Blogger Andrew Oh-Willeke of Dispatches from Turtle Island pointed me to an editorial in Science about the phrase scientific consensus.
The editorial argues that by referring to conclusions like the existence of climate change or vaccine safety as “the scientific consensus”, communicators have inadvertently fanned the flames of distrust. By emphasizing agreement between scientists, the phrase “scientific consensus” leaves open the question of how that consensus was reached. More conspiracy-minded people imagine shady backroom deals and corrupt payouts, while the more realistic blame incentives and groupthink. If you disagree with “the scientific consensus”, you may thus decide the best way forward is to silence those pesky scientists.
(The link to current events is left as an exercise to the reader, to comment on elsewhere. As usual, please no explicit discussion of politics on this blog!)
Instead of “scientific consensus”, the editorial suggests another term, convergence of evidence. The idea is that by centering the evidence instead of the scientists, the phrase would make it clear that these conclusions are justified by something more than social pressures, and will remain even if the scientists promoting them are silenced.
Oh-Willeke pointed me to another blog post responding to the editorial, which has a nice discussion of how the terms were used historically, showing their popularity over time. “Convergence of evidence” was more popular in the 1950’s, with a small surge in the late 90’s and early 2000’s. “Scientific consensus” rose in the 1980’s and 90’s, lining up with a time when social scientists were skeptical about science’s objectivity and wanted to explore the social reasons why scientists come to agreement. It then fell around the year 2000, before rising again, this time used instead by professional groups of scientists to emphasize their agreement on issues like climate change.
(The blog post then goes on to try to motivate the word “consilience” instead, on the rather thin basis that “convergence of evidence” isn’t interdisciplinary enough, which seems like a pretty silly objection. “Convergence” implies coming in from multiple directions, it’s already interdisciplinary!)
I appreciate “convergence of evidence”, it seems like a useful phrase. But I think the editorial is working from the wrong perspective, in trying to argue for which terms “we should use” in the first place.
Sometimes, as a scientist or an organization or a journalist, you want to emphasize evidence. Is it “a preponderance of evidence”, most but not all? Is it “overwhelming evidence”, evidence so powerful it is unlikely to ever be defeated? Or is it a “convergence of evidence”, evidence that came in slowly from multiple paths, each independent route making a coincidence that much less likely?
But sometimes, you want to emphasize the judgement of the scientists themselves.
Sometimes when scientists agree, they’re working not from evidence but from personal experience: feelings of which kinds of research pan out and which don’t, or shared philosophies that sit deep in how they conceive their discipline. Describing physicists’ reasons for expecting supersymmetry before the LHC turned on as a convergence of evidence would be inaccurate. Describing it as having been a (not unanimous) consensus gets much closer to the truth.
Sometimes, scientists do have evidence, but as a journalist, you can’t evaluate its strength. You note some controversy, you can follow some of the arguments, but ultimately you have to be honest about how you got the information. And sometimes, that will be because it’s what most of the responsible scientists you talked to agreed on: scientific consensus.
As science communicators, we care about telling the truth (as much as we ever can, at any rate). As a result, we cannot adopt blanket rules of thumb. We cannot say, “we as a community are using this term now”. The only responsible thing we can do is to think about each individual word. We need to decide what we actually mean, to read widely and learn from experience, to find which words express our case in a way that is both convincing and accurate. There’s no shortcut to that, no formula where you just “use the right words” and everything turns out fine. You have to do the work, and hope it’s enough.
4 gravitons 
There’s a related problem where beyond-Standard-Model theories are described in the press as “beautiful”. Of course, when a physicist says that, they actually mean, e.g. that it is particularly simple, sharply predictive, or able to resolve open problems. But this doesn’t seem to be conveyed to the public, which leads to responses like: “those physicists are being dumb because they’re obsessed with beauty, when they really should care about simplicity, predictivity, and solving problems!”
LikeLiked by 1 person
Yeah, though I think this also owes to a central problem journalists face: readers like emotions! If you want to keep a reader engaged, it helps a lot to have quotes from scientists having pithy emotional reactions to things, so if you can catch them saying something is beautiful that usually goes in the piece. But in turn that simplifies what is usually a deeper judgement that doesn’t express as nicely as a short quote.
LikeLike
There is literally a whole book that has been written on the problems with using beauty as a criterion in physics (and about the wrong paths that this and the concept of “naturalness” have led the discipline of physics down). The English title is “Lost in Math”, but the German title, which translated literally means “The Ugly Universe” is better. https://backreaction.blogspot.com/2020/06/physicists-still-lost-in-math.html
LikeLike
That’s literally the exact problem I was pointing out. You _think_ that physicists use beauty as a criterion because that book told you so. But in reality, they never did. They always cared about simplicity and predictivity (which are what “naturalness” actually means) and popsci journalists simplified that into “beauty” because that was easier to understand.
LikeLike
I mostly agree with Anonymous Chicken here, with the caveat that the core of Hossenfelder’s argument is that simplicity, in physics, is an aesthetic criterion. It’s not like we can use Kolmogorov complexity when we don’t know the universe’s programming language after all.
(And ultimately the physicists who really know what they’re talking about here know that the criterion isn’t actually simplicity, but a more complicated set of philosophical and methodological assumptions.)
In any case, I agree with the “social” point: if “beauty” hadn’t already been an established pop physics trope, Hossenfelder would have had to phrase her argument in a much less catchy way. I can fantasize that would have led to a more reasonable discussion, though I don’t actually think that’s realistic given the personalities involved.
LikeLike
“They always cared about simplicity and predictivity (which are what “naturalness” actually means)”
Naturalness in this sense has a much more technical sense that isn’t about simplicity and an ability to make predictions, and instead essentially places Bayesian priors on the idea that certain fundamental dimensionless constants should be expected to be close to the the order of a 1, rather than being very small or very large.
Per CERN: “Colloquially, a theory is natural if its underlying parameters are all of the same size in appropriate units. A more precise definition involves the notion of an effective field theory – the idea that a given quantum field theory might only describe nature at energies below a certain scale, or cutoff. The Standard Model (SM) is an effective field theory because it cannot be valid up to arbitrarily high energies even in the absence of gravity. An effective field theory is natural if all of its parameters are of order unity in units of the cutoff. Without fine-tuning, a parameter can only be much smaller than this if setting it to zero increases the symmetry of the theory. All couplings and scales in a quantum theory are connected by quantum effects unless symmetries distinguish them, making it generic for them to coincide.” https://cerncourier.com/a/understanding-naturalness/
Of course, none of our fundamental theories are a great fit to naturalness, and indeed, if anything, Nature seems to prefer parameters that are as extreme as possible without going over the line to impossible. See, e.g., the metastability of the universe due to the running of the Higgs boson related parameters of the Standard Model at high energies.
Hossenfelder’s argument is more that any methodology that prejudges what the laws of physics or its parameters should look like, without experimental evidence to suggest them, is a very poor way to generate hypotheses with a poor track record of success. Concepts like Naturalness presuppose what Nature should look like, without any basis for doing so.
LikeLike
A lot of accounts of naturalness start the story “in the middle”, positing an EFT with some parameters and saying “wouldn’t it be weird if those parameters differed by many orders of magnitude?” I think a lot of BSM theorists even generally understand it in that way.
But there is a deeper way to understand it, which I talk about in the post I link upthread. There, the problem is not “numbers different from one” itself. Rather, it’s whether or not all these EFT parameters can in principle be derived from a theory with fewer parameters…i.e., a simpler and more predictive theory. Experience shows that, when you have such a theory, the numbers you generate tend to be “close to one”: you don’t get differences of many orders of magnitude.
That’s still very much not a knock-down argument, but it means one has to bite some awkward bullets. For Hossenfelder, as I think we discuss in the comments on that post, she thinks we shouldn’t assume the existence of a simpler and more predictive theory at all. From her perspective, there might simply be no better theory (in that sense) available. When I wrote that post, I was expecting her to choose to bite a different bullet, and say that the argument from experience above is wrong because said experience is misleading due to the lamppost effect.
Overall, I’m not particularly confident in naturalness at this point, but I do think that its failure probably means there’s something interesting going on with how the world is set up at a higher energy scale.
And I’d still argue that the trope where physicists and mathematicians wax lyrical about “beauty” made the conversation much muddier than it needs to be. Hossenfelder is arguing that physicists are applying aesthetic criteria, and sure, there’s an argument there. But transhumanists argue that bioethicists are applying aesthetic criteria, and there aren’t any books with the subtitle “how beauty leads bioethics astray”. It’s just a very different conversation.
LikeLike
The rise in the use of “consensus” did not come from “social scientists [being] skeptical about science’s objectivity and [wanting] to explore the social reasons why scientists come to agreement” but from the fall of positivism — the belief that scientists could discover and had discovered universal truths. This epistemological doubt had its roots in the quantum world. Thomas Kuhn’s idea of “paradigm shift” became widely accepted — existing belief structures could be challenged by experimental anomalies that caused scientists to change their beliefs to a new, more explanatory paradigm. In the world of science, much of what we now see is couched in the language of models — models have consequences which can be measured by experiments in the real world to determine the extent to which the model can be an explanation of the real world. The model is not truth, as little as possible is postulated about the real world. In the popular mind, however, this humility has become the vulgar constructivism in which we now live, where whole sectors of the economy act on the belief that by changing what people think is true, they can change what is true.
LikeLike
That explanation doesn’t fit the Google Ngrams plots in the post I linked. The developments you’re talking about happened in the 60’s, but people only started heavily using “scientific consensus” in the 80’s and 90’s, so it’s much more likely the vocabulary was popularized by the STS folks mentioned in that post, and not by Kuhn and co earlier. I could see Kuhn contributing to the fall of “convergence of evidence”, though, that does start decreasing around the right time.
LikeLiked by 1 person
If scientists want to communicate more effectively their belief in the potential success of any particular theory, it would be helpful if they could elaborate a little more on their reasoning. For example, string theory. I think it was generally understood, but not always clearly stated, that it was not possible to experimentally verify string theory for the foreseeable future. Nevertheless, it certainly seems that the scientific consensus was, and probably still is, that string theory is the best bet for a deeper understanding of particle physics. So far, so good. But I was completely taken by surprise when Leonard Susskind stated that there was no current theory of string theory that works in de Sitter space, ie the real world. Because my view on the credibility of any theory strongly turns on how it works in the real world.
Anthropogenic climate change, on the other hand, is a certainty. It is based not only on scientific consensus but also empirical fact and solid theory.
LikeLike
So, the thing with string theory and de Sitter is controversial to a kind of silly degree, which explains why there isn’t consistent messaging.
Ask some string theorists, and they’ll say that we’ve known how to make de Sitter space work in string theory for over twenty years. If you’ve heard that “10^500 universes” number, that’s based on that work, it counts the different numbers of quanta a field that’s supposed to give rise to a cosmological constant can consistently have. They’ll admit that the construction isn’t rigorous, and thus it isn’t 100% clear that it works. But they’d argue it’s not that different from the level of rigor people use in other parts of physics, and that they expect that eventually the details will get filled in and it will all work fine.
Others string theorists argue that, not only does that method definitely not work, but it is impossible to have a de Sitter space in string theory. However, many of them go even further, and argue that it is impossible to have a de Sitter space in any theory of quantum gravity, due to a variety of suggestive paradoxes. They think that instead of a constant de Sitter, we live in a universe with a kind of “quintessence” field that mimics de Sitter but can vary. These people were excited to hear about recent evidence suggesting that dark energy might change over time, since this would better fit their expectations.
(There is also a much smaller group of people who argue that even the approximate de Sitter people have observed is a misinterpretation of the evidence.)
You mention it seeming like the consensus is that “string theory is the best bet for a deeper understanding of particle physics”, and I think that shows a (common) misunderstanding. The consensus is that string theory is the best bet for a deeper understanding of quantum gravity. That consensus exists, in part, because string theory has nice long-term side benefits for particle physics, because it shows a plausible way that the world could ultimately be described by a theory with no free parameters, so that every particle mass and charge is something physical that, with enough energy, could be changed. But string theory’s competitors aren’t other theories of particle physics, they’re other theories of quantum gravity.
(There was a brief period when a lot of people thought that string theory could genuinely be useful for particle physics in the near-term. That period of time was central in Brian Greene’s career, and because he popularized it so well it’s become central to the public picture of what string theory is. The metaphors he established, the way he talked about string theory, have become standard tropes, so that most people popularizing nowadays are on some level borrowing from him, despite his popularization work being influenced by a very specific moment in time.)
When you put string theory in its proper context, as a theory of quantum gravity, then you notice that every proposal out there has big foundational issues. It isn’t clear whether Loop Quantum Gravity can describe space like ours, that is approximately flat. It still isn’t clear that the core arguments behind Asymptotic Safety actually work. And in string theory, not only is it unclear whether it can reproduce a de Sitter universe, but there are even more fundamental issues. In the end, I think string theory is still the most promising option: its foundational issues seem milder, and more likely to be fixed. But that’s mostly because the other options are that much worse.
LikeLiked by 1 person
Thanks for the mention, and thanks even more for engaging with this interesting question in a very helpful and productive way.
LikeLiked by 2 people