Scientific consensus isn’t the same as consensus in politics, in business, or in deciding where to go for dinner. The conflation of the scientific and lay senses of the term are, I think, a primary cause of much of the general public’s distrust of the conclusions that scientists draw from their work.
NOTE: I was prompted to post this here after a longish commentary that I posted on Google+, based on an article in Scientific American about scientific consensus, became unexpectedly popular. I’ve actually written about this (briefly) before here and here, but my recent G+ post and the encouraging remarks I’ve received have prompted me to write more about it here.
If you read the SciAm piece, you will find that the authors, and those quoted by the authors, all assume that a “scientific consensus” is just a conventional consensus arrived at by scientists. But that’s not right. Like other terms used by scientists (e.g., theory1), “consensus” is a different critter when scientists reach it than when lay folk reach it. This is not a value judgement on how non-scientific consensus is reached; it’s just the way it is.
Conventional consensus (like political consensus, or organizational consensus) is often arrived at through argument and rhetoric, but not necessarily with evidence. Consider the public discussion around abortion, or legalization of marijuana, or euthanasia, or gun control, or creationism in science class, or whether a city should develop subway-based or surface-based public transport. In how many of these cases is privilege given to actual evidence? None. Sure, evidence has a role; but it’s largely the role of raw material, to be moulded and shaped to suit the agendas of the participants in the discussion, who usually have goals that preclude their accepting the evidence at face value. Even if you’re just interested in going out with friends for dinner, you’re unlikely to follow the evidence exclusively but rather just use it to reach your real goal. (Please note that this isn’t necessarily a bad thing. If your goal is to just spend quality time with your friends, then you may join the consensus to go to a restaurant you don’t like. There’s nothing wrong with that if you genuinely value time with your friends over the evidence of bad food.)
Notice the importance of social interaction and cooperation in the construction of conventional consensus. Town halls; debates; consultations; conferences and symposia. Even when done in the most sincere, authentic, and ethical way, building conventional consensus is about swaying those in disagreement. Thoughts and ideas are expressed and refined and posed in a variety of different forms, each with the goal of finding that magical combination of words that will convince the other to agree, regardless of what the evidence says. Parents will also relate to this: a parent will appeal to all manner of rhetoric to convince their child to eat their vegetables, or to brush their teeth; it keeps the peace if it can be achieved. Of course, the parent would never explain all this ulterior reasoning to the child, because it would undermine the attainment of the consensus.
Lay consensus can be good. It can help create a sense of community, it can strengthen the integrity of the citizenry, it can help rally support that leads to action on important and beneficial matters. It can of course also be bad. It can be a form of brainwashing, of indoctrination, of dogmatization of ideology. Like all tools, consensus is only as good or as bad as its use.
Scientific consensus is different.
Rhetoric plays a small (I’d say negligible) role in scientific consensus. The language of science, as evidenced by the writing style of scientists in peer-reviewed publications, is dry and direct, often lacking in any features that might make reading it enjoyable – except for the facts and evidence contained therein. Indeed, when I review papers, one of my basic criteria for quality is the directness and dryness of the prose. As soon as one starts reading florid language steeped in rhetoric, I immediately question the quality of the actual work. If the work is good enough, then it should stand on its own, unadorned by rhetoric.
Argumentation matters more than rhetoric in scientific consensus. Largely grounded in both deductive and inductive inference as well as mathematical derivations, argumentation in science infers new claims that might eventually form new models or theories or even laws. These inferences are such that they eventually allow predictions to be made, which in turn become the foundation of experiments. If sufficient experiments support the inferred claims, then those claims and the theories/models/laws they generate become part of the scientific body of knowledge.
But nothing beats evidence; it is the primary foundation on which scientific consensus is built. Evidence is not just an observation, but a collection of carefully collected observations, done in different ways, and by different people working independently. The evidence is shared, to be sure, but only for the sake of validating it using well-understood, rational, and verifiable techniques. Once the evidence reaches a sufficient level of robustness, which scientists are able to calculate, it becomes part of the scientific consensus.
Scientific consensus is what happens when a big pile of evidence (and all the information on how that information was collected and analyzed) is put before a group of scientists, who then work in relative isolation to recheck everything. Different scientists will use different techniques to perform these checks. New experiments done in new ways may be run to validate existing data. From all this work, various conclusions are drawn. But even those conclusions are, in a way, data; they are the results of rational argumentation from the evidence, and are subjected to the same scrutiny as the original evidence itself. When many scientists, each using their own methods, can all independently verify the evidence that they’re given, and then all reach the same conclusions by using their own, different analytic techniques – that’s a scientific consensus. Indeed, a scientific consensus can be reached (at least in principle) without any of the participating scientists doing any more than exchanging evidence. There need be no significant collaboration, no social interaction, at all.
So, I hope it’s clear by now that “scientific consensus” is different in very significant ways from conventional consensus. And, as far as I’m concerned, scientific consensus is far more robust than any other type of consensus.
Are mistakes made? Of course they are. Scientists are, after all, only human. Is new evidence ever found that invalidates some or all of the scientific consensus on a certain matter? Of course it is. Scientists are, after all, not omniscient. But the system is rigged to be self-correcting, and over time the errors will be eliminated, and the new evidence will be incorporated as necessary. This means that scientific consensus can change over time. This is as it should be, because scientific consensus is driven first and foremost by evidence, and as that changes, the consensus must also change to remain relevant.
A problem emerges, though, from people conflating scientific and non-scientific consensus. One example of this is apparent in the SciAm article linked above. In it, Richard Tol, a “climate change economist” (whatever that means), seems to be arguing from an assumption that scientific consensus is just like any other kind of consensus. Indeed, the whole SciAm article seems to waffle its way through that crucial distinction too. This is especially troubling as Tol, though not a scientist, has certainly undertaken to make his views well known to researchers and the public at large, and if a magazine like Scientific American plays into it too, the knock-on effects in the public sphere are magnified.
This results in too many people thinking that scientists don’t know what they’re doing; after all, they keep changing their minds. While I can understand how people may come to have such views, I have absolutely no patience with them. They are both ignorant of how science works, yet arrogant enough to believe they know better. Here’s one area where science journalism must help. (Yes, Scientific American, I’m looking at you.) Science isn’t science if it doesn’t change with new evidence. It’s up to people who know both about science and how to write well, to keep the public informed about these matters, about why scientific consensus is meaningful and important, and to ensure that the ignorant, arrogant anti-science crowd is marginalized.
UPDATE: 9 August 2014: Jerry Coyne has posted about a case that supports this notion of scientific consensus. The specifics relate to a letter signed by 139 population and evolutionary geneticists to the New York Times, regarding the scientifically impoverished content of Nicholas Wade’s fairy tale book about “race.” How did these scientists all come to agree on their verdict of Wade’s book? It was not because they all got together and talked it out until they’d reached an agreement. It was because they all – working essentially independently of each other on different aspects of genetics – reached certain common but independent conclusions that were driven by verified analysis of evidence.
- The term theory has a scientific meaning that is different than how the general public would use the term. In the lay language, a theory is more like a guess or a hypothesis. “I’ve got a theory,” says Anya in Once More With Feeling, “It could be bunnies.” Or, as creationists will often – and wrongly – argue, “evolution is just a theory.” Theories in science, like the Theory of Relativity and the Theory of Gravity, are explanations of how natural phenomena occur. Theories are complements to scientific laws, which are descriptions of natural phenomena. So, Newton’s Law of Gravity describes the forces that gravity exerts, while Einstein’s Theory of General Relativity explains how gravity arises from the curvature of space-time. Both laws and theories in science must be verified through multiple instances of repeatable controlled experiments that aim to falsify the proposed law or theory. As evidence mounts, scientists can actually calculate the odds that they’re wrong. For instance, the Higgs Boson’s discovery is robust to a level of “5 sigma” or about one in 3.5 million. It might surprise people to learn that this level of certainty is considered unethical in medicine, and that new drugs are released for use with far lower levels of certainty. This is not, however, evidence of some Big Pharma conspiracy, but rather simple statistics; one must trade off the odds of harm at lower levels of certainty from the possibility of error against the harm at higher levels of certainty from depriving people of the drug while the appropriate tests are carried out. In any case, the point is this: validated scientific theories are just as robust and just as “correct” as one can get in science; they are as good as humanity has got at explaining how the universe works.