home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!howland.reston.ans.net!paladin.american.edu!auvm!BEN.DCIEM.DND.CA!MMT
- Message-ID: <9212232302.AA07188@chroma.dciem.dnd.ca>
- Newsgroups: bit.listserv.csg-l
- Date: Wed, 23 Dec 1992 18:02:10 EST
- Sender: "Control Systems Group Network (CSGnet)" <CSG-L@UIUCVMD.BITNET>
- From: mmt@BEN.DCIEM.DND.CA
- Subject: Re: PCT Fanatics
- Lines: 79
-
- [Martin Taylor 921223 17:40]
- (Rick Marken 921223.1300)
-
- Nobody is in the office here either. Aren't we all a bit crazy?
-
- Rick, would you accept a small revision--
-
- > How could
- >geniuses like Freud (psychodynamic theory), G.A. Miller (information
- > theory),Skinner (reinforcement theory), Green and Swets (signal detection
- > theory), Estes (stimulus sampling theory), Rummelhart (parallel distributed
- > processing theory), Chomsky (transformational grammar theory), Guilford
- >(trait theory), Tolman (sign-stimulus theory) etc etc ALL BE WRONG?
- >...
- >
- >The answer is that ALL of these theories were based on a completely
- >incorrect view of behavior. They are ALL based on the idea that outputs
- >(neuroses, responses, operants, decisions, behavior, speech, intelligence,
- >movements) are caused by events in the environment or the brain.
-
- Not all of these examples are theories of behaviour, and so I might
- suggest adding to your first paragraph " ... ALL BE WRONG as applied
- to behaviour."
-
- I think specifically of information theory, signal detection theory,
- and PDP, all of which exist quite independently of whether there is
- a living organism in the neighbourhood. And the first two are mathematical
- statements which have universal applicability, even within PCT, if
- they are used appropriately. The problem with them is not that they
- are wrong in themselves, but that they are often wrongly applied and
- misunderstood.
-
- As for PDP, it is interesting that the simplest form of the connections
- within the perceptual side of a control hierarchy is exactly a
- multilayer perceptron, so that from PDP you can find an existence
- proof--anything a multilayer perceptron can discriminate is potentially
- controllable. And that means any configuration is controllable in a
- three-level hierarchy (maybe it is four, but I think three).
- So even if PDP is used poorly under normal circumstances (and I have
- long thought it is, before I heard of PCT), nevertheless you can point
- to it to show that simple structures simply connected CAN control
- perceptions of arbitrary complexity.
-
- The flip side of that is that if the PDP people ever, with any kind of
- node, demonstrate the possibility of making any kind of classification
- or recognition, you can immediately say that this is demonstrably a
- controllable kind of percept. So PDP provides an ever-rising lower
- bound on the known possibilities of PCT. PDP isn't a theory of
- behaviour, although people have used its possibilities that way (as
- you say, wrongly). But it does what it does, and you can use it without
- fear and trembling.
- >
- > I guess
- >Martin is preparing a thesis on the value of information theory
- >for understanding control . I?m waiting with great interest to see
- >what have missed by ignoring iformation theory. I have read
- >several rather unconvincing attempts to show that some versions
- >of behaviorism are equivalent to control models of behavior. If
- >anyone else out there has a non-PCT theory that they think provides
- >a real great explanation of some aspect of behavior then I?d sure
- >like to hear about it.
-
- I hope you aren't putting me in with people offering a non-PCT theory.
- I thought that what I had shown was thay information theory provided
- a demonstration of the necessity of PCT, but apparently I wasn't
- convincing. When I subsequently showed my posting around here, the
- reaction was that it made intuitive sense. So I have hopes that a
- more careful discussion will make intuitive sense to more people.
-
- People don't necessarily bring in ideas that have proved useful elsewhere
- just because they have a great fondness for the ideas and don't understand
- that PCT stands isolated. There IS the possibility that PCT hasn't
- reached the limits of its potential.
-
- Martin
- PS. You may be amused to know that I have observed in myself much the
- reaction you often express when reading or listening to psychologists
- talking in areas I used to think important. "What stupidities are they
- thinking of...Don't they know it's all control?"
-