Who Defends Science at Scale? And How?
Scientific Intelligence Needs Infrastructure
A recent Chemical & Engineering News article highlighted the growing community of scientific sleuths uncovering problematic research and large-scale fraud in the scholarly literature. Our work has been essential. And it also exposes a deeper problem.
The modern research ecosystem still relies heavily on individuals acting as watchdogs rather than on systems designed to detect problems at scale. While scientific sleuthing may have revealed systemic cracks, it’s also revealed the absence of infrastructure.
Florence Nightingale’s acted as an individual to bring about policy change. The data visualizations were used to translate the terrible conditions of the army in the Crimean war for the British politician understandings.
Here are a few places where that infrastructure needs to exist – and these are pointing to the need for a society, which is what FoSci is becoming.
1. Ethics and the Opportunistic weaponization of Scientific sleuthing
During a recent Bluesky discussion about the C&EN article, Dr David Bimler (aka Smut Clyde) posed a question about a comment I made in the piece: what evidence do we have that scientific sleuthing is being weaponized?
It is a fair question.
Here is one example. As reported in Inside Higher Ed , Black scholars faced waves of accusations tied to anti-DEI campaigns, raising concerns that misconduct allegations were being selectively deployed as part of broader political battles in academia.
Selecting researchers based on political beliefs, race, ethnicity, gender, or ideological position and then searching their work for errors is not scientific accountability. It is using the tools of integrity as a dagger to target opponents and advance an agenda. This is opportunistic, selection bias used for ideological means.
The power of sleuthing and forensic scientometrics lies in its ability to expose manipulation, fraud, and systemic weaknesses in the scientific record and environment.
This is precisely why norms and ethics matter.
FoSci must be grounded in principles as with any other rigorous field; within our group, those norms are being developed. But let me be clear about the above statement. I don’t just lay out norms and expect others to adhere to them. This is a collective group of people who all work very independently and that culture is important to uphold.
2. Funded Organized Research Investigations: Individuals cannot keep up with industrial-scale deception
Many of the most prominent sleuths/investigators work one paper at a time. Given, some work very quickly. This approach has uncovered remarkable cases of misconduct and manipulation. But it cannot keep pace with the scale of the modern publishing ecosystem.
Paper mills, citation cartels, and coordinated authorship networks operate industrially; this was true even before the recent uptake of Generative AI tools. They produce large volumes of papers across journals, preprints, disciplines, and national borders. Detecting these patterns requires something different than looking at individual papers, which is what we’ve been able to do with Dimensions Author Check (disclosure alert: this is a product of Digital Science, my employer, and I helped develop the tool).
Looking at one scholarly paper at a time is like trying to treat a pandemic by describing one person at a time. That part is critical in the early stages of an outbreak but insufficient to protect society at large. (See the history of HIV/AIDS in the US.) And in the age of social media and AI, let’s be honest, understanding networks will be critical to upholding scientific trust.
It requires the ability to analyze large bodies of connected scholarly data simultaneously – identifying unusual collaboration networks (or no collaboration networks), suspicious affiliations (or no affiliations), abnormal citation patterns, or improbable publication trajectories. The banking industry does this – those ‘fraud alerts’ are from understanding trust patterns and deceptive techniques.
In other words, it requires infrastructure.
This need for infrastructure presents a bit of a dilemma. First, these problems should be better internally governed by funders, publishers, and institutions – most of them have not sufficiently kept their own houses in order. Second, though, even if each of these sectors did a better job of housekeeping, manipulation of research happens across the sectors. Yet, there are tools and companies (my employer Digital Science included, with Dimensions, Author Check, and Dimensions Research Security).
The field of forensic scientometrics attempts to do exactly this: move integrity and security (see next item) investigations from isolated cases to pattern detection across the research ecosystem. Yet, we need funding and independence to run. One thing, however, is there is already a community of investigators and those will be the ones to be core to the infrastructure.
3. FoSci Mandate - Protecting Knowledge Infrastructure: Research integrity and research security are converging
For many years, research integrity and research security have been treated as separate domains. Integrity focused on misconduct: fabrication, falsification, plagiarism. There is the growing detection of coordinated misconduct with paper mills and authorship-for-sale.
Research security has focused on issues like intellectual property theft, now it has expanded into dual-use technology and critical technology with regulations growing.
But the boundaries between these areas are increasingly blurred.
Paper mills, coordinated publication networks, and manipulated citation systems can distort the scientific record in ways that influence policy debates, funding priorities, and technological development. They can also be leveraged for geopolitical agendas. In other words, the manipulation of scientific literature can have strategic consequences.
As science becomes more central to economic competitiveness and national security, protecting the integrity of the research ecosystem becomes part of protecting the broader knowledge infrastructure of society.
The convergence of integrity and security is already seen in some governments. In Australia, for example, there is growing recognition that the research ecosystem itself is a target for broader strategic exploitation. In an article from October 2025, Associate Professor in Law Brendan Walker-Munro challenges his country to do more to stop foreign spies. Australia is notably ahead here, moving quickly on research security. The nation can also boast a mature research integrity landscape, likely accelerated by sustained pressure from experts working at the intersection of these domains and the geographical proximity to adversarial countries.
Thus, the mandate of FoSci is broader than publishers or institutions. Science is central to national security and democratic decision-making. Perhaps there needs to be a body to protect the integrity of the research ecosystem as a societal concern. But at a national level? No. We are already an international collective. Yet finding local money to support international efforts is tricky.
Infrastructure matters for something much larger: the credibility of science in public life, in government decision-making, and in democratic societies that rely on scientific knowledge to function. If trusted science is a pillar of democracy then it needs defenses that operate at the same scale as democracy itself.
As Robert K. Merton wrote in 1942 during a previous time of societal upheaval: “Difficult as the notion may appear to those reared in a culture that grants science a prominent if not a commanding place in the scheme of things, it is evident that science is not immune from attack, restraint, and repression.”
Science underpins public policy, technological development, economic growth, and democratic decision-making. Protecting the integrity of the scientific record is therefore not only an academic concern, it’s a societal one.
Scientific sleuthing showed us the problem. Now we need the infrastructure to address it.
Other Reads
A few recent articles that might interest you.
A new article in Accountability in Research on, well, accountability in research: ethicists David Resnik and Mohammad Hosseini “… argue that GenAI hallucinated citations might qualify as a provable instance of research misconduct under the U.S. federal regulations”. Read more here.
An interesting thread on LLM-generated content going to researcher Carl Bergstrom, effectively shutting down potential communication with inquisitive people. In part of the thread this onslaught of questions is linked to DDOS attack on scientists. Read more on Bluesky.
- In David Hansen’s blog, AI and the False Promise of Control, he points out some of the seemingly ridiculous requests of AI systems. But the requests don’t seem out of proportion to me. David draws on and reacts to the University of Virginia Archival AI Protocol led by the Dean of University Libraries, Leo Lo. And Leo is taking a very applied and thoughtful approach to responsible AI.
General Security News (and it’s not yet April 1)
Possum Found Nestled in Plush Toys at Airport Gift Shop
Was this store past security? Do we know if it brought in liquids? So many questions to be answered.
