[Martin Taylor 951219 10:30]
Bill Powers (951218.1715 MST)
Martin Taylor 951218 15:00 --
The cost function I envisage can be seen by analogy. To a person
with $2, the "cost" of a $1 item is greater than it is to a person
with $10.If you ever go into business with someone, Martin, do NOT volunteer to
keep the books.
To the shopkeeper, it doesn't matter whether the purchaser has $10 or $2.
It's the shopkeeper who has the books "in business". The guy with $2
has his own books, and those are not the same books as are used by the
guy with $10. Each of the three keeps a separate set of accounts.
The "cost" in this case is psychological. Don't you ever have times when
you think you haven't got enough money to afford something you want, whereas
at other times you could have done so? I certainly have. At some moments
a $10 item "isn't worth it" whereas at other times I would buy the same thing
at the same price without a second thought. What something "costs" depends
enormously on how much money I have.
OK, you've invented a nonlinear cost function. This requires the rat to
know how much of the resource is left, as well as how much it is
getting,
What does "how much it is getting" (of food pellets) have to do with
how much of the resource (energy, time, number of paws...whatever) it
has available? These things come together only in some sense of value.
Is it worth stopping playing the piano in order to get a drink? You can't
do both at once. To the rat, is it worth pressing the lever to exhaustion
in order to get one lousy food pellet?
but what the hell, if you want to prove a point you just assume
whatever works.
These assumptions may be over specific, but I don't think it is in the least
unnatural, or suggested by "reinforcement" that it costs you more to give
up something of which you have only a little than to give up the same
quantity of something of which you have a lot.
You still didn't deal with the claim (in effect) that loss of
reinforcement from a running wheel is equivalent to a loss of food
reinforcements.
Where is the reinforcement from a running wheel? Does the rat find it fun?
This argument lumps all reinforcers together, so that it
is total reinforcement from all sources that increases the probability
of bar-pressing, not just reinforcements generated by bar-pressing.
Huh? I simply don't understand this comment. It comes "right out of left
field". I lump all resources of a given kind together, as one must. If you
get tired doing one thing, you can't do another energetic thing as well. If
you are using all ten fingers to play a chord, you can't play an eleventh
note in a different octave. If the rat is running on a wheel, he can't
be pressing a lever.
How this relates to reinforcements, I'm afraid I cannot see, except that a
reinforcement has a value and a resource usage has a cost. The value of
a reinforcement presumably relates to how much of it you have compared to
how much you want, just as I think the cost of a resource relates to how
much you have to expend out of how much you have.
But what this has to do with lumping all reinforcements of DIFFERENT
behaviours together as reinforcements for ONE behaviour, I cannot see. If
there is anything in reinforcement theory at all, shouldn't reinforcements
for different behaviours should combine subtractively, rather than additively?
If you reinforce behaviour X at level A, then reinforcing behaviour Y at
level B should lead to less of X if you increase B, shouldn't it? Or is this
just common sense and not reinforcement theory?
Is
this another of those varieties of reinforcement theory that we keep
hearing about?
As you should be well aware, I don't know any varieties of reinforcement
theory except what I read here on CSG-L or what I vaguely remember from
a course almost 40 years ago. All I am doing now is saying how I
interpreted a paragraph from Samuel Saunders, on which you commented in a
way I thought missed his point. As you miss mine. I assume that since
he brought it up, it has some relevance to reinforcement.
No, I didn't catch any algebraic errors. It would be interesting,
however, to see your justification for the cost function you chose as a
starting point, and how it relates to the problem of modeling
reinforcement.
The justification for the cost function is, like many of your justifications
for particular perceptions, simple everyday experience. It costs a lot
to give up your last dime. How the cost function relates to the problem
of modelling reinforcement I would leave to people who might be interested
in that topic.
All I wanted to say was that it seemed to me that Saunders had presented
what you asked him to present--a common-sense non-linear cost function
that prevented the runaway you asserted to be implicit in his presentation.
Martin