Re: Michalski's views on abduction/induction

Ryszard S. Michalski (michalsk@aic.gmu.edu)
Sat, 19 Oct 96 09:56:04 EDT

>>>>Ryszard Michalski's inferential theory, which could unfortunately not be
presented at the ECAI workshop, is certainly interesting and
thought-provoking -- Richard, could you provide a pointer to one of your
papers?
---
The best sourse on the inferential theory is probably my chapter
"Inferential Theory of Learning: Developing Foundations for Multistrategy
Learning" in Machine Learning: A Multistrategy Approach,  Morgan Kaufman
Publishers, 1994.

>>I would like to respond to one of the above points, namely that induction is falsity-preserving. This has always struck me as something odd: why is falsity worth preserving? In my PhD thesis I have proposed an alternative view: explanatory induction is a form of induction that preserves (amplifies) explanatory power from premises to inductive hypothesis. Of course, if we identify explanatory power with deductive closure, then the two views coincide. However, the explanatory view has the merit that it is expressed in terms of something worth preserving.

---
The importance of "falsity preserving" of induction is that its  results
entail (jointly with BK) the observations (inputs to induction);
thus reasoning from the hypotheses to the facts ("consequents") is
truth-preserving. This property characterizes all forms of
induction (such as inductive generalization, inductive specialization, 
abduction, inductive concretion, etc.). Hope this helps a little.
Details (and a more precise elaboration of the above) are in the paper 
listed above.

Ryszard