[ExI] experiment regarding ethical behaviors vs status

spike spike66 at att.net
Wed Mar 28 00:30:58 UTC 2012

>... On Behalf Of Anders Sandberg

On 27/03/2012 18:07, spike wrote:
>> ... It's just deep within the basic nature of the beast: sloths are slow,
you are nice.

>...But we can try to measure the change in niceness? There ought to be some
effect size?!  I *promise* not to donate the money to charity! :-)  --
Anders Sandberg,


If d(nice)/dt is negatively correlated with d$/dt, then...

What we need are metrics for nice.  This would be a really interesting
artificial intelligence exercise.  Anders, I have followed your posts on ExI
for over 15 yrs now, and I have never seen a comment or word from you that
was unkind, even when unkind words would be amply justified.  You and Johnny
Grigg are the champions of nice, and I do deeply admire that, sir.

Humans can easily estimate nice from reading archives, but can we write
software to measure nice?  We have a huge body of work to use as a test
subject, the ExI archives from about 2003.  We could get a group of
randomly-chosen readers to estimate a kindness metric between 1 and 10,
although I might have already accidentally wrecked it by identifying a
couple of high NQ scorers.  Would not it be cool to try to write a script
that can take those archives and match the general profile of the readers?

Any code hipsters have any ideas on how to do this?  If so it would be a
classic AI app: having an (apparently) unfeeling machine try to read human
language and measure a human emotion.


More information about the extropy-chat mailing list