I browsed an essay of Norton's and pulled out a cool quote:
We can infer inductively from the evidence that some samples of the element Bismuth melt at 271 oC to the universal conclusion that all samples melt so. The warrant is a fact about chemical elements:
Generally , all samples of one element agree in such physical properties.
The “generally” accommodates the existence of allotropes of elements, which typically differ in their physical properties. Without this qualification, the inference would be deductive. The qualification gives the inference an inductive character. In accepting the conclusion, we take the inductive risk that this element has no allotropic forms that would have a different melting point. As result, the inference is ampliative, even with explicit adoption of the warranting fact as a premise. The material theory does NOT assert that all inductions are enthymemes, that is, deductive arguments with suppressed premises.
I'm having trouble getting on board with this, because I view successful induction as something which can be retroactively deduced from the facts of the matter. For Norton, it seems that you can't perform induction without risk of error.
Another way to say this is that induction doesn't add facts -- but only probabilities -- to our growing body of knowledge.
Ed
|