Behaviorist terms. Theory, imagery, language
Unedited posts from archives of CSG-L (see INTROCSG.NET):
Date: Tue Apr 11, 1995 3:42 pm PST
Subject: Re: Extremism
[From Rick Marken (950410.2145)]
Bruce Abbott (950410.1100 EST) --
> So, Rick, how about offering a reasoned reply to my argument? What DOyou suggest we use in place of "predictive cue" or "anticipation?"
Well, as you can tell, reason is not my strong suit. But I'll give it ashot.
I suggest calling them what they are. Here is the beginning of a glossarytranslating the animistic terms of behaviorism into the scientific terms ofPCT:
Animistic term Scientific term
Predictive cue Disturbance variable
Anticipation Control of imagination
Reinforcement Controlled variable
Schedule of Reinforcement Feedback function
Discriminative stimulus Perceptual variable
Stimulus control Response to disturbance
Control by consequences Control of consequences
Feel free to add to the list. Because animistic terms are based on amagical view of the world they often have more than one scientific meaning. Forexample, a predictive cue can refer to a disturbance variable (like the targetin a tracking task) or to the value of a perception that is influenced by thatdisturbance variable (such as the rate of change in the distance between cursorand target). I used the scientific term that seemed to capture the most commonuse of the animistic term but feel free to use all relevant scientific termsfor an animistic term when completing the glossary.
> I'm really confused at the debate we seem to be having about"anticipation," as I didn't think I was saying anything controversial;
That's why I said PCT is no fun for conventional psychologists. You areconfused because you assume that PCT accounts for phenomena that conventionalpsychologists think are important. But, as Bill Powers (950410.0900 MST)pointed out, the "phenomena" of conventional psychology are contaminated bytheoretical interpretation. Even the things you think of as pure, objectivephenomena (like "anticipation") contain theoretical assumptions that are beingmade in order to avoid facing the fact that organisms control.
> Meanwhile, here's Rick, off on some tangent having little to do withthis discussion (so far as I can tell) ranting about "myths" and asserting thatphenomena like "anticipation" don't exist at all.
Thanks for providing the opportunity to present a nice, clear example ofhow you try (probably unconsciously) to make it seem like Rick is off allalone, ranting about extremes. See if you can see anything familiar in thefollowing rantings:
> So some APPARENT [emphasis mine] anticipations might arise fromcontinuous control of a relationship.
> If we want to model anticipatory behavior, let's use the theory athand and see what it can do. I think we'll find that many APPARENT EXAMPLES [ofanticipation --my emphasis again] (like a few I've mentioned above) can be handled with amodel that doesn't actually involve any anticipation at all.
Whoever made these statements seems to believe that examples of the"phenomenon" of anticipation probably don't involve anticipation (prediction ofthe future) at all. Was this said by Rick, the ranting extremist? Why no. Itwas none other than Bill Powers (950410.0900), the (closet) rantingextremist.
> Of course, I offered the counterargument that these things are not"myths," they are objective phenomena, which, as it turns out, can be nicelyexplained via PCT.
And PCT shows that they don't involve anticipation or prediction.
> Do I get a response to this argument?
I've tried to explain the PCT position on "anticipation" several times --not very well, apparently. I think my best attempt so far is in my reply thismorning (950410.0920) to Wayne Hershberger.
> How about a reply that addresses the argument?
See my post to Wayne. It's a start. The basic answer is "control ofperception".
> Predictive cues are rarely 100% reliable, and they don't always arriveat the most opportune moment. This does not prevent them from being powerfullyuseful on most occasions on which they occur.
Well, it sounds like a lot of faith is involved here. What, for example,does the organism do on those (not infrequent) occasions when the predictivecues turn out to be completely wrong? Die?
Me:
> Oh no. I thought you had become a PCTer.
Bruce:
> I'm just saying that I'm not Rick Marken.
Lucky for me;-)
> I'd guess the one that comes closest to your description [of apredictive "control" system] would be the one presented a short while ago byBill Powers, proposed as a model for classical conditioning.
I don't remember Bill's model of classical conditioning as being anythingat all like my description of predictive control. What I described was not evena control system. It was a stimulus-responsesystem that keeps the cursor on target because the stimulus (x(t)) is one ofthose "helpful" little predictive variables that happens to generate just theright responses. If x(t) goes south (as a predictor) so does tracking --and there's nothing the system can do about it.
By the way, thanks for the report on the BAAM talk. And don't bedisappointed by the turn-out.When we go to conventional psychology meetings we count it as a great victoryif we get more than two; 20 is a rock concert;-)I was a little disappointed with your description of the talk, however. Itsounds like the emphasis was on the theory rather than on the phenomenon ofcontrol. It seems like nobody was "blown away" by the theory, which is not verysurprising. Indeed, I would imagine that many in your audience were alreadyfamiliar with control theory. What they might have been less familiar with isthe nature of control as it appears in operant studies? Did you tell them howto tell whether or not an organism is controlling a particular variable in anoperant experiment? Did you explain how reinforcement is actually a controlledvariable and that it's apparent effect on behavior is an illusion? Did youexplain why conventional operant research tells us almost nothing about whatorganisms are doing (controlling)? Did you explain why attempts to controlbehavior using reinforcement are an almost sure fire way to createinterpersonal conflict? Or would these little points (facts) have been too"extreme"?
Best Rick
Date: Wed Apr 12, 1995 7:38 am PST
Subject: Re: Extremism
[From Bruce Abbott (950411.1620 EST)]
>Rick Marken (950410.2145)]
>>Bruce Abbott (950410.1100 EST) --
>> So, Rick, how about offering a reasoned reply to my argument?What DO you suggest we use in place of "predictive cue" or"anticipation?"
> I suggest calling them what they are. Here is the beginning of aglossary translating the animistic terms of behaviorism into the scientificterms of PCT:
> Animistic term Scientific term
> Predictive cue Disturbance variable
> Anticipation Control of imagination
> Reinforcement Controlled variable
> Schedule of Reinforcement Feedback function
> Discriminative stimulus Perceptual variable
> Stimulus control Response to disturbance
> Control by consequences Control of consequences
Well, that's a start. I don't follow the one for reinforcement. I suggest acloser approximation would be "reinforcER."
I also don't think "disturbance variable" quite captures the essence ofwhat I have defined as a "predictive cue," as it fails to differentiate what Ideem to be a crucial difference: predictive cues not only act as disturbancesat one level, they predict disturbances at another level. I have a similarproblem with your definition of "stimulus control," but I am aware that in boththese cases you see no need for special terms for the phenomena that definethese terms.
> Even the things you think of as pure, objective phenomena (like"anticipation") contain theoretical assumptions that are being made in order toavoid facing the fact that organisms control.
I don't think we should let that possibility interfere with scientificanalysis. Objective phenomena are objective phenomena; we are free to describeand explain them any way we please, with whatever theoretical assumptions wechoose to make.
> Thanks for providing the opportunity to present a nice, clear exampleof how you try (probably unconsciously) to make it seem like Rick is off allalone, ranting about extremes.
Sorry, but it just seems to me that your concerns are off the mark; I'mtalking about ordinary control systems and you're responding as if I'm talkingabout "feedforward" and S-Rmechanisms, which I am not. Bill seems to have made the same mistake. Also, I'mnot really concerned whether you and Bill are of like mind on these issues;what does concern me is when that position appears to be at odds with mine. Itry very hard to understand why you take the view that you do, because it islikely that I've missed something important and could learn a valuable lessonfrom the attempt. However, I don't believe that you, or Bill, or I, for thatmatter, have a lock on the "truth;" we are all fallible human beings.Therefore, my criterion for belief is not whether Bill or you agree with myposition, but whether that position makes sense to me.
> Whoever made these statements seems to believe that examples of the"phenomenon" of anticipation probably don't involve anticipation (prediction ofthe future) at all. Was this said by Rick, the ranting extremist? Why no. Itwas none other than Bill Powers (950410.0900 MST), the (closet) rantingextremist.
The problem here is that you want to define anticipation as involving someexplicit algorithm by means of which the system generates an output in responseto the "predictive cue." I don't. For me, the heart of anticipation is that thesystem begins reacting to a lower-leveldisturbance in advance of (or timed with the occurrence of) that disturbance.Describing how it achieves this miracle is the job of the explanation. Giventhese different definitions of the term, it is hardly surprising that wedisagree whether a given example involves anticipation. It turns out that ourexplanations are basically identical, which is why I've had such a hard timeunderstanding all the fuss ("ranting").
> I've tried to explain the PCT position on "anticipation" several times --not very well, apparently. I think my best attempt so far is in my reply thismorning (950410.0920) to Wayne Hershberger.
Yes, I read it and found little to disagree with, except for the questionWayne subsequently raised as to how is just happens that the disturbanceproduced by the predictive cue produces an action that tends to reduce theeffect of the disturbance to the lower-levelsystem. Seems more than coincidental to me.
>> Predictive cues are rarely 100% reliable, and they don't alwaysarrive at the most opportune moment. This does not prevent them from beingpowerfully useful on most occasions on which they occur.
> Well, it sounds like a lot of faith is involved here. What, forexample, does the organism do on those (not infrequent) occasions when thepredictive cues turn out to be completely wrong? Die?
Usually, there is an unnecessary action on the part of the lower-levelsystem to counter a disturbance that fails to materialize, these often thenproduce disturbances of their own which must be countered. As I mentionedbefore, failing to prepare for the disturbance-to-comeusually leads to more serious problems. If the fly doesn't move, then yes,death is a likely consequence.
>> I'd guess the one that comes closest to your description [of apredictive "control" system] would be the one presented a short while ago byBill Powers, proposed as a model for classical conditioning.
> I don't remember Bill's model of classical conditioning as beinganything at all like my description of predictive control. What I described wasnot even a control system. It was a stimulus-responsesystem that keeps the cursor on target because the stimulus (x(t)) is one ofthose "helpful" little predictive variables that happens to generate just theright responses. If x(t) goes south (as a predictor) so does tracking --and there's nothing the system can do about it.
Correct. And, as I never tire of saying, your model (which I believe youthought, for reasons I can only guess, was my model) is NOT my model. Imaginemy surprise when I applied Bill's model to the "anticipation" problem and foundmyself facing stiff opposition from both of you! In at least some of theanticipatory situations we've been discussing, the "predictive cue" I've beenspeaking of is nothing more (or less) than the CS in that model. Ironic, isn'tit?
Regards, Bruce
Date: Wed, 19 Apr 1995 09:04:34 -0700
Subject: Language, Models
[From Rick Marken (950419.0900)]
Bruce Abbott (950418.1515 EST) --
> I like Chuck Tucker's (950418) suggestion, copied below.
>>CHUCK TUCKER 950418
>> Let us get back to the glossary idea and preface all posts with "Imean by X ....." so such silly discussions can be avoided.
I don't like this idea at all. The notion that these "silly discussions"are based on a failure to define terms strikes me as misleading. We are usingthe language many of us have been using for over 30 years to do what many of ushave been trying to do for over 30 years --to communicate our thoughts and experiences to others. When it comes to talkingabout PCT, I have found that this ordinary language, combined with "pointing"at the behavior of working models, communicates the basic ideas of PCT justfine. There are some terms that require more careful definition --"control", for example --but there are not many of these special terms so I think we should be able totalk about PCT without having a glossary in hand.
It seems to me that linguistic "nitpicking" occurs (people saying thingslike "what, exactly, did you mean by "prediction", "reward", "information","consequence", "of", "by", "the", etc" ) when people are trying to make a non-PCTidea seems consistent with PCT --or vice versa. I think this is what is going on in the discussion of"prediction" and "anticipation". Bruce Abbott, for example, sees nothing wrongwith viewing the perception of the rate of change of a variable as an exampleof "prediction" or "anticipation" in control. Bruce said, for example:
> The use of the target velocity to predict future target position canbe considered a form of "anticipation"
which elicited the following beauty from Bill Powers:
> Gravitational acceleration can be considered a form of affinity;momentum can be considered a form of impetus ... Control of consequences bybehavior can be consider a form of control of behavior by consequences.
Why all the fuss? Because Bruce's statement evokes the wrong imagery aboutthe behavior of control system and running to the glossary won't help. Controlsystems don't "use" perceptions; they control them. Target velocity is not usedby a control system to predict the future state of anything; it is part of apresent time perception that is being controlled relative to a present timereference --the way control always works. So whether what Bruce described can be"considered a form of anticipation" is moot because control systems don't "usetarget velocity to predict future target position".
Language does matter. And it seems to me (since it worked for me) thateveryday language (sans glossary) is completely up to the task of describingthe phenomenon of control and the model thereof. If you look carefully, youwill see that the only time people want to be VERY precise about what they meanby a term is when they want to make ideas that are inconsistent with PCT seemlike they are NOT. A wonderful example of this occurred in the "sillydiscussion" of information theory. "Information" was defined and redefined inthe hopes that one could talk about it in a way that did not contradictPCT.
There are ways to talk about "information" (and "prediction" and"anticipation", etc) that ARE consistent with PCT. But this can be done withoutlooking for the "precise" definitions of these terms. It can be done byapplying the term (as ordinarily understood) appropriately to control.Perception does have information about the state of a controlled variable. Thisis just another way of saying p = f(i). Perception just doesn't have anyinformation about the cause of the state of the controlled variable. The outputof a control system sometimes does anticipate the disturbance to a controlledvariable. But this is just another way of saying that o = -dand that d(t) = f(d(t-dt).The control system itself does not operate by anticipating of predictinganything.
Best Rick