The Rapidly Growing Problem of Scientific Fraud
- In ICU
- Fri, 8 Aug 2025
Over the past several centuries, science has evolved into a critical enterprise supported by states and societies worldwide. This contract between scientists and society promises that in exchange for creating new knowledge and training skilled workers, scientists receive rewarding careers, public recognition, and resources. This model has driven remarkable growth in scientific research, closely tied to national wealth and global development. However, recent evidence reveals that this system faces unprecedented threats from coordinated scientific fraud that may fundamentally undermine trust in the scientific record.
Science functions as a complex public good relying on trust among numerous stakeholders: scientists depend on their peers and publishers for knowledge sharing and quality control; universities and funders rely on scientists for peer review; and society depends on science to generate reliable knowledge for wellbeing and security. The increasing complexity and specialisation in science amplify these interdependencies, but also make the system vulnerable to defections, i.e., failures by some participants to contribute genuine work while still benefitting from the system.
To evaluate contributions, proxies such as journal impact factors and citation-based metrics have become widespread, but these often poorly reflect true quality. This has fuelled intense competition and resource inequality, potentially incentivising misconduct. Scientific misconduct ranges from falsifying data to selling authorship slots and using so-called paper mills that mass-produce fabricated research. Surveys reveal alarming admission rates of misconduct, and investigations show that fraudulent articles may comprise a growing fraction of published work.
Studies of editorial practices at journals like PLOS ONE and Hindawi expose small groups of editors disproportionately associated with retracted or suspicious papers. These editors, making up a tiny fraction of all editors, handle a large share of problematic articles and often collaborate in tight-knit networks facilitating fraudulent publications. Similarly, analyses of conference proceedings highlight entire events flagged for high rates of questionable publications.
One hallmark of fraudulent science is the duplication of images across multiple articles. Network analyses reveal large connected clusters of papers sharing duplicated images, suggesting the existence of organised batches of fraudulent articles. These clusters tend to be concentrated within specific journals and publication years, consistent with a modus operandi where paper mills or their brokers coordinate rapid, large-scale insertion of fake papers into the literature. This batch publishing requires collusion by editorial insiders or compromised journals.
Fraudulent entities appear highly adaptive. For example, the organisation Academic Research and Development Association (ARDA) openly advertises services to guarantee publication and shows a history of journal hopping, shifting the journals it targets in response to deindexing by bibliometric databases like Scopus and Web of Science. Many ARDA-associated journals are deindexed at rates far exceeding the baseline, indicating concerns about their editorial quality.
Moreover, articles in these journals frequently fall well outside the stated scope of the journals, such as a paper about roasting hazelnuts appearing in an HIV/AIDS journal. Authorship patterns also vary: while some fraudulent papers have authorship restricted to local scholars, others display improbable international collaborations, implying different models of selling authorship slots globally.
Retractions of fraudulent papers have increased exponentially over the past three decades, doubling approximately every 3.3 years. Similarly, the volume of critical post-publication commentary on platforms like PubPeer has grown at a comparable rate. Yet the total scientific output grows much more slowly, doubling only every 15 years. Suspiciously, suspected paper mill papers are doubling even faster, every 1.5 years, and now outnumber both retracted and PubPeer-commented articles, underscoring the scale of the problem.
Despite these alarming trends, current mechanisms to combat fraud remain insufficient. Deindexing of journals occurs but affects far fewer journals than those publishing suspected paper mill content. Retractions remain relatively rare compared to the number of problematic papers, and a significant portion of fraudulent articles remain embedded in the scientific record. This persistence threatens to undermine scientific integrity both retrospectively and for future research.
Underlying these problems are systemic issues in the scientific ecosystem. Intense competition for limited funding and academic positions drives a relentless push for growth in metrics like publication count and citation impact. While a privileged few access the resources needed to achieve these goals legitimately, many scientists operate under conditions of inequality, limited opportunity, and inadequate training.
Research links such perceived injustice and lack of opportunity to increased risk of misconduct. For some, fraud becomes a rational risk-reward calculation: why invest years in uncertain research when for a small fee one can buy publications and citations? Unfortunately, the normalisation of fraudulent practices appears increasingly entrenched, especially among junior researchers in some regions or fields.
Addressing this crisis requires urgent, systemic change. Current measures, such as retractions, deindexing, contract terminations, are insufficient and unevenly applied. Moreover, conflict of interest undermines accountability: journal editors may profit from rapid publications, and institutions are conflicted when investigating their own researchers. Detection efforts have largely relied on isolated, courageous individuals who face dismissal or legal threats.
A more robust accountability system must separate detection, investigation, and sanctioning functions, removing them from parties with conflicts of interest. Substantial investment in human and technological resources is essential to keep pace with the scale of fraud. Furthermore, clearer definitions and a unified vocabulary to describe diverse fraudulent behaviours are needed, moving beyond vague terms like paper mill towards frameworks that capture the complex reality of coordinated misconduct.
Game theory’s concept of defection offers a useful lens, framing fraudulent behaviour as rational deviation from normative cooperation rather than mere ethical failure. Recognising this shift is crucial for designing interventions and changing incentives. Without concerted action, for many emerging scientists, engaging in fraud may sadly become the new norm.
Beyond immediate damage to scientific fields, large-scale fraudulent science poses risks to emerging technologies aimed at synthesising and disseminating scientific knowledge. Machine scientists and large language models, which hold promise to assist researchers and the public, currently lack the capacity to reliably distinguish legitimate research from fraudulent work. As the volume of fraudulent literature grows, these technologies may inadvertently amplify misinformation and reduce trust in scientific outputs.
The scientific enterprise stands at a crossroads. While historically known for progress and trust, the system now faces coordinated, industrial-scale fraud that exploits vulnerabilities in editorial practices, evaluation metrics, and resource inequalities. Failure to act decisively threatens to erode public trust, degrade the scientific record, and compromise future discovery. Building a more transparent, accountable, and equitable research culture is essential, not only to root out current fraud but to safeguard the future of science itself.
Source: PNAS
Image Credit: iStock
References:
Richardson RAK, Hong SS, Byrne JA et al. (2025) The entities enabling scientific fraud at scale are large, resilient, and growing rapidly. PNAS. 122(32):e2420092122
