Why Academic Fraud-Fighting Always Loses to the System — From Six U.S. Cases to Bazerman's Five Reform Pathways

1. The Lead: Classmate Geng Gets Permanently Throttled — and Six Cases Ten Thousand Kilometers Away

On May 28, 2026, the Douyin account of "Classmate Geng," a Chinese academic-fraud watchdog blogger, was permanently throttled, and permanently banned from Xingtu commercial brand deals. His serial exposés, begun in April, had in succession led to the removal of a dean at Tongji University — demoted two ranks — and triggered investigations at Nankai University and Sun Yat-sen University. Still in his hands: evidence that five Distinguished Young Scholars (Jieqing) awardees are suspected of research fraud, published in Nature's flagship journal and its sub-journals.

Classmate Geng's own résumé is unusual: a bachelor's and a master's in biology from Jilin University, a fifth-year doctoral dropout at Beihang University (Beijing University of Aeronautics and Astronautics). He is not the protégé of some senior academician, nor someone with a patron inside the system — he is "a person who has already left academia." This very status is the source of his fraud-fighting power: he has no fear of offending colleagues, no fear of his projects being blocked, no fear of being ostracized by the academic community. Even after state-media coverage, he was still throttled.

Back home, people already explain why big-name academics are so hard to check with a four-step cycle: "contractor → project → subcontracting → honorary title." But that analysis stops at "why it is this way in China." This case pushes the lens to a global scale — the United States has deeper funding, more top journals, and a more independent judiciary, yet it too has produced generation after generation of "Classmate Gengs," who are likewise ground down, marginalized, and made to disappear.

The conclusion is blunt: no number of Classmate Gengs will make a difference, because the existing machinery of academic production itself encourages fraud.

2. The Mechanism Problem: Why the Lone Warrior Always Loses to the System

The core contradiction at the mechanism level is an asymmetry of incentives:

• The fraudster's payoff: grants won, top-journal publications, titles conferred, citations from colleagues, students flocking to follow. Every item is quantifiable capital accumulation.
• The fraud fighter's cost: being ground down, marginalized, throttled by platforms, shunned by peers, cut loose by collaborators. Every item is irreversible capital depletion.

This asymmetry shows up again and again in four landmark American cases:

Duke University (2019)

Joseph Thomas, a technician in a pulmonary research lab, happened upon the fraud through a credit-card fraud case — the fabrication was not confined to ledgers; it reached into experimental data. The principal investigator used falsified data to defraud grant funding; the university shielded him and retaliated against Thomas in the workplace. Thomas ultimately sued under the False Claims Act, and in 2019 Duke University acknowledged that, between 2006 and 2018, it had applied for more than 30 research grants while knowingly relying on fabricated data, and agreed to pay the U.S. government $112.5 million. Thomas personally received a whistleblower award of $33 million to $33.75 million.

Stanford University (2023)

Neuroscientist and former Stanford president Marc Tessier-Lavigne had long been dogged by rumors that images in his papers had been digitally manipulated. It was not until Theo Baker, a 17-year-old freshman reporter for the Stanford Daily, spotted an anonymous post on PubPeer and began investigating that the matter broke. In July 2023, an independent investigation found that in four of the five papers on which Tessier-Lavigne was the lead author, data had been manipulated by others. Multiple papers were retracted, and Tessier-Lavigne resigned.

Harvard's Dana-Farber Cancer Institute (2025)

Sholto David, a British biotech industry insider and amateur academic detective, had spent years flagging problems with papers on PubPeer. In January 2024, he published a concentrated exposé on his blog: 57 Dana-Farber papers containing image duplication, 180-degree rotations, and other fabrication. Among the executives implicated were CEO Laurie Glimcher, COO William Hahn, and SVP Irene Ghobrial. In April 2024, David sued under the False Claims Act; the case settled in December 2025, with DFCI paying $15 million and David receiving about $2.63 million.

University of Minnesota (2022)

In a landmark 2006 Nature paper, neuroscientist Sylvain Lesné claimed to have discovered an amyloid-beta oligomer called Aβ*56, made up of 56 subunits, that induced memory loss when injected into mice. Cited more than 2,300 times, the paper is one of the central pillars of the amyloid hypothesis of Alzheimer's disease. Sixteen years later, a Science investigation revealed that the paper's images had been copied and pasted together from unrelated experiments. Multiple papers Lesné published over the following two decades were also found to contain manipulated images, and pharmaceutical companies and research institutions around the world spent 16 years pursuing a dead end.

3. Field Distribution: The High Visibility of STEM and the "Landfill" of the Humanities

Fraud is not uniformly distributed. Fraud in STEM is more likely to be exposed, bigger in impact, and more strictly regulated — but that is all it is.

3.1 Three Salient Harms of STEM Fraud

First, enormous economic damage. One fabricated paper by Lesné sent pharmaceutical companies around the world pouring in billions of dollars in R&D spending, only for it to evaporate into bubbles. The 57 fraudulent papers from Dana-Farber alone may involve tens of millions to hundreds of millions of dollars in federal funding. Duke alone was fined $112.5 million.

Second, threats to public health. In 1998, British physician Andrew Wakefield published a paper in The Lancet claiming that the MMR (measles-mumps-rubella) vaccine could cause autism in children. Although the paper was fully retracted by The Lancet in 2010, the anti-vaccination movement (anti-vaxxer) had already taken shape. The global resurgence of measles is a fact.

Third, the absurdity of cross-domain cons. Theranos, founded by Elizabeth Holmes, claimed that "a single drop of blood can run more than 200 tests." Former Secretary of State Kissinger and former Secretary of Defense Mattis both lent it their endorsement, and its valuation once reached $9 billion. In reality, its device ran data on modified third-party commercial machines. In 2022, Holmes was convicted on four counts and went from "the female Steve Jobs" to inmate. The fraudster did not seek peer review from the medical community; she sought endorsement from political grandees — then substituted "an endorsement by a former Secretary of State" for "FDA approval."

3.2 Low Visibility, High Permeation: Fraud in the Humanities and Social Sciences

Fraud in STEM has the hard constraints of data, images, and repeatable experiments, which is why it is more easily exposed. Fraud in the humanities and social sciences operates under a different logic — mediocre research questions, casual methods, statistical manipulation, and conclusions that are hard to falsify.

Diederik Stapel, a social psychologist at Tilburg University in the Netherlands, is a textbook-level case. He would decide on his conclusion first (for instance, "looking at photos of handsome men and beautiful women makes people feel inferior"); when his own experiments failed to produce the expected results, he fabricated the data by hand. Some studies were never conducted at all — they were fabricated from beginning to end. In the end, data fabrication was found in at least 55 of his papers and 10 doctoral dissertations. After the exposure, the entire Dutch academic community, stung into reflection, pushed through the publication of the Netherlands Code of Conduct for Research Integrity.

Even more dangerous is that "seemingly boring research questions can also influence public policy." In 2012, Dan Ariely of Duke and Francesca Gino of Harvard, along with others, published a joint paper in PNAS concluding that signing at the top of a form (committing first) makes people more honest than signing at the bottom. This conclusion was adopted directly by the U.S. Internal Revenue Service, countless government agencies, and insurance companies, which revised their declaration-form templates. Beginning in 2021, the behavioral scientists Uri Simonsohn, Joseph Simmons, and Leif Nelson continuously exposed the data fabrication in this paper on their DataColada blog. Gino was later stripped of tenure at Harvard (the first such case since Harvard established its tenure-revocation rules in the 1940s).

3.3 The "Gray Zone" of P-hacking

P-hacking sits between fraud in STEM and fraud in the humanities — it is neither "fabricating data" nor "forcing conclusions." Rather, at every step — research design, statistical procedure, sample selection — there is room to operate flexibly until the results reach the p < 0.05 threshold of "statistical significance." One researcher read through 2,578 social-science papers and found statistical methods riddled with holes, and conclusions that could not withstand scrutiny. The vast majority of the papers had sample sizes that were far too small — experiments with some 20 subjects — yet they were still published in top journals, still cited, and still shaped policy.

4. The Reproducibility Crisis: Putting Numbers on the "Anti-Movable" Mechanism

The most direct quantitative evidence of the mechanism problem comes from two large-scale replication exercises in the pharmaceutical industry:

Academia itself does not do much better. A review published in eLife in 2021 pointed out: in academia, at least more than half of cancer-biology experiments are highly likely to be irreproducible.

The harm of the reproducibility crisis does not lie in "some studies turn out to be wrong" — it lies in the fact that it is self-reinforcing. Researchers tend to cite publishable results ("positive findings"); non-significant studies get locked away in file drawers. Citation networks turn a handful of exploded "positive findings" (whether true or false) into consensus; the next generation of researchers builds new studies on top of that consensus; erroneous conclusions are stacked, verified, and cited over and over. Lesné's Aβ*56 paper has been cited more than 2,300 times — those 2,300 citations are at once proof of scientific impact and 2,300 opportunities for the error to be amplified.

5. Bazerman's Five: Actionable Reform Pathways

In his book on academic scandals, Harvard Business School's Max Bazerman writes: a great deal of academic research is nothing more than meaningless activity carried out by amateurs who want to enjoy the perks of academic status. Together with other American think tanks and scholars, he has put forward five actionable reform pathways:

1. Preregistered hypotheses + transparent data platforms. Before conducting an experiment, researchers register their hypotheses, study designs, and data-collection plans on a public platform; afterward, the preregistered plan is compared against the actual data. This cuts off the possibility of "working backward from conclusion to hypothesis."
2. A dedicated 10% replication fund. Ten percent of research funding is set aside specifically for replicating other people's studies — this 10% is the quality-control checkpoint of the research system. If the replication rate does not rise, the entire field's direction is wrong.
3. Tighten the p threshold from 0.05 to 0.005. The p < 0.05 standard was proposed in 1925, when experiments were expensive and sample sizes were small; today a single cancer study routinely involves thousands of variables, and the 0.05 threshold has become far too lax.
4. Abolish the pseudo-citation doctrine. The current "impact factor" game has papers citing one another — junk citing junk. Highly cited papers are often not good papers, but papers sitting on hot tracks where citations accrue "naturally."
5. Authors put up their own money. Paper authors must stake their own hard cash, betting that their research can be successfully replicated by a third party; if replication fails, the author loses the money. This aligns the "publication incentive" with the "truthfulness incentive."

These five are not idle fantasy. After the Stapel case, the Netherlands spent ten years pushing through the Netherlands Code of Conduct for Research Integrity, spelling out five principles — "honesty, rigor, transparency, independence, accountability" — and systematically raising the profile of research integrity. Reform is not impossible — but the precondition is that someone is willing to bear its costs.

6. Why This Path Applies to China Too

The horsepower of paper production is far greater than in the United States, yet the mechanism is more fragile.

The reason is straightforward: the American academic system has relatively independent judicial intervention (the False Claims Act lets whistleblowers like Thomas and David walk away with awards in the tens of millions of dollars), relatively independent peer review (despite the entrenched habit of "immunity for big names"), and relatively independent investigative journalism (Science's investigation of the Lesné case took nearly a year). Even so, the reproducibility crisis is still severe.

In China's system, independent judicial intervention is almost entirely absent, the independence of peer review is weaker still, and independent investigative journalism has almost no room to operate in the academic sphere. Classmate Geng's fraud-fighting path has reached a dead end in China — platform throttling, institutional self-investigation, truth and falsehood impossible to distinguish. This mirrors the same "the fraud fighter gets ground down" trajectory as the Thomas and David cases, but without the corresponding judicial rewards.