Skip to content
Coordinates of ThoughtHome of Philosophy
Join
Debate

Thesis A model can work for decades without being true, and instrumentalism is the honest response

Abdulrahman Al-Qahtani Contributor
· 6 replies · 230 views
Scientific methodAnalyticModern EuropeanBas van FraassenHilary PutnamPierre Duhem#instrumentalism#models#prediction#scientific-realism
Follow

Where it stands

I build computer models of oil reservoirs. We tune them to match years of production data, a process called history matching, and then use them to forecast. Here's the uncomfortable part: several quite different models of the rock underground match the same history equally well. One has a sealing fault where another has a low-permeability layer. They forecast similarly for years, then diverge.

So when my model works, I don't believe the rock is as drawn. I believe the model is a good tool for this field, for now.

Astronomy had a two-thousand-year version of this. Ptolemy's epicycles predicted planetary positions well enough to be used for centuries; nobody now thinks planets ride on circles on circles. Duhem wrote a whole book on the ancient and medieval view that astronomy's job was to “save the phenomena”, to fit the observations, not to say what the heavens are really like.

My thesis: predictive success shows that a theory is useful, not that it is true. The honest attitude toward unobservable entities is instrumentalism. Treat theories as instruments, judge them by how well they work, and don't add a belief in hidden things that does no extra work.

I expect the physicists to object.

LikeReply
Follow6 replies

Members reply here. Reading is always free.

Sign in to reply

6 replies

For the thesis 2

  1. Wanjiru Kamau

    For Contributor

    Statisticians have an old saying, usually attributed to George Box: all models are wrong, but some are useful. I live by it. In my work, believing a model is “true” is exactly what gets people into trouble, because they stop checking where it fails.

  2. Tomasz Wójcik

    For Fellow

    A clarification that may help Abdulrahman's case. Van Fraassen's constructive empiricism is not instrumentalism. He thinks theories are literally true or false, including about electrons. He just says the aim of science is empirical adequacy, and that accepting a theory need only involve believing it gets the observable phenomena right. That's a more defensible position than “theories are mere tools”, and I think it's the one you actually hold about your reservoir.

    Helpful · 3

Against the thesis 3

  1. Emeka Nwosu

    Against Fellow

    Your reservoir models are tuned to data you already have. That's accommodation, and realists agree accommodation proves little. The no-miracles argument is about novel prediction: theories positing electrons predicted phenomena nobody had observed or was looking for. A tuned model can't do that. So your example shows something true about curve fitting, not about electrons.

    Helpful · 2
    1. Abdulrahman Al-Qahtani

      Contributor

      Emeka, the novel-prediction point is the strongest reply, but it doesn't fully save realism. Fresnel's wave theory, which posited the ether, predicted a bright spot at the centre of a circular shadow. Poisson derived it as an absurd consequence, and then it was observed. A genuine novel prediction, from a theory whose central entity doesn't exist. So even novel success doesn't guarantee the entities.

      Helpful · 1
  2. Lars Eriksen

    Against Fellow

    I'm a realist, but a cautious one. What survives from Ptolemy and from Fresnel is not their entities but their mathematical structure, the relations they got right. So I'd say: believe the structure, be humble about the stuff. That's not instrumentalism; it's a claim that our theories are true about something.

    Helpful · 3
  3. Arvind Poojary

    Against Fellow

    My worry is practical. In my rationalist work, instrumentalism is often used as cover: “science is just a tool, so my healer's tool is as good.” I know that's a confusion, because instruments can be compared by how well they work. But the realist language, that there really are bacteria, does important work for ordinary people. I'm against the thesis partly for that reason.

    Helpful · 2