Teach early-career researchers to navigate distrust and disinformation

The image of science as an ‘ivory tower’ is constantly being challenged nowadays. Researchers work in an environment where fabricated papers can enter the literature, political actors can determine which research questions are acceptable, and paper mills and generative AI can produce unreliable text at unprecedented speed. These pressures create a threat to the ‘epistemic security’ of research.

Epistemic security means protecting the people, practices and institutions that allow knowledge to be created, scrutinised, corrected and communicated. It depends on both research integrity and academic freedom. The academic system cannot produce trustworthy knowledge if misconduct is ignored, scholars fear harassment, or institutions avoid sensitive questions.

Research misconduct

The sharp rise in retractions in recent years reveals both the strength and the vulnerability of science. Retractions demonstrate that science can correct itself, but correction often comes late. By then, unreliable findings may already have been cited, circulated and incorporated into new studies. More than 10,000 research papers were retracted in 2023 alone; these papers had received more than 35,000 citations by 2024. The original claims can therefore travel much further than the notices explaining why they should no longer be trusted. Media and publishing platforms can reinforce this imbalance by giving striking findings far greater visibility than their subsequent retractions.

A quick look at Retraction Watch reveals the range of misconduct involved: fabricated credentials, falsified findings, hallucinated references, manipulated peer review and even fake authorship. Reports on PubPeer also show how frequently researchers identify duplicated images, unexplained inconsistencies and other concerns in the published literature. Retractions represent only the most visible part of the problem. A much larger body of research may be poorly designed, weakly reported or rhetorically misleading without meeting the threshold for formal retraction. Such papers remain part of “normal science”, where they can be cited, taught and used as foundations for further studies.

Generative AI raises the stakes. The central question is not whether a text is human or synthetic, but whether its claims are true, traceable and supported. Much attention has been devoted to detecting whether a text was written by AI. This has also contributed to the rise of “AI accent-ism”, which can be defined as an ideological attitude towards words and expressions associated with AI writing. However, the more important question should be: “Is this text, and the research reported in it, verifiable?” Researchers need habits of verification that apply regardless of who, or what, produced the text.

Navigating distrust and disinformation

Early-career researchers are especially exposed in these scenarios. They rely on published literature to learn disciplinary norms, locate research spaces, and design studies, yet they may receive little preparation for recognising unreliable work. They also hold less power when authorship, supervision or integrity disputes arise. Universities therefore need to treat critical reading, research integrity, academic freedom and researcher safety as connected parts of doctoral education.

One practical approach is to work with retracted papers as an entry point. Rather than presenting science only through polished success stories, educators can invite doctoral researchers to examine how studies fail and how correction works. Misconduct can leave rhetorical warning signals: sources that are fabricated or misrepresented; methods that sound elaborate but omit essential detail; inconsistencies among claims, results, figures and conclusions; exaggerated certainty; or distorted terminology, including “tortured phrases” created to disguise copied or machine-generated text.

Epistemic security

None of these signals proves misconduct. They are prompts to slow down and use further verification strategies. This “detective” reading should create informed trust. That preparation should begin early, before publication pressures harden habits, and continue throughout research careers, because epistemic security is a shared responsibility rather than a one-off compliance exercise or checklist.

Additionally, universities must support that work institutionally. They should make correction notices visible, provide clear guidance on responsible AI use, train supervisors and reviewers, protect scholars facing harassment, and communicate openly when research goes wrong. They should also tell researchers what support exists, which public platforms carry particular risks and how threats will be handled. The recent Ctrl-Z award for self-retractions is a great example of how integrity can be supported.

Protecting science requires more than fact-checking its final products. It requires guarding the people, practices and freedoms that make reliable knowledge possible. If early-career researchers learn how distrust is manufactured, how misconduct travels and how academic freedom can be weakened, they will be better equipped to interrupt disinformation before it becomes part of the evidence base. Science needs critical readers as much as productive authors, and universities must train and protect both.