Piet schreef op zondag 4 november 2007, 18:02:
[knip]
> Dit vond ik in mijn mailbox...
> Omdat ik hier pas heb beweerd dat we van de we niet
> alles van de vier vormen van mogelijke training
>
> Piet
> The opposite of positive reinforcement is no reinforcement,
> not correction
Primo: Skinner is niet gek maar dood.
Secundo: een straaltje noorderlicht.
Het lijkt me wel handig en eerlijk de complete bijdrage van die poster even te kopiëren, zodat de kreten in context staan.
CITAAT OPENEN:
" I'll weigh in on this, perhaps a bit controversially, on a couple of points.
First, I once again will denounce so-called "four quadrants" analysis. There are basically three ways to respond to a behavior: (1) reinforce it (reinforcement), (2) suppress it (punishment), and (3) ignore it and let it die a slow death (extinction). You can then divide reinforcement and punishment into types: positive and negative. So you generally have three ways to respond to behavior, and five if you do the subdividing. The only way you get to four is if you take the average of the two. Everyone who talks about "four quadrants" simply forgets about extinction.
It's most important to remember that if your response can be called "reinforcement" (whether positive or negative) then by definition it builds or strengthens behavior; conversely, if what you are doing serves to build or strengthen behavior then it is reinforcement (whether positive or negative). It is critical that one remove moral judgments from the terminology.
The other problem with "four quadrant" theory is that by definition it connotes equality among the parts. A quadrant is one-quarter of a whole. However, in our training we do not use the responses in any way close to equal measure. Those who defend "four quadrants" say "well, we really don't mean that", but if that is so then the term "four quadrants" is completely misplaced and inaccurate.
The fact is that 99.something% of what we do involves positive reinforcement; +R is in no way used in equal measure with any of the other responses available to the trainer. If you look at ABE's history, though they did use punishments in one or two situations, statistically it was a non-event given the vast numbers of animals they trained over their 45 year history. So in any given case it is not a matter of whether one "violates a taboo", but whether one can justify the use of a punishment in light of the stringent criteria for doing so.
I referred earlier to Annie (the filly we're boarding). She nips, and sometimes will turn suddenly and fire a kick at you. Now, in the immediate situation an aversive response may be called for. But the bigger issue is doing the ground work and manners training necessary to condition her away from these behaviors. So comparatively speaking, perhaps one punishment to keep her from kicking my teeth in
and 10,000 trials using positive reinforcement to teach her proper manners.
Second, we have to remember that not all aversives are punishments, and further that something initially aversive can become a stimulus or cue for a behavior and its initial aversive quality fades and sometimes disappears entirely. One example is using heel pressure on a horse's flank to make the horse move. The horse moves in response to the pressure and you remove the pressure (-R). Another example is the "2-reward system", which involves (1) removal of an aversive and (2) positive reinforcement; a common example of the use of this is in conditioning a dog to accept the approach of strangers. The stranger approaches to a point where the dog -- though not wanting the stranger present -- remains calm. The stranger then leaves (-R) and the dog gets a C/T (+R). Another recent example, i think on ClickRyder, had to do with teaching a horse to accept trailer loading. The trainer moved the horse in with +R, then allowed the horse to back out under control (-R) so the horse did not feel "trapped" in the trailer.
The trainer is not condemned to spend eternity in the lake of fire for using -R, nor does he achieve eternity in the company of the saints by using +R. It's a matter of using the tools appropriate to the situation, always emphasizing +R over the others for the fundamental reason that we are above all in the business of building not suppressing behavior. That it also improves our relationships with our animals at so many levels is also an important consideration, but it begins with the purpose just stated because, after all, if it didn't get us where we wanted to go in our training there would be no point to it regardless of how good it made us or our animals feel.
So while I agree that I don't want to spend a lot of time on a list that bogs down in discussions of shock collars and such, I also think we have to be open to training solutions that don't necessarily involve merely the click-and-treat. As I say, that's where the vastly overwhelming number of solutions will lie. But training is dynamic and complex, and we can only help each other if we are from time to time and case to case willing to consider all the possibilities and ramifications."
CITAAT SLUITEN.
Er zijn nog wel meer mogelijkheden om op gedrag te reageren hoor, maar wat bedoeld wordt is natuurlijk modelmatig
kwantitatief modificerende mogelijkheden.
Deze poster komt op mij bijna even warrig over als de vorige die je mij (privé) doormailde. De vier
actie kwadranten lijken mij vrij van mogelijke verwarring en ook de poster maakt er feitelijk 4 van (r, p, + en -) dus dat wordt eigenlijk wel begrepen doch heel eigenaardig verwrongen verwoord. Poster heeft gelijk dat extinctie d.m.v. negeren (= géén actie!) hier niet inbegrepen is (is ook pas veel later onderzocht en toegevoegd). Er is, denk ik terzijde, nog wel meer niet inbegrepen anders zou er geen "Skinner-paradox" zijn.
Skinner's model van de vier kwadranten is een kwalitatief model, geen kwantitatief. Het zegt dus niets over de mate waarin je de kwadranten gebruikt; Skinner zelf vond, op grond van uitgebreide proefondervindelijke uitkomsten, dat je wanneer mogelijk straffen moest vermijden.
Interessant vooral is nmm "Second, we have to remember that not all aversives are punishments,"
Zo, waarom niet dan?
"and further that something initially aversive can become a stimulus or cue for a behavior"
Nee, dat IS HET AL, nml. een stimulus ter vermijding van... (zoals in
wijken-voor-druk)
"and its initial aversive quality fades and sometimes disappears entirely."
Alweer nee, tenzij
desensibilisatie veroorzaakt wordt maar dat is de bedoeling niet want dan zou je cue niet werken!
Wat de eindconclusie betreft (het gaat toch om het verder kijken dan C/T?) denk ik dat zowel klikkeraars als anderen er vooral goed aan doen nauwkeuriger proberen waar te nemen.
Blabla! , Egon
Keep it greasey
so it'll go down easy
Hi-ho Silver!
Warboel
