This came up in a discussion thread a few years ago. In response to some thoughts from Danielle Navarro about the importance of model checking, I wrote:
This makes me think of an analogy between the following two things:
– Model checking in Bayesian statistics, and
– The self-correcting nature of science.
The story of model checking in Bayesian statistics is that the fact that Bayesian inference can give ridiculous answers is a good thing, in that, when we see the ridiculous answer, this signals to us that there’s a problem with the model, and we can go fix it. This is the idea that we would rather have our methods fail loudly than fail quietly. But this all only works if, when we see a ridiculous result, we confront the anomaly. It doesn’t work if we just accept the ridiculous conclusion without questioning it, and it doesn’t work if we shunt the ridiculous conclusion aside and refuse to consider its implications.
Similarly with the self-correcting nature of science. Science makes predictions which can be falsified. Scientists make public statements, many (most?) of which will eventually be proved wrong. These failures motivate re-examination of assumptions. That’s the self-correcting nature of science. But it only works if individual scientists do this (notice anomalies and explore them) and it only works if the social structure of science allows it. Science doesn’t self-correct if scientists continue to stand by refuted claims, and it doesn’t work if they attack or ignore criticism.
In short, science is self-correcting, but only if “science”—that is, the people and the institutions of science—do that correction.
Similarly, statistical methods are checkable, but only if the users of these methods actually check them, and only if the developers of these methods develop methods for users to perform these checks. Which is where I come in, as a methodologist.
As Thomas Bayes famously said, with great power comes great responsibility.
This particular blog entry ends with, “As Thomas Bayes famously said, with great power comes great responsibility.” Further, it directs us to a 2004 entry which repeats the phrase. From Wikipedia
https://en.wikipedia.org/wiki/With_great_power_comes_great_responsibility
the phrase takes center stage and somehow Spider-Man is invoked as well as “the release of Donald Trump’s tax returns.” Presidents Grant, McKinley,Teddy and Franklyn Roosevelt make cameo appearances. Likewise, the French Revolution’s infamous, Committee of Public Safety.
As far as I can tell, however, there is no website which states, “with no power, comes no responsibility at all.”
I strongly disagree!! Criticism is only appropirate in limited circumstances!! Only certified and authorized expert individuals should be allowed to criticize. Other than that, people should have to check with the relevant authorities before criticizing to see if their criticisms are valid!! In general, the mass of the people need to trust the science, women, “Palestinian Officials” in Gaza, all minorities, legislators who come from, and advocate for the practices used in, countries that are in total chaos and – by no means least – anyone who has married into a family established in the candy business. We know these are the Certified Sources of Truth and trusting them will lead to a Great Society!
Stop criticism now!! Let’s have a great society!
The methodology has to have antenna for picking up on a problem, and a rationale for deeming it a problem. Else, what are the grounds to either correct or take account of it? So, for example, if Bayesians claim their assessment of evidence at hand is insensitive to data-dredging, optional stopping, multiple testing and the like, what is the rationale for criticizing resulting inferences as unwarranted?
Then there’s the issue you often raise: the resistance of some Bayesians to model checking. To many of them, you observe, “any Bayesian model necessarily represented a subjective prior distribution and as such could never be tested” (Gelman 2011: “Induction and Deduction in Bayesian Data Analysis”
https://errorstatistics.com/wp-content/uploads/2019/12/gelman-rmm.pdf)
The description in the top article is rather different to the “self-correcting” nature of science as commonly understood at least in the physical/bio sciences which is something more akin to “the truth will out”. It doesn’t matter if scientists continue to stand by refuted claims or ignore criticism since the body of evidence in progressing fields leaves incorrect ideas marginalised (they are found to be uninteresting or not useful) whether or not scientists stand by incorrect claims. The scientific literature accumulates loads of incorrect stuff which isn’t formally “corrected” but that doesn’t matter since self-correcting involves the movement of scientific fields propelled by preponderance of evidence in productive directions. The incorrect stuff doesn’t have to be “corrected” – that’s really not what “the self-correcting nature of science” means. So Linus Pauling’s incorrect model for the structure of DNA wasn’t ever “corrected” – it just wasn’t a useful model (it had some clear flaws) and the Watson/Crick model was not only obviously correct but it had massive powers of explanation and hypothesis generation. Likewise the bunch of scientists who resisted the chemiosmotic theory of metabolic energy generation or continental drift or Felisa Wolfe-Simon’s resistance to accepting that her interpretation of bacterial incorporation of arsenic in place of phosphorus in DNA was flawed, and myriad other examples.
Perhaps things are different in the Social Sciences where there isn’t a broad movement of research fields with an imperative to get to the bottom of an invariant “natural reality” (and, a la Feyerabend, it doesn’t matter too much how one gets there!). Or maybe things are just different nowadays since the Internet allows detailed real-time scrutiny of published research by everyone in a way that wasn’t possible 20 years ago and so “hunting” flaws in other’s work has become a scientific sport. This is also fine, especially in a Fayerabendian sense, although it can lead to a cheapening of discourse, but maybe that doesn’t matter either.
It does matter to people who want, eg, a cure for cancer in their lifetime. Or who have priorities other than academic research they also want to fund. Ie, in a world of finite time and resources.
Like I could write some really slow code and rent out supercomputer time for years waiting for the result. Or write efficient code that will get the same answer in under a second running on my phone.