Roy Sorensen & Michael Weisberg

Roy Sorensen (left) and Michael Weisberg (right) on idealization and scientific realism.

Idealization is the intentional introduction of distortion into scientific theories. If science aims at the truth, as scientific realists believe, then why are scientific theories routinely idealized? To answer that question, Weisberg take a pluralistic approach. He distinguishes three kinds of idealization (Galilean, minimalist, and multiple-models), and recommends that realists pursue different accounts of each kind. In contrast, Sorensen proposes that realists can develop a unified account of idealization if they can show that idealized scientific theories are merely supposed rather than asserted.

Related works

by Sorensen:
Bald-faced Lies! Lying without the Intent to Deceive” (2007)
Logically equivalent—but closer to the truth” (2007)
Thought Experiments (1992)

by Weisberg:
Who Is a Modeler?” (2007)
Three Kinds of Idealization” (2007)


9 comments to Roy Sorensen & Michael Weisberg

  • “If pigs can fly, then phlogiston causes fire.” This statement is vacuously true. If Sorenson is correct, all claims/theories in science which depend on idealizations are similarly vacuously true since the antecendent is false. Now Sorenson says at 42-42 minute mark that there has to be some “premises” which are added to render certain claims/theories less accurate or correct. But this is not clear and it seems to me to be a little too imprecise to make much sense of his supposition theory.

    In any case, I suspect it will not appease realists and may be ammunition for the instrumentalist and other anti-realists that all theories, true or otherwise (under more intuitive notions of scientific truth which seems to be about the world and which are true or false simpliciter and not true under logical material implication with an added false supposition as antecendent), in science or otherwise are similarly vacuously true even if they happen to be less or more accurate, etc, under some premise. In this way, no theory is really true in a more robust sense required by the realist because they will always be vacuously true under some (false) idealization.

    What I am missing from Sorenson’s account here?

  • Sorry, “Sorensen” and “42-43 minute mark”.

  • Neither of Michael’s papers listed as “forthcoming” is properly forthcoming any longer. The citations are here:

    Three Kinds of Idealization, 2007, The Journal of Philosophy, 104 (12) 639-59.

    Who is a Modeler?, 2007, British Journal for Philosophy of Science, 58, 207–233.

  • david

    Thanks Daniel. The dates are fixed.

  • In logic class, the one-sided diet of simple examples gives the impression that conditionals are asserted just to set up deductively valid inferences such as modus ponens. A vacuously true conditional then appears incapable of further service as a premise. But embedded conditionals show that a vacuously true conditional can convey an interesting connection between the antecedent and the consequent:

    Humphrey Davy: If Michael Faraday is awake, then he is in the laboratory.
    Mrs. Davy: If that is true, Faraday is a hard worker.

    Mrs. Davy’s assertion is not undermined if Faraday is asleep.
    Employment opportunities expand when vacuous conditionals move beyond deduction. Take the appeal to verisimilitude:

    If P then Q.
    P is close to the truth.
    Therefore, Q is close to the truth.

    Illustration: If the earth is sphere, then its surface area is 4pi times the square of the earth’s radius. `The earth is a sphere’ is close to the truth. Therefore, the surface area of the earth is 4pi times the square of its radius.
    This type of reasoning is inductively strong because a little difference rarely makes a significant difference. With threshold phenomena, some little difference may make a big difference (such as an avalanche). So background knowledge is needed to assess the applicability of the vacuous conditional.
    Standards also matter. For some purposes, we will need more accuracy and so need to accommodate the fact that the earth’s spin makes it bulge in the middle. But we can deal with this with another round of idealization; suppose that the earth is an oblate sphere . . . .
    Even if we view a theory as a conjunction of assertions, the scientific realist is entitled to use suppositional reasoning in applying these assertions. For the realist is entitled to classical logic and that legitimates both reductio ad absurdum and conditional proof.

  • Thanks for the response, Prof. Sorensen. That explanation seems far more plausible.

  • stuart

    And the mp3 version has vanished 🙁

    • david

      Stuart–We’ll have an mp3 for this episode up later today (and we’ll also post an mp3 for Keller-Stroud). Sorry for the delay.

  • LittieCheese

    Around minutes 38 and 42, it is said (paraphrasing) that premises are not asserted but supposed, and that the contact of suppositions with the world requires additional premises concerning application conditions (such as closeness to frictionless planes). I’m sure that details are lacking due to time constraints, but as it stands, I’m unsure of how the suppositional approach doesn’t collapse into scientific empiricism as opposed to realism.

    Take for example models of particle collisions, the supposition of their structure will only make contact with predictions through the addition of additional application assumptions. But what is the nature of these assumptions? For example, one models particle collisions by taking atoms to be rigid spheres (among other things too). This is really not that true, but it is fairly accurate in its predictions. So what sort of application assumptions would be supplied to allow this supposition to make contact with the world?

    Is it that atoms really are close to rigid spheres (as in they are very much like BBs or whiffle balls)? Or is it that they behave as though they are rigid spheres? The first assumption does not fit nicely with our modern atomic models, and the second approaches instrumentalism by treating behavior as the only relevant characteristic for models.

    Additionally, it seems as though the suppositional approach allows one to maintain conditionals and observations, but allows one to jettison assumptions about supposedly unobservable objects. This does not sit well with my realist inclinations. However, I may be misunderstanding the nature of suppositions or missed a key detail of applicability assumptions.