How to program emotions
Crazy
I need to design a library to provide functions to program emotions.
Please suggest what you think will be outputs from a system which
simulate emotions. The inputs will be present stimuli and past
experiences. Also, I would be glad if you can provide me with what is
expected from a program which promises to simulate emotions. I will be
dealing with rudimentary emotions like hatred, lust, obsession and
happiness. Love is very complex for me but I humbly accept any comments
on love or any others.
Falazar
acieve it?
The problem I find with your statement, is that this is a very hard
problem without building a total general AI. Its very difficult to
realistically extract this single issue.
But on that note, I have thought a little about it, and most of it
boils down to only two emotions, or two sides of an emotion:
like/dislike.
Love, hate, obsession, all of these are really just offshoots and
degradations of like/dislike.
So at the very lowest level of randomness, we can just assign a value,
0-100 for everything in the universe.
I am depressed, Like:myself 20
I love Her, Like: her 95
Then you really need to get back to studying the actual human emotions,
and the easiest way to do that is to look to the writing. There are
things peopel have strong feelings about, and then others they dont.
Searching through large amounts of text can give you these
relationships often, even something as simple as "I like *" or "I hate
*" gives a large amount of information.
You could build up a good database of things "generally liked or
disliked" by people at large.
With that you could randomly choose a few of these, and assign them to
your personality, and build up a set of distinct individual
personalities.
A secondary set of issues here, is Why. That is the big question here,
I like Dr Pepper. Ok, why? Well........ Some are not as easy to
define. I like DP because it is sweet and tastes good.
I like her because she has long red hair and is nice.
So then in theory you can search out some other individuals that have
long red hair, and you have a higher chance of liking them.
Now you still have some arbitrary terms in here, that are not able to
be defined in the AI easily, like, "it tastes good" and "is nice"
These you have to handle carefully because they are a human measure.
They could be simulated fairly well using the above DB, and meta
information in the system.
A second part of the like/dislike scenario, is goal based programming.
This is where everything the AI does is based upon min/max goals.
Maximize my happiness, and minimize my unhappiness. So therefore your
AI will always try to make itself happy. This is reasonable, and will
still lead to the AI potentially taking a job it is not happy with, so
it can pay for an apartment that he is happy with etc. just depending
on which happiness vector is more important at the time.
One thing here I would like to see implemented is the law of
diminishing returns.
I like Dr Pepper. So I go buy Dr Pepper. I like Dr Pepper ALOT, so I
go buy 16 cases of Dr Pepper.. I start drinking it and tis wonderful,
but by the 5th one, it starts become not so wonderful. And so on, you
know how it goes.
And so this would naturally limit the AI's actions towards these goals,
so it would no be Stuck, continously doing the One thing it liked the
most, always, and such.
James Ratcliff
KaisaHann...@webinfo.fi
by the natural evolution to be a force which guides us toward best
survival in life. So feelings tie us to things which help us in
survival, like natural healthy life does - so we get tied to things
which we recognize to be its components. That is the task of positive
feelings.
And feelings repell us away from unhealthy directions. That is the task
of negative feelings.