Re: [ba-unrev-talk] Connecting the Dots...
John Sechrest wrote:
>
> Perhaps this is the point where it would help
> to have a philosophical background, so that we know what the
> classic "Previously proved problems" are.
>
> So that the whole argument did not have to be drawn out to each
> node before we gain understanding.
>
It's possible, but the education in philosophy I received in
college (back in the middle of the last century) doesn't
suggest much that would of use here, that I recall. (01)
Instead, I'm reminded more of a game theory situation, where
we have a variety of past situations that have led to
strategy-formulation. Like chess players, the circumstances
on the board suggest potential strategies, but as a result
of an association process that links circumstances to strategies. (02)
The question, "what is the right thing to do?" is therefore
tantamount to asking a whole series of questions:
* What is the current situation?
* What aspects of the current situation are relevant?
* Which strategies have been successful or not successful
in dealing with situations of this kind?
* Given a strategy is execution possible in this case?
(Example: The king is behind a wall of pawns. That
immediately suggests a back-rank mate as a possible
strategy. But can it be employed? If all files are
blocked, or there are no rooks or queens on the board,
then the answer would be no.) (03)
A "resolution", then, would require examining strategies,
coming to an agreement as to whether they are appropriate,
and then coming to an agreement as whether they are executable. (04)
Interestingly, there is a model of chess analysis, where experts
evaluate a position and decide who has the upper hand. But there
are frequently disagreements, the analysis of even the best
experts is sometimes wrong, and even the apparently "best"
strategy could suffer from a failure of execution. (So a player
may choose an inferior strategy, and succeed with it, because
it suits their strengths, or exploits the oppositions weaknesses.) (05)
"Drilling down" in a such a scenario would require lots of
additional analysis, both to provide evidence for one's own
claims, and to examine the claims of another. But the time
spent doing that analysis would defeat the goal that heuristics
achieve in the first place -- allowing a decision to be reached
in time for it to be of any use. (06)
Again, I welcome contributions from anyone who has ever facilitated
such a discussion. Because at the moment, I don't see any basis for
solving the problem, with or without a tool. (Which again leads to
the notion of delineating cases, so we can separate "computationally
unsolvable" problems from those that we can profitably approach. (07)