Re: 50 years later, Marvin Minsky still doesn't get it

Glen M. Sizemore wrote:


"J.A. Legris" <jalegris@xxxxxxxxxxxx> wrote in message
news:1156434816.740376.106410@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

OK, I'm back. Thanks again to you and Olea - Catania's book is quite
good ( I have the 1979 edition). Chapter 4 (Consequences of Responding:
Reinforcement) is particularly interesting - he seems to anticipate the
"response-deprivation" principle you mentioned above.

There's a little footnote on page 78 that I nearly missed: "The
detailed operation of Premack's principle depends on how probabilities
are calculated. Choice among simultaneously available responses seems
more satisfactory than the proportion of time occupied by each
response..."

The degree of choice among a set of responses can give us the average
amount of information obtained in making those choices.

I'm not sure what is meant by this ("degree of choice"), but the way that
Olea talks about the function of responses in entropy reduction could
easily be applied, I would guess, to some choice procedures.

I jotted down some comments, but I'll have to wait a bit to elaborate -- I'm
going out of town for a few day, leaving in about 10 minutes. I'll be
attending a seminar on Alex, an African Gray parrot that has been the
subject of many behavioral studies.

Anyway, "degree of choice" needs to be made precise. Is this just the number
of freely available choices, or the entropy of the distribution over
choices, or something else?

The information gained by the message that a particular choice had been
made, would be information gained by an observer. This would be the
reduction in entropy from its value before the receipt of the message, just
the entropy of the distribution over choices, to the entropy after receipt
of the message, which would be zero. It is not clear to me how this could
be applied to ranking the relative effectiveness of reinforcers. One idea
is to do a baseline study, in which choices are all freely available,
record some ratios, and use these as predictors of behavior when available
choices are made contingent on each other (no running on your wheel till
you finish your food pellets, young rat). But I'm suffering from ratio
strain -- gotta break and run.

-- Michael

.

Relevant Pages

  • Re: 50 years later, Marvin Minsky still doesnt get it
    ... Reinforcement) is particularly interesting - he seems to anticipate the ... "response-deprivation" principle you mentioned above. ... Olea talks about the function of responses in entropy reduction could ...
    (comp.ai.philosophy)
  • Re: 50 years later, Marvin Minsky still doesnt get it
    ... Reinforcement) is particularly interesting - he seems to anticipate the ... "response-deprivation" principle you mentioned above. ... The degree of choice among a set of responses can give us the average ... amount of information obtained in making those choices. ...
    (comp.ai.philosophy)
  • Re: Maximum entropy principle
    ... [Min entropy production versus Max entropy production ] ... >> equilibrium with fixed boundary conditions*. ... >> The MaxEP principle is relevant for *processes that do not have fixed ...
    (sci.bio.evolution)
  • Re: An Inflationary Account - #6a - missing information - partial retraction
    ... Thermodynamic entropy should be thought of as 'missing information'. ... Shannon entropy is closely related to thermodynamic entropy. ... One example of this is provided by the 'Data Processing Inequality". ... so that there are only two possible responses to each question. ...
    (talk.origins)
  • Re: Entropy maximisation
    ... entropy production and self-organized criticality in non-equilibrium ... a little further what the "selection principle of maximum entropy ... around and use that same data to prove that entropy maximizer are good ... survivors. ...
    (comp.ai.philosophy)