The Weaponization of “Scientific Consensus”

by Roger Pielke Jr.

In September, 2022 California Governor Gavin Newsome signed into law a bill that prohibited medical professionals from sharing “misinformation” with patients. Specifically, the law stated that it would be

[U]nprofessional conduct for a physician and surgeon to disseminate misinformation or disinformation related to COVID-19.

The law defined “misinformation”:

“Misinformation” means false information that is contradicted by contemporary scientific consensus contrary to the standard of care.

The law reflected a common perspective: A scientific consensus represents truth and views outside that consensus are misinformation. Thus, to identify those who are spreading misinformation we simply need to identify the relevant scientific consensus. Those out of step with the consensus, the argument continues, can then be called out or sanctioned for spreading misinformation — and public discourse can proceed based on accepted facts, not falsehoods.

The notion of consensus-as-truth has been operationalized in various forms: journalistic “fact checkers,” academic “misinformation” researchers, and content moderation on social media platforms. The practical effect is the creation of self-appointed arbiters of truth — journalists, academics, social media platforms, and even governments — who render judgments on acceptable and unacceptable speech according to conformance with an acceptable view.

There are many problems with the notion of consensus-as-truth and the (self)appointment of misinformation police to regulate discourse, whether of the public or, as in the case of the California law, of experts themselves.

A scientific consensus is not a single view, but a distribution of views. Almost 20 years ago I participated in an exchange in Science with Naomi Oreskes on this point. Professor Oreskes shot to fame by publishing a commentary that argued that the consensus on climate change was universal, based on a review of 928 papers. Oreskes argument quickly moved from characterizing science to a call for political action, based on the asserted universal consensus.

I responded by arguing that a consensus is not a single thing, but a distribution, and policy should be robust to that distribution:

The actions that we take on climate change should be robust to (i) the diversity of scientific perspectives, and thus also to (ii) the diversity of perspectives of the nature of the consensus. A consensus is a measure of a central tendency and, as such, it necessarily has a distribution of perspectives around that central measure. On climate change, almost all of this distribution is well within the bounds of legitimate scientific debate and reflected within the full text of the IPCC reports. Our policies should not be optimized to reflect a single measure of the central tendency or, worse yet, caricatures of that measure, but instead they should be robust enough to accommodate the distribution of perspectives around that central measure, thus providing a buffer against the possibility that we might learn more in the future.

A further complication is that the notion of a “consensus on climate change” is incoherent. In a 2011 study of how the Intergovernmental Panel on Climate Change (IPCC) represented uncertainty in its Fourth Assessment Report, Rachael Jonasson and I identified 2,744 “findings” across the AR4 report — each finding was a specific scientific claim.

There was a degree of consensus associated with each of those claims — the distribution of views may have been narrow (e.g., climate change is real), bimodal or wide (e.g., future hurricane incidence), and with the advantage of hindsight, utterly wrong (e.g., a high emissions scenario is business-as-usual).

Last week, the Global Catastrophic Risk Institute (a non-profit funded by a group called Social and Environmental Entrepreneurs) released a survey of expert views on the origins of COVID-19. The results of that survey were quickly invoked on social media to claim support for whatever position that was held before the survey was released.

The far more important result of the survey was not what it says about expert views on the origins of COVID-19, but what it says about the futility of trying to use surveys of experts as a short-cut to truth.

Consider:

• 1,138 experts were invited to participate in the survey, 184 chose to respond;
• About 65% of those who responded believe that there is a >10% that COVID-19 resulted from a research-releated incident;
• About 23% believe that there is a <5% chance that COVID-19 resulted from a research-releated incident.

There is a little something there for everyone, and the results can be spun this way or that. But it gets more interesting. The survey asked participants whether or not they were familiar with key studies related to Covid origins. The results are shocking:

• 43% of “experts” acknowledge that they had never heard of Proximal Origins;
• 78% were unaware of the DEFUSE grant proposal;
• And perhaps most incredible, 33% of respondents claimed familiarity with a study that does not exist (Hanlen et al. 2022).

You can see the full results in the table below. It would have been quite interesting to have seen the results of the survey according to familiarity with the relevant literature.

 
To the extent that the GCRI survey conveys useful information, it tells us that there are an extremely wide range of views on the origins of COVID-19, and some but not all of those views are well informed among credentialed experts. Less charitably, the survey tells us that identifying experts as a proxy for evaluating information — that is, actually doing the hard work of science, forensics and investigation — is problematic, as many of these “experts” are no such thing.

The notion of consensus-as-truth can create obstacles to improving understandings. In re-reading Oreskes 2004 on climate consensus for the first time in a while I was struck by this comment:

This analysis shows that scientists publishing in the peer-reviewed literature agree with IPCC, the National Academy of Sciences, and the public statements of their professional societies.

This is completely backwards — scientific assessments are an interpretive snapshot of what a scientific literature says about specific scientific claims. When done properly, they are a useful characterization of what is often a large amount of published research. But make no mistake — the scientific literature does not “agree” with assessments, the literature informs the assessments.

The notion that scientists should agree with a consensus is contrary to how science advances — scientists challenge each other, ask difficult questions and explore paths untaken. Expectations of conformance to a consensus undercuts scientific inquiry. It also lends itself to the weaponization of consensus to delegitimize or deplatform inconvenient views, particularly in highly politicized settings.

I have witnessed the weaponization of consensus many times. For instance, I have been asked in Congressional hearings and by colleagues whether I “endorse” the IPCC consensus, as a sort of litmus test. The correct, simple answer is that my work is cited in all three AR6 IPCC Working Groups and is thus a part of the IPCC consensus. A more complex answer is that the IPCC makes thousands of findings and my expertise allows me to make informed judgments on dozens or hundreds of them — And based on my expertise I conclude that for most of its findings the IPCC correctly summarizes the literature (e.g., human-caused climate change is warming the earth and presents significant risks) but for some other claims, the IPCC does not accurately represent current understandings (e.g., there are not more intense hurricanes and RCP8.5 is not BAU).

A recent study of scientific censorship by scientists by Park et al. 2023 finds that pressures by scientists on their peers to conform to a consensus are fairly common within the scientific community:

Read more