In my years of working with leadership teams in science-based organisations, one scene repeats. The team is meeting to agree a recommendation it must issue, and the evidence itself is not in dispute. Someone talks about confidence intervals. Someone else talks about acceptable risk. The director weighs what the institution can defend in front of its board or its donors, while the operational lead thinks about what the field offices will be asked to carry. Four conversations run through one room, and whenever they collide, someone closes the discussion with the same sentence: the science is clear.

The science may well be clear. The question in front of the team is a different one: what to do. Deciding that requires judgements no finding can supply, and the sentence works in the room because it can allow those judgements to remain unnamed. Usually no one is hiding them; a long chain of reasoning has been compressed into one defensible sentence. By the time the recommendation goes public, the judgements can be hard to find at all. They have disappeared into the phrase “the evidence”.

In July, Nature published a special collection on trust in science, and one of its most useful corrections is to a story I often hear inside the institutions I work with. The picture is not one of collapse: a study of 71,922 people across 68 countries found no country with low average trust in scientists, and 78% consider scientists qualified to do consequential research. The fractures are specific. In Britain, a different survey found that 84% still report at least some trust in science, though only 34% trust it “a lot”. The two surveys measure different things, trust in scientists in one case and trust in science in the other, but they share the pattern that matters here: trust remains widespread, and much of it comes with conditions.

A less quoted number from the same study is more revealing. Only 42% said that scientists pay “somewhat” or “very much” attention to other people’s views, against the 78% who consider them qualified. The sharper doubt concerns listening.

My perspective is that of a coach, not a science-policy adviser. I work with the leaders and teams of institutions that operate where science meets policy, mostly in the UN system and in research and civil-society organisations. From this perspective, the collection’s advice is necessary but incomplete. Its strongest contributions argue for involving the public earlier in research and in policy advice. The collection says less about the internal leadership process through which expert teams turn evidence into institutional judgement. Heidi Larson and David Bersoff argue in the same collection that science’s larger problem is a loss of influence rather than a loss of trust. I would add an organisational dimension. Whether advice carries influence is partly shaped by how it is produced, and the move from evidence to institutional advice happens in meetings like the one above. Trust is given or withheld by others; trustworthiness is the part the institution can work on.

A team in that scene is doing more than weighing evidence. It is deciding how much uncertainty it can tolerate before acting, and which risks take priority. It is deciding which kinds of knowledge are allowed to inform it, and whether the practitioners who will implement the recommendation helped form it or will simply receive it. All of these are judgements, unavoidable ones. Evidence can inform them; it cannot make them for the institution. The trouble starts when they enter the decision invisibly and the conclusion is then presented as though the evidence made the choice unaided. “Follow the science” becomes misleading at exactly the moment it hides the judgement about where science should lead. What a team has not distinguished internally, it will struggle to explain publicly.

Trust in evidence is not the same as agreement with policy. People can accept a finding and still dispute the risk threshold chosen in response, or the authority of the institution acting on it. From inside an expert organisation, that disagreement is easily classified as rejection of the evidence. The classification can flatter the institution: it locates the problem in the audience. The 42% is consistent with another possibility: people may respect the expertise and still doubt that the institution has listened.

What can a leadership team do with this? The teams I would describe as trustworthy share a small set of disciplines. They separate the layers of a decision, so that everyone in the room can tell evidence from interpretation, and both from the judgements the institution is making on its own authority. And they take internal disagreement as information about the decision, because a team that treats every objection as resistance to the evidence will do the same to the outside world.

They also keep a reasoning trail an outsider could follow: what was known and assumed, and what would have changed the recommendation. The UK’s Joint Committee on Vaccination and Immunisation, for example, publishes its advice together with detailed statements and meeting minutes. That does not end disagreement; it makes part of the reasoning available for scrutiny. The fourth discipline is the least practised: the knowledge of the people who will implement a recommendation, and of those who will live with it, enters the room before the answer settles, instead of arriving afterwards as a communications problem.

In practice, these disciplines can be tested with six questions, asked before the recommendation leaves the room:

What do we know? What are we inferring? What remains uncertain? What are we choosing rather than discovering? Who will carry the consequences? What would cause us to revise?

A team that can answer these six questions has made its judgement legible, and given anyone outside the room grounds to decide whether trust is warranted.

I wrote in the first of these notes that the quality of a team’s conversations is one of the few variables still largely in its own hands. This is where that claim meets the outside world. A leadership team cannot control the political weather, or the histories that shape who is inclined to trust it. It has more agency than it sometimes recognises over how its own advice is formed, and the way an institution reasons on the inside eventually becomes visible on the outside.

Thinking well together will not repair the information environment. It can make the institution’s own contribution more honest. A team can name where evidence ends and judgement begins, and it can bring the people who will carry the consequences into the room before the answer is fixed. None of that guarantees trust. It is the part of deserving trust that lies within the team’s reach.

Sources: the Nature special collection on trust in science (July 2026); V. Cologna et al., “Trust in scientists and their role in society across 68 countries“, Nature Human Behaviour; More in Common, “Britain Talks Trust in Science“; H. Larson and D. Bersoff, “Science’s big problem is a loss of influence, not a loss of trust“, Nature.