Science and the quest for truth

20 Feb, 2018 at 19:01 | Posted in Economics | 4 Comments

28mptoothfairy_jpg_1771152eIn my view, scientific theories are not to be considered ‘true’ or ‘false.’ In constructing such a theory, we are not trying to get at the truth, or even to approximate to it: rather, we are trying to organize our thoughts and observations in a useful manner.

Robert Aumann


What a handy view of science.

How reassuring for all of you who have always thought that believing in the tooth fairy make you understand what happens to kids’ teeth. Now a ‘Nobel prize’ winning economist tells you that if there are such things as tooth fairies or not doesn’t really matter. Scientific theories are not about what is true or false, but whether ‘they enable us to organize and understand our observations’ …

Mirabile dictu!

What Aumann and other defenders of scientific storytelling ‘forgets’ is that potential explanatory power achieved in thought experimental models is not enough for attaining real explanations. Model explanations are at best conjectures, and whether they do or do not explain things in the real world is something we have to test. To just believe that you understand or explain things better with thought experiments is not enough. Without a warranted export certificate to the real world, model explanations are pretty worthless. Proving things in models is not enough. Truth is an important concept in real science.


  1. I have to chuckle a the “Science is based on consensus” comment. It leaves out the scientist, after all, for few scientists are passionately pursuing their ground breaking ideas with the end of goal of creating a “consensus” 😉 Of course, science as social institution evolves slowly and is retrained by consensus, but even at times when one view reigns supreme there are always minority views. Imagine Newton striving for consensus rather than truth (he was pretty explicit on the latter ;-), or Einstein pursuing consensus rather than truth; rather silly I think and very ahistorical and narrow. But then sounds just like homo economicus, doesn’t it?

    • Your chuckling is intended to reinforce the consensus that science is not based on consensus.

      Aristarchus of Samos proposed a heliocentric model of the solar system in the third century BC, but consensus on fundamental assumptions prevented astronomers from considering it.

      Newton proposed one hypothesis; a consensus accepted it. Remember Pirsig’s Zen and the Art of Motorcycle Maintenance: therevare infinite hypotheses for any phenomenon. Today Newton’s consensus has been replaced, by consenting to another hypothesis from another powerful personality, Einstein. But Einstein’s gravity equations don’t explain the observed velocity of stars rotating at the outer reaches of galaxies. By consensus, we accept dark matter theory; but many other hypotheses exist. 100 or a thousand years from now, will we have different gravity equations? What is preventing us from discovering them now? I say consensus.

      Individuals propose hypotheses, consensus selects among them, mostly based on characteristics of the individual. Some measurements are used, but that requires consensus for the accuracy of the sensors. Once again, the epicyclists disproved heliocentrism by measurements showing there was no observed parallax motion of the stars.

      If Newton was pursuing truth he failed because you can’t build GPS with his equations; if Einstein was pursuing truth he failed because he could not reconcile the fundamental inconsistency of quantum physics with relativity … But they each achieved a consensus which reigns for a time, running roughshod over minority views such as Aristarchus’s, until the next powerful personality gains a following and creates a new consensus.

      In AI for example there was a consensus that neural networks were unable to solve complex problems; Minsky and Papert proved it in 1969. Research was reduced by consensus (and lack of funding). Since then, popular figures such as Hinton have created a consensus around deep learning neural networks. I used to be mocked for talking about neural networks; now it seems every other post on social media is about deep learning. Consensus …

  2. “Truth is an important concept in real science.”
    Science is based on consensus. As the consensus changes (epicycles, ether, dark matter, dark energy) so does truth. Epicycles had an export warrant to the real world, and Aristarchus’s heliocentric theory was not supported by evidence since their instruments were too blunt to measure parallax motion of the stars … Physics made all sorts of export warrants before dark matter changed everything, and now they make different export warrants but somehow the previous ones were valid when they made them back then …
    “To just believe that you understand or explain things better with thought experiments is not enough.”
    I want to program the thought experiments. If perfect hedges work in practice, public banks can generate riskless revenue without needing taxes … Alternatively, if you want to create a virtual world where no arbitrage conditions are strictly enforced by the system, you could do that and choose to spend your time playing with money inside that virtual sandbox while others can choose their own constraints.

  3. Lars,
    You say: “Without a warranted export certificate to the real world, model explanations are pretty worthless. ”
    How would an “export certificate” look?
    In your view, what would be required to validate a model?

Sorry, the comment form is closed at this time.

Blog at
Entries and Comments feeds.