[MUD-Dev] Evolutionary Design

[Sasha Hart]

[Phinehas]
> [Sasha]

> I only mean to say that, while it is quite easy for me to
> understand that the seed is a "carrot" it does not seem nearly as
> natural to think of the key press as the "carrot." The key press
> seems more like a causal relationship set up to elicit specific
> behavior based on our knowledge of what functions like a "carrot"
> in motivating a pigeon.

I agree that it doesn't seem nearly as natural.

Watching an adult pigeon, it is clear that the pigeon doesn't have
to mess around to get the seed into its mouth. As you'd probably
expect, pecking is relatively stereotyped, very fast, and very
efficient.  I used to suppose, vaguely, that pigeons are just made
to peck at grain; that they come with the functionality. In any
case, there seems to be this obvious disparity between how hard it
is for a pigeon to learn key-pecking, and how obviously easy it is
for a pigeon to peck for grain: it just knows how already. And this
disparity has a pretty intuitive relation to the contrivedness of
the situations - pecking for seed is usual, pecking a key for seed
somewhat unusual, an artifact of the lab.

Somewhat surprisingly, a pigeon's adult seed-eating is actually a
developed skill (check out
http://www.pigeon.psy.tufts.edu/avc/balsam/default.htm#development).
the gist is that a pigeon raised on a powder diet has trouble
grabbing seed; make of it what you will, the point is that certain
conditions need to obtain for the pecking response to get its form;
pecking at seed, like pecking a key, isn't a given, though the
process of learning either is probably not so complicated).

I'd argue from the above that the intuitive distinction you pointed
out, basically between the pigeon being coerced to peck the key and
the pigeon eating the seed, is not a matter of naturalness or
contrivedness.

In an adult pigeon, it is quite reasonable for us to take
seed-eating for granted and to suppose that it is inherently
motivating, because 1. almost all pigeons know how to do it very
very well and 2. it is easy to demonstrate that the pigeon will work
for the opportunity to pick up seed. The question of whether or not
picking up seed itself is a contrivance of the environment doesn't
matter, because it is in the nature of the adult pigeon to know how
to pick up seed and to be motivated by it.

>> In that example, *your* carrots and sticks

This is my mistake. I meant the designer's carrots and sticks, like
XP and dying. Erk.

> Perhaps a better distinction would be to decide what *are *
> motivations and what simply *leverage* those motivations.  So, is
> the pigeon motivated to peck the key?  Perhaps not.  Perhaps it is
> only motivated to eat seeds, but a causal relationship between key
> presses and seeds might very well leverage this motivation.

Yeah, I think so too. It seems important to me that anything I call
a 'motivation' be motivating and not, well, un-motivating or
ineffective. The main difference I see in what you say above from
the intrinsic/extrinsic idea is that you are not calling things
which don't motivate motivators, which gets around one of the main
objections I presented. Correct me if I'm wrong, but you're bringing
out a question like "what things REALLY ARE motivators, regardless
of anything else, any other experience?"

The definition of motivation I'm using is operationally simpler - if
it motivates, it's a motivator, even if it became a motivator as a
result of some past events rather than always being one. That
(trivially) ensures that anything I call a motivator really
motivates. The difference between this idea and the one I think you
just presented is principally that mine doesn't necessarily imply
persistence or theoretical content in itself. Seed might be
motivating one day and not the next, for all I know, but today it
was motivating (descriptive). The equivalent assertion in the terms
I think you just gave is that seed may be a motivator, or it may
only leverage some more basic motivation, we don't necessarily know
for sure (descriptive with an eye toward theoretical).

It's notable, and it's only ethical that I point this out, that this
is very similar to the idea of reinforcement (as flogged by
Skinner).

I think there are other useful distinctions that account for a lot
of the intuition running into intrinsic/extrinsic. For example,
there is doing something and saying "This is GREAT" versus doing
something and saying "I hate this!" There is doing something and
making attempts to do other things which are quelled by punishment,
and there is doing something and not trying to do anything else.

There is doing something and being so ready to give it up that one
hair will break the camel's back, and there is doing something that
is well enough motivated that you would really need a lot of
counter- incentive to stop. But none of these distinctions require
an external-internal, inherent/non-inherent distinction per se. And
I think such a distinction is confusing in that we don't really have
a good idea of how to distinguish intrinsic from extrinsic
empirically.  (Though I'm certain someone or other does, in which
case the whole argument moves to the arena of word choice - for
example, if it's defined in terms of the distinctions I give
above. But at least in the process it would become more clear what
is being talked about specifically, and how it relates to events in
the world).




_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev