From: Michael Roy Ames (firstname.lastname@example.org)
Date: Mon Nov 25 2002 - 18:53:25 MST
> > Your loose description is a good start... now we just have
> > to fill in all those pesky definitions for words like
> > 'significant', 'positive/negative' and 'outcomes' ;)
> Heavens, no. If you want to approach ethics as an objective,
> scientific problem the last thing you want to do is make
> unsupported assumptions about anything. [snip]
Thus, the wink ;)
> You certainly can't justify assuming that your own values are
> more correct than anyone else's, or that you have a better
> grasp of who matters than some other group does - the very
> fact that there is disagreement on these points is a sure
> sighn that you need hard data before you can come to
> any conclusions.
Yes, agreement. That is why I am currently uncomfortable with the
idea of encoding an ethical/moral system into software, even if - or
especially if - it is a sentient piece of software. When I am
uncomfortable with something, particularly something as important as
this, I try to do something about it. We need 'hard data' on morals.
FAI currently plans to get it from humans. This is the only place we
no of... but that doesn't mean current human morality is the best
morality. Could we do better? Gee, I hope so.
> But the data has to come first, and it has to be interpreted
> independantly of any one person's preconceived notions of
> what the final result 'ought' to be. Otherwise you're just
> manufacturing excuses for whatever ethical system you
> started out with.
The data is definitely very, very important. Also important is how
that data is compiled, and used. It would seem an utter waste if the
data were great, but FAI got stuck with the 'lowest common
denominator' for everything. Do we want a FAI to have 'average'
morals, or the best available? As soon as you start asking questions
like this, and really try to figure it out how it would work...
morality starts to look awfully arbitrary. And so: here I am, asking
the question: is there a way to make them less arbitrary?
Michael Roy Ames
This archive was generated by hypermail 2.1.5 : Sat May 25 2013 - 04:00:35 MDT