[FRIAM] AI and argument

[┣glen┫]

On 10/03/2017 07:51 PM, Nick Thompson wrote:
> Well, as a Peircean, I am certainly NOT allowed to believe that all valid logic is deductive, so Got Me There!

Heh, I'm not playing "gotcha".  What's important to me about my question is whether you think abduction can be formalized.

> But to the extent that we were talking about logic, is not logic the formalization of good thought?  So, then, it behooves one who would claim that an argument is logic to formalize it. So, in which logical world (if not deductive logic) does the statement that Einstein is usually right lead directly, without an intervening premise, to the conclusion that I should provisionally believe him.  I think the argument IS deductive (in this case) and that the suppressed premise is that I should treat all people who are usually right provisionally as authorities.  (i.e., as people to be believed until contrary evidence teaches us otherwise. )

1) My argument held Einstein up as an authority on relativity (or physics), *not* in any other sense.  For me, the following argument would be fallacious for appealing to an unqualified authority:  Einstein believes in God.  Therefore, we will find evidence God exists.  It's fallacious because Einstein is not an expert on God, as far as I'm concerned.  And even if he were, his reasoning (like all metaphysics) is at least somewhat opaque.

2) We weren't really talking about logic.  We were talking about reasoning and argument.  I find it interesting that you (and lots of others) conflate the two.  As we've discussed recently, there are many types of logic, including those that reject the rule Frank raised (that a false premise implies every/anything).  The attempt of these different logics (PLURAL, damnit) is to find one or more that *better* formalizes the reasoning/argument we know works in different circumstances.

So, while you'd be right to say we were talking about logics (PLURAL), we were definitely NOT talking about your own particular pet logic.  We were talking about the entire breadth of argument/reasoning used by both humans and computers (and dogs and honey bees ...).

FWIW, I would answer "yes" to my own question: I do believe (without much evidence) that all reasoning can be accurately formalized (sound or not).  But I don't see evidence of formalisms that capture huge swaths of (human) reasoning, including abduction.  (John Woods comes close with abduction, though, I think.)  That should make it obvious that I wasn't playing "gotcha" ... methinks the lady doth protest too much. 8^)

-- 
␦glen?