Google Analytics and Mathjax

Friday, August 20, 2021

Trusting Scientists

 A tweet that made me think.

The point here is subtle. We don't get on a plane because we "trust scientists", rather we do so because of the strong safety record of commercial aviation. I knew some physicists who won't get on a commuter plane because they worry about the science. Never stopped me.

It is science that we trust to tell us why planes fly, or the water is our tap is (mostly) safe and healthy. I'm not a big fan of beans but not because of the science. Of course I trust science that created the vaccines.

It's not just science, but solid engineering and lots and lots of testing. 

Science isn't always right or consistent. When I was a kid not that long ago, we had nine planets in this solar system, dinosaurs were killed off by climate change and homosexuality was a mental illness. Science is fluid, updating as we learn with new data, models and experimentation. Science is at its best when it doesn't trust itself.

Sometimes people say trust in science to reinforce their beliefs. I've seen smart people say "Trust in the science" about whether vaccinated people should wear masks with completely different conclusions.

I'm a scientist, should you trust me? Let me quote another Paul G. 

“There’s a slightly humorous stereotype about computational complexity that says what we often end up doing is taking a problem that is solved a lot of the time in practice and proving that it’s actually very difficult,” said Goldberg.

The quote comes from a recent Quanta Magazine article about Paul's recent work with John Fearnley, Alexandros Hollender and Rahul Savani on the hardness of gradient descent. Even many NP-complete problems these days can often be solved in practice.

Let's end with the quote attributed to statistician George Box, "All models are wrong, but some are useful". Science gives us ways to understand the world and we need to both trust in the science but know the limitations of what it has to say.


  1. I think you have put your finger on a very important paradox when you point out that "science is at its best when it doesn't trust itself." The motto of the Royal Society is "Nullius in verba," meaning "on the word of no one," or more loosely, "don't believe something just because someone says so." It is a basic principle of science that you should be able to verify its claims yourself.

    On the other hand, in practice, taking the principle of "nullius in verba" to an extreme forbids us from standing on the shoulders of giants, thereby severely limiting how much scientific progress we can make. If you Google the word "zetetic" then you will see what can happen when you try to pursue "nullius in verba" to its logical conclusion. Marcello Truzzi founded the journal, "The Zetetic Scholar" and popularized the motto, "An extraordinary claim requires extraordinary proof." He was a founding member of CSICOP and insisted on strict standards for evaluating paranormal claims; most people who say, "trust the science" would nod in approval. On the other hand, Truzzi was also critical of what he called "pseudoskeptics" who, as Truzzi saw it, had already made up their minds that paranormal phenomena did not exist. If you are sufficiently skeptical of the skeptics then you might end up with the "Universal Zetetic Society" (not something that Truzzi was involved with), which was primarily known for its belief in a flat earth.

    Modern science is so complex that it is not possible for me to check all its claims myself, and so it seems that in practice, I am forced to violate its own motto and blindly trust its verdicts. I find this state of affairs unsatisfactory, but unfortunately I have no silver bullet to offer. I have often wondered if theoretical computer science has something to offer with its concept of an interactive proof. I am Arthur, "science" is Merlin, and maybe I can attain high confidence that Merlin isn't cheating if I cleverly ask the right questions. I hold out hope that there's something to this idea, but the trouble is that in practice, scientific knowledge isn't as rock-solid as (say) graph isomorphism. It's almost always possible to poke little holes here and there in an established scientific claim, so Arthur is hard-pressed to tell the difference between a lying Merlin and an honest but fallible Merlin.

  2. > We don't get on a plane because we "trust scientists", rather we do so because of the strong safety record of commercial aviation.

    Are you sure that you are not trusting the scientists/engineers here?

    The aircraft and technologies used in commercial aviation change all the time, and you can't just say that a modern aircraft is safe because older aircrafts have safely flown for hundreds of thousands of hours. You have to trust the scientists and engineers who built the modern aircraft to conclude that it is safe.

    > Science gives us ways to understand the world and we need to both trust in the science but know the limitations of what it has to say.

    Well this is true, but from a practical (risk minimization) point of view, a scientist is far more likely to be right as compared to me or any other random person. So my best choice is to follow their advice.