My personal experience with academic fraud arose from being a co-author on a paper described as a “clusterfake” by Data Colada, a blog written by researchers Leif Nelson, Joe Simmons, and Uri Simonsohn.Reference Simonsohn 1 Data Colada exposed two different and likely independent frauds in the same paper that I co-authored with Lisa Shu, Nina Mazar, Francesca Gino, and Dan Ariely in 2012, entitled “Signing at the Beginning Makes Ethics Salient and Decreases Dishonest Self-Reports in Comparison to Signing at the End” (the signing-first paper).Reference Shu 2 In the paper, my co-authors and I claimed that if people sign a statement promising to answer truthfully before filling out a form, they are more likely to be honest than if they signed such a statement after filling out the form. Possible applications for this idea could include filling out your tax return or signing a reimbursement request. The idea that signing first might encourage people to act more ethically is intuitively compelling: if true, it would offer an effective nudge toward more ethical behavior.
For many years, I have thought a great deal about the ethicality of my behavior and the behavior of those around me. Some view me as being obsessed with ethics, as I frequently examine my own behavior on ethical issues and also research and write about such questions. A specific ethics topic that has fascinated me is the complicity of other people in the actions of harm-doers, from Nazi collaborators during World War II to the corporate board that failed to provide proper oversight in the Theranos scandal.Reference Bazerman 3 For the last decade and a half, behavioral ethics has been the central focus of my academic career. I have written three books on behavioral ethics: Blind Spots: Why We Fail to Do What’s Right and What to Do About It (with Ann Tenbrunsel), Better, Not Perfect: A Realist’s Guide to Maximum Sustainable Goodness, and Complicit: How We Enable the Unethical and How to Stop. More recently, I have been writing Inside an Academic Scandal, a book that will be published soon after this paper. The book describes the evolution of the signing-first paper, which I will overview in this paper.
The signing-first paper was influential: It was cited over 500 times, and its central finding was implemented in multiple real-world organizations. Data Colada provided convincing evidence that fraud occurred in at least two of the three experiments overseen by two different co-authors: Study 1 by Francesca Gino, and Study 3 by Dan Ariely. I will describe how the paper developed and offer insights into the institutional changes that are needed to prevent similar episodes of fraud in the future.
I appreciate the opportunity to contribute my experience and social science perspective to the biomedical perspectives that are the core focus of this special issue of the Journal of Law, Medicine & Ethics. I believe that many of my core conclusions connect well to insights from other papers in this volume, including the papers of Schrag, Patrick and Bik, Oransky and Marcus, and FitzGerald.Reference Schrag 4
Evolution of a “Clusterfake”
In their coverage of the alleged frauds in the signing-first paper, Data Colada and the media focused on Dan Ariely and Francesca Gino, the two authors most closely connected to the fabricated data. Ariely was my co-author on the paper but not someone I knew well. Gino, in contrast, was my former advisee, longtime Harvard Business School (HBS) colleague, and friend. Gino was the prolific author of 135 empirical papers written with 148 co-authors, most of them junior scholars. Data Colada also documented fabricated data in three other papers that Gino co-authored.
The signing-first paper combined two separate empirical efforts to test the hypothesis that signing first leads to more ethical reporting. One of these efforts was Study 3, a field experiment that Dan Ariely allegedly conducted with The Hartford Insurance Company. The other effort included two laboratory experiments on which Lisa Shu, Francesca Gino, and I collaborated. Gino managed this data collection through a lab manager at her prior university, the University of North Carolina, after joining the Harvard Business School faculty.
At the time the data was collected, I had never considered the possibility any co-author of mine would fabricate data. But as I read the first draft of the combined paper in 2011, I had questions about peculiar aspects of the insurance company data that Ariely provided, questions that I asked in email exchanges with all co-authors of the paper. My questions had to do with pieces of data that were counterintuitive. I did not suspect fraud. I was simply trying to understand the data. Ariely’s evasive responses to my questions should have led me to investigate details of the data collection in more detail. But, when another co-author provided a viable explanation, I was too pleased to accept the explanation and move on.
In 2021, Data Colada made public allegations against Study 3, the insurance study led by Ariely.Reference Simonson 5 The accusations of fraud focused significant media attention on Ariely. Data Colada’s allegations against Study 1, led by Gino, came two years later, in 2023, and attracted much more media attention. 6 The repercussions were massive. First, Gino’s co-authors, beyond the four papers associated with the fraud allegations, didn’t know whether they could trust their data anymore, whether to tell their universities about their uncertainties, and whether to ask journals to retract their already-published papers. Second, universities spent significant time and effort investigating the alleged frauds. Third, other researchers were misdirected based on the publication of a fraudulent result in the literature. Michael Sanders, then the chief economist of UK Behavioral Insights Team (BIT), told Science that about $250,000 was spent to try to induce honesty in tax returns based on the effects of the signing-first paper — an experiment that did not find a significant difference between taxpayers who signed a return before versus after filling out the form. The BIT team’s efforts had been wasted “because we were digging in the wrong place,” according to Sanders.Reference O’Grady 7
Why People Cheat
Having a tenured professorship at a prestigious university is a pretty good position. Many have asked why anyone would risk such a privileged position by making up data. The most obvious reason for cheating in academia, and beyond, is self-interest that breaks through the guardrails of clear ethical principles — people may perceive that they will be better off if they cheat. Drawing on the nineteenth-century writing of English philosopher Jeremy Bentham, economist Gary Becker presents an economic perspective on criminal behavior, stating that a person will break a rule (e.g., commit a crime) if the expected benefits of doing so exceed the expected costs.Reference Becker 8 Becker includes benefits beyond financial gain, such as psychic benefits. Scholars obtain financial rewards and prestige from the visibility that comes from publishing interesting results.
Beyond this straightforward explanation, much has been written on the psychology of unethical behavior. In 2011, psychology professor Diederik Stapel confessed to his employer, Tilburg University in the Netherlands, that he had engaged in data fabrication for many years. In a subsequent memoir, Stapel listed the many factors that contributed to his fraud: “The need to score, ambition, laziness, nihilism, want of power, status anxiety, desire for solutions, unity, pressure to publish, arrogance, emotional detachment, loneliness, disappointment, ADD, addiction to answers.”Reference Stapel 9
Soon after Gino was accused of fabricating data in June 2023, Uri Simonsohn, one of the three members of Data Colada, described to me his interest in creating an online platform that would allow all of Gino’s co-authors to identify who had controlled the data in each study of the many papers she co-authored. I joined this effort, along with Don Moore (Haas School of Business at Berkeley), Julia Minson (Harvard Kennedy School), Juliana Schroeder (Haas School of Business at Berkeley), and Maurice Schweitzer (Wharton). The platform we launched can be found at manycoauthors.org. I was motivated to get to the bottom of this scandal, first and foremost, because of my own involvement in some of the research. I was also motivated to provide a forum that would assist more junior scholars, doctoral students, and recent Ph.Ds with clarifying the provenance of their data. Some questioned whether Gino’s co-authors could be trusted to report this information honestly; I concluded, as I believe the other organizers did, that it was appropriate to trust them. As Simonsohn has noted, academic fraud is rarely a collaborative effort. Across the many scandals he has investigated, he uncovered no evidence of fraudsters working with a partner. At the same time, co-authors are likely to have the greatest access to hints of data fabrication.
Toward Solutions
Questionable scientific practices have been found in scientific research as long as the scientific method has existed. Psychology is no exception. In the late nineteenth century, early scientists were at odds with sorcerers, wizards, charmers, shamans, medicine men, and seers.Reference Benjamin and Rieber 10 Pseudo-therapies that have since been refuted by the scientific establishment include mesmerism (healing with magnets), phrenology (measuring aspects of the head), physiognomy (observations of the face), and spiritualism (a variety of invalid paranormal views). 11 Many pseudo-therapies also often had racist overtones. But questionable psychological claims never went away. In 2011, in a leading social psychology journal, Cornell University psychologist Daryl Bem published the unbelievable claim that people can be influenced by an unforeseeable future event (because of precognition, a type of extrasensory perception).Reference Bem 12 Few suspected Bem of fraud, yet his results, as expected, were not replicable. Many point to this paper as prompting a crisis in psychology that led to the open science movement, or what Data Colada has called “psychology’s renaissance.”Reference Nelson 13
Nelson, Simmons, and Simonsohn published an influential paper on p-hacking in 2011Reference Simmons 14 and launched Data Colada in September 2013 as a blog focused on investigating the replicability of social science research. P-hacking involves inappropriately manipulating data to increase the likelihood of finding a significant result. Enthusiasts of the open science movement have highlighted changes that are needed to reform our research practices. Yet many well-known social scientists have argued against reform.Reference Bazerman 15 Increasingly, the evidence has supported the views of the reformers, and scientific journals have incorporated several guardrails to help authors protect themselves against easy fraud.
Seven years after their paper on p-hacking was published, the Data Colada team argued that if a research psychologist had hibernated between 2011 and 2018, they would not recognize their own field upon awakening. 16 Changes included psychologists showing great interest in symposiums and articles on research methods, authors posting their data online, journals valuing replication projects, the preregistering of studies, and labs working together on replication projects that a journal reviews before the data are even collected. 17 It was fascinating and gratifying to observe how much the field had changed.
But the field still has far to go. Along with other data reformers, Nelson, Simmons, and Simonsohn have emphasized the need for two major changes to the research process: disclosure and preregistration.Reference Vazire 18 Disclosure refers to researchers presenting their work in a way such that any reader or reviewer could fully understand what the researchers did at each stage of the research process. This includes knowing when the researcher created the hypothesis, such as before or after the data were collected. Disclosure should also include descriptions of all the conditions in an experiment, all the procedures followed by participants, and all the measures collected.
As Nelson, Simmons, and Simonsohn note, many social science researchers have begun to provide disclosures, even when not formally required to do so by journals. 19 This shift occurred in part due to a widely read suggestion from The Center for Open Science that reviewers could include the following in their reviews: “I request that the authors add a statement to the paper confirming whether, for all experiments, they have reported all measures, conditions, data exclusions, and how they determined their sample sizes. The authors should, of course, add any additional text to ensure the statement is accurate.”Reference Nosek 20
In another important change, social science researchers are increasingly preregistering their research plans. Preregistration involves submitting predictions and a data analysis plan to an online platform, typically either AsPredicted.org or osf.io, before collecting data. Because preregistrations are digitally time-stamped, others can clearly identify which analyses are testing predictions made in advance and which are exploratory. Preregistration does not prevent researchers from engaging in exploratory analysis but simply clarifies what they predicted in advance. 21
A useful third procedure that can improve integrity is for researchers to make their data publicly available by posting it online. Knowing that others can easily examine data could reduce the temptation to engage in fraudulent behavior.
These three processes, now commonplace for many researchers, provide a useful signal that data has been collected with integrity. Obviously, a researcher can implement these processes on their own, but co-authors and various institutions can also play an important role in promoting them.
Co-authors
Journals, universities, and funding agencies should encourage co-authors to play a larger role to improve the integrity of research. I recommend two new roles for co-authors. The first is to be more responsive to peculiar results. If something seems off or very surprising, co-authors should investigate thoroughly until they fully understand their data. I am sure that some co-authors have effectively implemented this suggestion in the past. I simply think that more of us need to be engaged in following up suspicions. As I document in Inside an Academic Scandal, I had hints that something was off in Study 3 of the signing-first paper but was pleased when a co-author gave me a plausible explanation of the peculiarities that I observed. In contrast, when a co-author sees something that doesn’t make sense, they should follow up until they have a complete understanding of the data.
I also recommend that institutions encourage all authors of co-authored papers to critically review some basic data provided by the author doing the analytics (perhaps in a lab meeting) and to have at least one co-author review the data and details of the analysis. High-integrity researchers should be delighted to have an extra person review their data and analysis. Notably, anyone who might be tempted to fabricate data would be less likely to do so if they knew someone on their team would be looking at it closely. If researchers can verify that their data was reviewed by a second set of eyes, the paper should have more credibility. A second look would also reduce any honest errors made in the publication process. By making such co-author reviews normal, it also destigmatizes the request to review the work of a co-author.
Many journals have moved to require that co-authors clarify who contributed to various parts of the process to create the paper that is being submitted. And Data Colada is creating a new platform called AsCollected. AsCollected will allow co-authors to provide details about their data collection and analysis, including who was involved in each aspect of the research and where and when data collection took place. I look forward to seeing how this platform develops and is diffused to bring co-authors into the story of improving data integrity.
Finally, in the aftermath of the signing-first episode, some co-authors are providing a new model of data transparency in their labs. Many of my junior colleagues at HBS have created a combined document on good research practices to guide their labs. Scholars who oversee lab groups have the power to change research norms within their labs, and department chairs and deans have the power to influence data integrity policies in their institutions.
Journals
Journals are playing a critical role in moving researchers toward better research practices. Psychological Science, following the lead of Science, was the first highly visible psychology journal to publish very short papers, a shift that often stood in the way of full disclosure. But in 2014, just a few years after the open science movement started, Psychological Science began implementing stronger disclosure rules, removed word limits for Methods sections (which could have worked against full transparency), and offered “badges” to authors who preregistered their studies and shared their data and materials. In 2022, 69% of articles published in Psychological Science earned open science badges.Reference Hardwick and Vazire 22 Even stronger changes occurred with Simine Vazire’s appointment as the editor of Psychological Science in 2024. Vazire, a vocal advocate of open science, required that data, original research materials, and analysis scripts be made public in all papers published in Psychological Science. Exceptions are allowed when ethical, legal, and practical reasons for protecting confidential data exist.
Vazire introduced “registered reports,” a new review process that allows authors to submit their research plan (hypotheses, methods, etc.) for peer review before they collect data and to be potentially granted “in-principle acceptance,” committing Psychological Science to accept the paper if it is competently executed as proposed. This process aims to keep author bias from affecting data collection and analysis and to prevent “bias stemming from editors’ decisions during peer review,” wrote Vazire and Tom E. Hardwicke, a new senior editor at the journal. 23
Hardwicke was appointed to lead a new group of editors at the journal focused on statistics, transparency, and rigor. These editors examine the methodology of submitted manuscripts by performing light transparency checks on all manuscripts sent for peer review and more in-depth checks on some manuscripts that have been conditionally accepted and are likely to soon be published. The journal also created a partnership with the Institute for Replication aimed at verifying the reproducibility, robustness, and replicability of articles published in Psychological Science. 24
Psychological Science is a role model for other journals, but many journals will change more slowly. For example, in February 2024, Mike Baer and Maryam Kouchaki, the co-editors of the Organizational Behavior and Human Decision Processes, wrote:
Over the last several years, OBHDP has taken important steps … such as requiring pre-registration for all new studies conducted during the review process, mandating clear reporting, and encouraging authors to make all study materials and data available to reviewers and readers. It is time to take the next step. Over the next months, we will be codifying the journal’s position on each aspect of the research process … we want to assure authors that our policies will balance the need for integrity and transparency without being exclusionary.… A measured approach will ensure that the diverse research approaches we welcome are not excluded from submission.Reference Baer and Kouchaki 25
This statement highlights the conflicts associated with adopting the best practices recommended by the open science movement. The editors clearly feel the need to move in the direction of the open science movement yet are concerned with being “measured”.
Universities
As open science reformers have provided clear guidance about needed changes to scientific practices and journals have begun to implement changes, universities, which employ the majority of social scientists, have remained on the sidelines. When universities allow questionable research practices, their reputations suffer. It is time for universities to become active participants in the open science movement. They can begin by creating task forces to review their social science research practices and to consider requiring many of the ideas summarized in this chapter. Rather than outsourcing the task of reforming social science to lawyers and journals, universities should engage in discussion about how to move forward.
In the aftermath of data fraud allegations, American universities often close ranks and keep their investigations and decisions secret. When professors have resigned, their universities typically say little about why they did so. Some have speculated that these universities have avoided investigating all the papers published by the resigning professor, with the exception of retracted papers. American universities also tend to be remarkably slow. I appreciate the need for universities to be thorough and fair to the accused. Universities may also have good legal reasons for these behaviors. But the slow speed of investigation can also be morally problematic, as delay leaves the professor’s co-authors, the broader research community, and the public less informed than a faster and more transparent investigation could provide. While the investigation meanders along, leakage and rumors occur, often harming innocent scholars related to the inquiry. The need to be fair to the accused must be balanced with the need to keep the accused from potentially committing further fraud, to ensure the integrity of published research, and to protect the intellectual communities that surround the accused, particularly junior colleagues. The opaque nature of these investigations leaves colleagues of the accused in the dark when students and junior faculty ask about whether to work with them. My call for greater transparency matches the recommendation of multiple other papers in this special issue, including those by Briskin and Gunsalus and by Oransky and Marcus.Reference Briskin and Gunsalus 26
Universities also need to give far greater support to the co-authors of the accused, particularly junior researchers, after the investigation begins. In a 2023 article, Marret K. Noordewier, a former advisee of Stapel’s, reported that the scandal led her to face years of wasted work, retracted papers, unwanted media attention, and unpleasant investigations.Reference Noordewier 27 Noordewier managed to reboot her career by pursuing a second Ph.D.; she is now a psychology professor at Leiden University. She recently gave the following advice to universities dealing with allegations of data fraud:
[A]sk the involved researchers what they need. No one can be prepared for the consequences of being affected by scientific misconduct … Those affected may need legal help, advice on how to deal with the media and psychological help with picking up the pieces…. [I]f we do not want researchers to be slowed down beyond the time they have already lost, there needs to be room for nonstandard decisions in (for example) grant applications, hiring decisions and promotion criteria. Recognize what people have accomplished and appreciate the work that is still valid, meaningful and useful.
Make sure researchers feel safe and free to share their concerns and needs. Responsible research practices are not only about research, but also about researchers and the culture we work in. This includes the support of those who are affected by scientific misconduct.
Noordewier raises important issues for universities dealing with allegations of research fraud to consider.
I believe that universities can also do a better job of protecting scholars who speak up. While the three members of Data Colada were all employed by universities, it took a GoFundMe campaign to pay for their legal defense when Francesca Gino sued them for defamation, charges that were dismissed by the judge. 28 If we view creating high integrity research as part of the scholarly goals of research universities, these universities should be more supportive of those who speak up to protect the integrity of science.
In addition to researchers, lab leaders, journals, and universities, governments and other research funders should also be involved in discussions of how to reform the field. These entities are involved in social science research because they believe it can make the world better. Implementing these proposed reforms would result in more replicable and powerful research.
The Need for Equitable Reform
As documented above, some journals have adopted the ideas from the open science movement, including requiring scholars to pre-register their hypotheses and data analysis plans, to describe their research methods in sufficient detail that other researchers can replicate their studies using very similar procedures, and to make data publicly available. These are clear steps for creating more replicable research. Yet, universities and government funders (e.g., the National Science Foundation and National Institutes of Health) have been too slow to connect to the open science movement. As a result, their resources are being wasted, and they are allowing bad practices to continue.
These recommendations of the open science movement will create additional work and increase research costs. Yet these changes are needed and justified for the long-term benefit of science and society. The bad behavior of some researchers requires the rest to accept new costs to ensure continued trust in science. Some institutions can more easily absorb the costs. But researchers at less-funded institutions, already at a disadvantage, will be further disadvantaged. Thus, the broader community, including the government and funders, should consider ways to subsidize reform for those least capable of absorbing its costs.
I close on a personal note. I only connected to the data integrity issue as a result of being a co-author in a clusterfake. And my career has been far less enjoyable for the past three years as a result of this connection. Quite simply, if I had followed the practices recommended in this article, either on my own, or by being encouraged by the institutions around me, I do not think that I would have been part of the story. In fact, I do not think that the frauds that I am most closely associated with would have reached the printed pages of a leading journal. I hope that my reflections are useful to help us reform our institutions.
Acknowledgements
This paper benefitted from the insightful comments of Mahzarin Banaji, Steven Joffe, and participants at the 10/10/24 Penn conference on Advancing Trust in Science, as well as the numerous colleagues who spoke to me about their experiences as I was writing Inside an Academic Scandal. This paper borrows recommendations also discussed in Inside an Academic Scandal.
