How to Read a Consensus

A scientific consensus is not a vote, and it is not an authority claim. Both readings are common and both are wrong in the same way: they treat the number of people who agree as the thing carrying the weight.

What actually carries the weight is convergence. When several independent lines of evidence (different methods, resting on different assumptions, vulnerable to different errors) arrive at the same answer, the probability that they are all wrong in the same direction becomes small. That is the structure of a strong consensus, and it is why a hundred studies sharing a single flawed assumption count for much less than four genuinely independent ones agreeing.

This gives a non-specialist a question they can actually answer. Not who is more qualified, and not how many signed a statement, but whether the supporting evidence is independent. Four thermometers of the same faulty design agreeing tells you about the design. A thermometer record, a satellite measurement, a borehole temperature profile and glacier retreat agreeing is a different kind of evidence entirely, because no single fault is common to all of them.

The reason this matters practically is that an industry exists to obscure it. A 1969 tobacco industry memorandum stated the strategy with remarkable candour: doubt is our product, since it is the best means of competing with the body of fact. The aim was never to win the scientific argument, which was already lost. It was to sustain the appearance of a live dispute long enough to delay regulation.

The tactics are consistent wherever the approach is used: fund contrarian work, amplify genuine uncertainty at the edges of a field to suggest the core is unsettled, demand a standard of certainty that empirical science never provides, and promote a small number of credentialed dissenters as evidence of ongoing debate. The approach has been reused on lead, acid rain, ozone and carbon, in several cases by the same consultancies and occasionally the same individuals.

The asymmetry is what makes it work. Establishing a position requires evidence; making one look unsettled requires only noise. And the strategy does not need to win, only to delay, which changes what counts as success.

Which leaves scientists with a genuinely hard communication problem rather than a simple one. State a range and you invite the reply that nothing is really known. Omit it and you invite the charge of overclaiming as soon as the estimate moves, which it will. Both failure modes are exploitable and both are on display in every contested area.

What tends to work is giving the range together with an explicit statement of what is and is not in dispute, and saying in advance what evidence would change the estimate. What reliably fails is presenting a central figure as settled and then revising it, because the revision is then evidence of unreliability rather than of a functioning process.

The deepest difficulty is that this is all a matter of degree, and public argument is conducted in categories. Science produces confidence intervals; decisions require yes or no. Nothing dissolves that gap, and the best anyone manages is to be clear about which side of it they are speaking from.