[Avodah] psak halacha

[Micha Berger]

On Fri, Jan 09, 2009 at 8:54am EST, Rich, R Joel wrote:
: [R' Eli Turkel:]
:> BTW we once had a discussion of a automatic psak computer. Such a
:> computer of course could not account for such extra-halachic issues

: I think I would disagree in theory in 99% of the cases- just need a bigger
: computer and better fuzzy logic/AI programmers and poskim who can detail
: how they reach a decision. Of course it's hard to detail how you would react
: to a black swan.

To explain the reference...

Inductive reasoning is imperfect. You see a swan, it's white. And another,
it's white. And another and another... At some point you conclude that
all swans are white. Now, what do you do when you encounter your first
black swan?

I'm not sure pesaq has much role for inductive reasoning. Perhaps qal
vachomer, to establish the chomer, we would say "if X is stricter in ways
A, B, C and D, it is obviously stricter in general". Induction over cases
is called chazaqah disvarah, and we have rules limiting its authority. In
short, I'm not sure "black swans" really poses a major issue.

On Mon, Jan 12, 2009 at 8:08am IST, R Eli Turkel replied:
: If the psak depends on the community then the computer would have to
: account for that.
: AFAIK fuzzy logic relies on probability and so one has a probabilistic psak.

Fuzzy Logic (FL) isn't statistic. I don't think the difference matters
to our discussion, but since it was raised...

FL isn't about the chance that someone is tall (for example). Rather, it's
about the intensity of membership to a set. Some people are definitely
tall. Some shades of color are definitely red. However, others are only
somewhat tall or red. FL is about logic involving claims about membership
in sets with fuzzy edges, gray area.

The odds that the next coin you flip lands heads is 1/2. The odds that
the next card you pick is hearts is 1/8. Since the two are uncorrelated
events, the probability that the next coin lands heads AND the next card
is a heart is 1/8. Statistical AND of two unrelated facts is computed
by multiplying.

In FL, though, AND is performed using a MIN. It's not about stats, so
correlated vs uncorrelated is a non-issue. A man who is .8 tall and .6
red-headed has a .6 membership in the class of tall red-heads.

...
: Obviously these are 2 poskim with different attitudes to including the
: general populace
: and not just rabbonim or even shomrei mitzvot in dancing with the torah.
: How could one put that into a computer program?

On Mon, Jan 12, 2009 at 5:34am EST, RJR answered:
: The same way a preference towards needs of the tzibbur vs. individual
: etc. - I assume this would be one of a number of variables that would
: need to be adjustable to mimic any particular posek.

Yes. In a fuzzy or statistical, or any other multivalent system.
(Multivalent: having multiple values; typically between "true" and
"false".) If the weighing of factors can be done algorithmically, then
all of these can "simply" be settings you put into the system.


However, I don't personally believe that "if". I believe there are
problems solvable by true intelligence that can't simply be solved by
getting "a bigger computer and better fuzzy logic/AI programmers".

This is a matter of major debate, is true AI possible? I personally do
not believe that it's possible on a computer. And therefore I have room
to believe in a process that a rav can do that AI can not.

(My own beliefs are based on The Chinese Room thought experiment of
John Searle: http://www.consciousentities.com/stories.htm#chineseroom
My experience having qualia
http://www.consciousentities.com/stories.htm#mary
and the notion of koach hadimyon
http://www.aishdas.org/asp/2006/12/ruach-memalela.shtml

(I believe that the first-hand experience of thought, that reflective
self-awareness, is what Unqelus means by "ru'ach memalela". It's not
that man can speak, it's that we think in speech, we can hear and see
our thoughts, think about thinking, and modify our thought in response.
That's the topic of that third link.)


Assuming a computer can't do AI...

How important something is could very well not be something reducible to
numbers and compared. It could be an issue of semantics, not symbols
and algebra.

This impossibility is what I meant when I called pesaq a creative
process. Not just that different settings would produce different
answers, but that it can't be reduced to any objective system, including
any algorithm.


On Sun, Jan 11, 2009 at 10:06pm EST, R Richard Wolpoe wrote:
: I wouldn't want a computer to spit out a psak.
: but I would like it to present the relvant factoids related to a p'sak

: For example the Abudarham's point re: 10th of Teves has been rejected by the
: BY.  That fact must be noted. It's not jsut that Rashi and Rambam ot
: Abuarhma, it's that the BY sets it aside himself!

: The various postings omitted this factoid, and are somehow rehabilitating
: the Abudarham as something more than what apperas to be a patently rejected
: da'as yachid.  A computer wouldn't do that.

OTOH, it might. It depends whether the common acceptance of the idea
offsets its being a daas yachid explicitly rejected by the BY. How the
dials are set on RJR's program.

Tir'u baTov!
-Micha

-- 
Micha Berger             I slept and dreamt that life was joy.
micha at aishdas.org        I awoke and found that life was duty.
http://www.aishdas.org   I worked and, behold -- duty is joy.
Fax: (270) 514-1507                        - Rabindranath Tagore