Clark, list,
I think that abduction as inferring back to the object pertains to
plausibility, that is, to abduction as forming plausible explanations.
Also, not all inductions are in familiar (statistical) forms, and so we
won't always recognize them as attempted inductions by noting their
statistical style.
The characterizing of both abduction and induction as ampliative is
harder-core and lets us separate the question of what is an induction or
abduction from the question of what is a good or bad induction or
abduction, though it doesn't go far enough obviously, since it doesn't
furthermore distinguish induction _/from/_ abduction - but a distinction
as 'hard-core' between them as the distinction between deductive and
ampliative would be desirable (but my candidate for that has never made
a splash).
I agree that it is not to be expected that any of the promise aspects of
inference are usefully computable. My impression is that Danko wants to
overcome such problems with his 'three-transverse' system and the
involvement of continual and evolving interaction with the environment
(though how that would work is over my head). I wanted to point out to
him that plausibility (natural simplicity) is not the only challenging
inferential aspect that he might find worth dealing with, and that the
aspects seem to belong to a family.
Best, Ben
On 4/3/2015 3:06 PM, Clark Goble wrote:
On Apr 3, 2015, at 12:04 PM, Benjamin Udell wrote:
You wrote,
While not everyone would agree, I think Peirce is somewhat
influenced by virtue ethics with regards to abduction. Thus
abduction arises out of an optimistic development of an aesthetic
and ethical life. Abduction isn’t defined except in terms of the
practices of people who conduct good a[b]ductive reasoning. This
reasoning can’t be formalized however. It arises out of consciously
engaging with the living of each of our own lives and adjusting
ourselves as we live.
Peirce doesn't quite _/define/_ abductive inference in terms of the
practices of people who conduct good abductive reasoning. Most
generally he defines it as a conjecture that arises in the mind in
order to explain (successfully or not) a surprising phenomenon, and
ultimately concludes that abducing has very little to do with rules,
and that the pragmatic maxim is necessary and sufficient as a rule
for it, to the extent that it follows rules at all. He does offer
particular forms of abductive inference - in earlier years, a
reordering of the Barbara syllogism, and in 1903 another form. I
agree that we should not regard those forms as _/definitions/_ of
abductive inference, but likewise I think that we should not regard
the Barbara syllogistic form as the _/definition/_ of deduction.
Well to talk about good or bad abduction seems independent from saying
what abduction is. When he talks about abduction most clearly I think
he opposes it to induction and deduction when he makes what abduction
is most clearly. The problem is that once you remove rules you then
have practices that are successful or not without saying what makes
them successful beyond whether they achieve those ends. Contrast this
with induction or deduction where we can often tell reasoning was done
poorly without knowing if the answer is correct.
I’ve often thought a good way to think about abduction is as the
reversal of the semiotic process. Especially in his late thought on
signs, such as in the letters to Lady Welby, he talks of the gap
between the object to the interpretant in a sign. This gap must be
bridged by a guess when one moves backwards towards ones object. The
clearing of this gap is fundamentally what abduction is about.
Really, none of these fruitful/promising aspects (nontriviality,
novelty, verisimilitude, plausibility / natural simplicity) lend
themselves to useful quantification, yet A.I. research could take
some sort of interest in all of them, since, if they were to be
generally absent, no mind would bother to infer. In the case of
deduction, special forms or schemata (e.g., the categorical
syllogisms) are developed in order to assure some modicum of novelty
or nontriviality in deduction, but how would one teach an A.I. what
such novelty or nontriviality are? How would it catch on to the idea
in general? If an instinct is not involved, then what is involved?
I’m a bit biased here, but I think that most of these aspects of
reasoning are tied to embodied practices such that they are inherently
non-computable. (Say here the Heideggarian line of reasoning that
critiques of AI such as that by Dreyfus take) There are ways to
simulate simplicity, such as variants of Ockham by minimizing
categories. But that’s really not simplicity of the sort I think we
use in our day to day experience. More an inference from simplicity.
The common error in AI, in my opinion, is in assuming some simplified
representation of simplicity, plausibility or so forth is the same
thing as our experience of such in daily lives. Again, I’m not saying
these aren’t extremely useful. I think the computational methods we’ve
discovered the past decades have been extremely beneficial. I just
don’t think they are the same was what humans do.
-----------------------------
PEIRCE-L subscribers: Click on "Reply List" or "Reply All" to REPLY ON PEIRCE-L
to this message. PEIRCE-L posts should go to [email protected] . To
UNSUBSCRIBE, send a message not to PEIRCE-L but to [email protected] with the
line "UNSubscribe PEIRCE-L" in the BODY of the message. More at
http://www.cspeirce.com/peirce-l/peirce-l.htm .
