‘Open science’ advocates warn of widespread academic fraud
Scandals at Stanford and Harvard show manipulation of research remains an issue despite growth of ‘data detectives’
A decade since Brian Nosek launched an initiative to tackle academic fraud, efforts to impose greater accountability in research have in the past few weeks claimed two of their most high-profile scalps.
Marc Tessier-Lavigne, a distinguished neuroscientist and president of Stanford, resigned and pledged to retract a series of papers in prestigious journals after an independent inquiry concluded they used manipulated data. Harvard has similarly demanded retractions of papers co-written by professor Francesca Gino, a leading dishonesty expert in its business school who is currently on administrative leave.
“There is a culture of greater transparency,” said Nosek, a professor at the University of Virginia and co-founder of the Center for Open Science. “The observation of fraud is certainly increasing, but we don’t know the denominator.”
Nosek cautions that manipulation of research remains widespread. He said there were still many incentives for authors, their institutions and journals to fabricate and misinterpret data.
He has pushed for wider sharing of methods and underlying data on academic findings. This has fuelled a mini-industry of “data detectives” who scrutinise and seek to replicate findings.
“Even if all [Tessier-Lavigne’s papers] were retracted, it would be a spit in the ocean of the amount of actual misconduct taking place,” said Ivan Oransky, co-founder of Retraction Watch, which flags research papers withdrawn from academic publications.
“This literally happens every day. We track 5,000 retractions a year,” he added. This is up from less than 120 in 2002, according to Retraction Watch data.
In theory, academic journals demand rigorous peer review by other experts ahead of publication. In practice that scrutiny is often limited, many findings are still taken largely on trust and editors are often reluctant to retract work if concerns later emerge.
Improvement has come with rising requirements for studies to be “preregistered” publicly before they are conducted. That helps prevent “p-hacking” — efforts to find patterns once data has been generated, rather than respecting the scientific method of first developing and then testing hypotheses.
Academic journals also often now request disclosure of underlying data sets. This enabled Elisabeth Bik, an “image integrity” specialist whose scrutiny has led to nearly 1,000 retractions, to download and spot Photoshopped, rotated and duplicated cell and DNA images in papers co-written by Tessier-Lavigne.
Data disclosure has fuelled the blog PubPeer, where researchers can anonymously raise concerns. DataColada, another blog which has exposed numerous frauds, scrutinised the spreadsheets on which papers by Gino and her colleagues were based and identified a series of anomalies.
But Leif Nelson at Berkeley’s Haas School of Business, who is one of DataColada’s contributors, said still more sharing of the raw underlying data of published research was needed, as well as greater co-operation from faculties and universities.
DataColada’s work led Harvard nearly two years later to request retractions to Gino’s journal articles. But Leif said the university refused to confirm an investigation was under way, let alone share its scope or findings. “We do not know exactly what happened,” he said.
Both Gino and Harvard declined to comment.
By contrast, Stanford used independent experts from outside the university and made its report on Tessier-Lavigne public, but limited its scrutiny to 12 academic papers. It only started an investigation in late 2022, years after red flags were first raised by Bik and others, when one of its own undergraduates reported concerns in the student newspaper.
Ellen Evers, an associate professor at Haas, said more institutions should learn from the approach Dutch universities, such as Tilburg, Groningen and Amsterdam, which in 2012 jointly conducted and published a detailed report into the entire research career of Diederik Stapel. That led to at least 58 of his papers being retracted — to the benefit of many others working in his field of social psychology.
“They looked at every single paper which had his name on it,” said Evers, who helped blow the whistle. “It was good for science and really reduced the cost for some of his co-authors. Now we know which papers we can and can’t cite.”
While many universities’ responses remain ambivalent, other incentives to manipulate research remain in place. For Oransky, “this stuff will keep happening as long as there is a publish or perish culture”, driven by university rankings based on academics’ publications in prestigious journals.
Nosek stressed that adherence to open science principles still remained patchy. For example, data sharing is more widespread in science and medicine than disciplines such as business, where there is less pressure from funders or external regulatory scrutiny.
He has been pushing for a new model in academic publishing, which changes the incentives by getting journals to reward important research questions with publication, regardless of whether the findings are positive. Academics would register proposed experiments and editors would commit to publishing the interesting ones regardless of the outcomes.
“In science, any outcome is informative. It’s the questions that matter,” he said, pointing to a number of leading journals which have signed up, including Nature.
However, he remains cautious about progress in tackling academic fraud over the past half century. “There is still a lot of work to do.”