study was cited over 3000 times as somehow relevant. Behavioral science done this way is no more relevant to the world than philosophy. That is exactly what Ariely has done in his work: He is a philosopher that deceives people into believing that his ideas and his storytelling have been proven correct by some experiments, which are actually scientifically meaningless.
Actual scientific studies define the population to be studied, then try to find a statistically representative sample from that population, then take detailed peripheral information from that sample - in case of people this is often income, national background, gender, age, beliefs, etc. - and have an error discussion of their results, and may attempt to correct for the error by accounting for numerous variables.
Almost all researchers at universities are knowledge-driven. They are studying and examining things in a quest for knowledge and discovery of the truth and how things work. So when Ariely’s peers loudly proclaim “Dan Ariely is a fraud”, what they mean is that he is not one of those knowlege-driven people. He is there to storytell and self-promote, betraying the foundation of a university as an institution in society.
OK I am puzzled at one test.
"Just to double-check, I re-examined the same data using an additional analysis of the 2nd leading digit with Benford’s law, which predicts the distribution of digits in natural number sets. Examining the first digit likely would not work on this data set."
But Benford's Law (which is sort of cool) only applies to the lead digit
https://statisticsbyjim.com/probability/benfords-law
Shame, I enjoyed Dan Ariely's books. I didn't know the underlying research had so many issues.