A promising finding, an enthusiastic press release, then a headline announcing a revolution. Within hours, a laboratory observation becomes the promise of a treatment or technology. The problem is not always a scientific error: it is often the disappearance of caveats from one stage to the next. In the media landscape of September 2026, learning to say “we do not know yet” is a major communication challenge. Not to dampen curiosity, but to give it a reliable direction.
Discovery moves more slowly than the story told about it
Research rarely produces clear-cut answers on the first attempt. It gathers evidence, tests hypotheses and refines its instruments. Media narratives favour a different sequence: a problem, a discovery, a solution. Between the two, preliminary findings occupy a delicate position. Interesting enough to report, they are not necessarily robust enough to guide a medical, industrial or political decision.
Preprints illustrate this tension. Released before peer review, they allow teams to share their work quickly. Their visibility increased during the Covid-19 pandemic, particularly through the bioRxiv and medRxiv platforms. This process accelerates scientific exchange, but also makes conclusions still under debate accessible to the general public. And peer review, when it takes place, is no guarantee of infallibility either.
The challenge, then, is less about choosing between speaking out and staying silent than about making the status of what is being reported clear. An appealing hypothesis, an isolated finding and a phenomenon replicated by several teams do not warrant the same degree of confidence. Yet on a phone screen, they can receive exactly the same presentation.
Waves of hype that teach caution
In the summer of 2023, LK-99 provided a textbook example. Researchers presented the material as a possible superconductor at room temperature and ambient pressure. The prospect was spectacular: it appeared to open the door to far-reaching applications. Replication attempts and subsequent analyses, however, did not confirm the claim. The episode showed how a promise can spread much faster than it can be verified.
The lesson is not that early findings should remain secret. Replication attempts, objections and explanations also made the scientific process visible. But the public had to piece together a fundamental distinction: “a team claims to have observed” does not mean “science has established”. This distinction should appear in the initial story, not just in its correction.
Artificial intelligence poses a similar difficulty. AlphaFold represented a major advance in predicting protein structures. Yet predicting a structure does not amount to fully understanding how a protein works, let alone having a drug available. Between computational performance, experimental validation and clinical benefit, many stages remain. Describing them does not diminish the advance: it allows its significance to be assessed precisely.
Humility, a skill that can be developed
Saying “we do not know yet” can feel like an admission of weakness. A researcher is making the case for funding, a laboratory director is trying to attract talent, a start-up is preparing a funding round. Each may fear that a caveat will be interpreted as a lack of conviction. This is precisely where soft skills come in: resisting inflated claims, listening to expectations and explaining a limitation without becoming defensive.
A useful answer does not stop at acknowledging a lack of knowledge. It defines its scope. We observe an effect under these conditions, but we do not know whether it persists under others. We have identified an association without demonstrating a causal link. We achieve good performance on this dataset without knowing how the system will behave elsewhere. Uncertainty becomes understandable when it concerns a concrete question.
Four points to consider before an interview
- The finding: what was actually measured or observed?
- The level of evidence: an exploratory study, a controlled experiment, an observation or a synthesis of several studies?
- The main limitation: what could change the interpretation?
- The next test: what experiment would help reduce the uncertainty?
This preparation helps researchers respond without reciting a list of caveats. It also makes cooperation with journalists easier: a clearly defined finding can make a good story without having to be turned into a historic breakthrough.
Explaining without overwhelming the reader
Precision does not mean piling up technical terms. When discussing a medical trial, explaining that the study involves a small group and focuses primarily on safety can be more illuminating than immediately detailing every statistical analysis. In research on AI, specifying that the model was tested on selected data helps explain why its effectiveness in everyday use remains to be demonstrated.
Different sources of uncertainty must also be distinguished. A shortage of participants is not the same as disagreement between studies. An imprecise measurement is not the same as a poorly defined hypothesis. Depending on the problem, the response will be to collect more data, control the experiment more carefully or reconsider the question. Explaining this distinction gives readers something more than a vague warning to “treat with caution”.
Care must be taken, however, not to manufacture artificial doubt. Not everything is equally uncertain. Open questions in climate research do not invalidate firmly established knowledge about human-caused warming. Honest communication therefore requires a twofold approach: highlighting the weaknesses of a new finding and recalling the established knowledge on which it rests.
Responsibility does not rest with the researcher alone
The press release is often the tipping point. If its headline promises a treatment when the experiment involves cells grown in a laboratory, caveats placed in the final paragraph will struggle to undo the initial impression. Communications teams would do well to incorporate limitations into the opening lines, then have scientific wording checked as carefully as the figures.
In newsrooms, a simple method is to seek an independent opinion, check publication status and ask what conclusions the study design actually supports. The headline must stand up to this scrutiny. It is better to explain why a line of inquiry is worth pursuing than to promise what it might, perhaps, deliver several years down the road.
Looking ahead to September 2026, one prospective hypothesis deserves particular attention: the proliferation of AI-generated summaries could exacerbate the loss of nuance as stories are repeatedly repackaged. A caveat included in the source article may disappear in a short summary. This makes the case for limitations expressed simply, close to the finding, rather than relegated to a secondary note.
What now? The next desirable development is not more timid science communication, but better-structured communication. Interview training, press releases that distinguish findings from interpretation, and clearly visible updates could make uncertainty more familiar. Trust will not emerge from a promise of permanent certainty. It will also be built through this sentence, provided it is completed: we do not know yet, and here is how we will investigate.
